The man who made India digital isn’t done yet

Nandan Nilekani can’t stop trying to push India into the future. He started nearly 30 years ago, masterminding an ongoing experiment in technological state capacity that started with Aadhaar—the world’s largest digital identity system. Aadhaar means “foundation” in Hindi, and on that bedrock Nilekani and people working with him went on to build a sprawling collection of free, interoperating online tools that add up to nothing less than a digital infrastructure for society. They cover government services, digital payments, banking, credit, and health care, offering convenience and access that would be eye-popping in wealthy countries a tenth of India’s size. In India those systems are called, collectively, “digital public infrastructure,” or DPI.

At 70 years old, Nilekani should be retired. But he has a few more ideas. India’s electrical grid is creaky and prone to failure; Nilekani wants to add a layer of digital communication to stabilize it. And then there’s his idea to expand the financial functions in DPI to the rest of the world, creating a global digital backbone for commerce that he calls the “finternet.”

“It sounds like some crazy stuff,” Nilekani says. “But I think these are all big ideas, which over the next five years will have demonstrable, material impact.” As a last act in public life, why not Aadhaarize the world?

India’s digital backbone

Today, a farmer in a village in India, hours from the nearest bank, can collect welfare payments or transfer money by simply pressing a thumb to a fingerprint scanner at the local store. Digitally authenticated copies of driver’s licenses, birth certificates, and educational records can be accessed and shared via a digital wallet that sits on your smartphone.

In big cities, where cash is less and less common (just trying to break a bill can be a major headache), mobile payments are ubiquitous, whether you’re buying a TV from a high-street retailer or a coconut from a roadside cart. There are no fees, and any payment app or bank account can send money to any other. The country’s chaotic patchwork of public and private hospitals have begun digitizing all their medical records and uploading them to a nationwide platform. On the Open Network for Digital Commerce (ONDC), people can do online shopping searches on whatever app they want, and the results show sellers from an array of other platforms, too. The idea is to liberate small merchants and consumers from the walled gardens of online shopping giants like Amazon and the domestic giant Flipkart. 

In the most populous nation on Earth—with 1.4 billion people—a large portion of the bureaucracy anyone encounters in daily life happens seamlessly and in the cloud.

At the heart of all these tools is Aadhaar. The system gives every Indian a 12-digit number that, in combination with either a fingerprint scan or an SMS code, allows access to government services, SIM cards, basic bank accounts, digital signature services, and social welfare payments. The Indian government says that since its inception in 2009, Aadhaar has saved 3.48 trillion rupees ($39.2 billion) by boosting efficiency, bypassing corrupt officials, and cutting other types of fraud. The system is controversial and imperfect—a database with 1.4 billion people in it comes with inherent security and privacy concerns. Still, in the most populous nation on Earth, a big portion of the bureaucracy anyone might encounter in daily life just happens in the cloud.

Nilekani was behind much of that innovation, marshaling an army of civil servants, tech companies, and volunteers. Now he sees it in action every day. “It reinforces that what you have done is not some abstract stuff, but real stuff for real people,” he says.

By his own admission, Nilekani is entering the twilight of his career. But it’s not over yet. He’s now “chief mentor” for the India Energy Stack (IES), a government initiative to connect the fragmented data held by companies responsible for generating, transmitting, and distributing power. India’s grids are unstable and disparate, but Nilekani hopes an Aadhaar-like move will help. IES aims to give unique digital identities not only to power plants and energy storage facilities but even to rooftop solar panels and electric vehicles. All the data attached to those things—device characteristics, energy rating certifications, usage information—will be in a common, machine-readable format and shared on the same open protocols.

Ideally, that’ll give grid operators a real-time view of energy supply and demand. And if it works, it might also make it simpler and cheaper for anyone to connect to the grid—even everyday folks selling excess power from their rooftop solar rigs, says RS Sharma, the chair of the project and Nilekani’s deputy while building Aadhaar.

Nilekani’s other side hustle is even more ambitious. His idea for a global “finternet” combines Aadhaarization with blockchains—creating digital representations called tokens for not only financial instruments like stocks or bonds but also real-world assets like houses or jewelry. Anyone from a bank to an asset manager or even a company could create and manage these tokens, but Nilekani’s team especially hopes the idea will help poor people trade their assets, or use them as loan collateral—expanding financial services to those who otherwise couldn’t access them. 

It sounds almost wild-eyed. Yet the finternet project has 30 partners across four continents. Nilekani says it’ll launch next year.

A call to service

Nilekani was born in Bengaluru, in 1955. His family was middle class and, Nilekani says, “seized with societal issues and challenges.” His upbringing was also steeped in the kind of socialism espoused by the newish nation’s first prime minister, Jawaharlal Nehru.

After studying electrical engineering at the Indian Institute of Technology, in 1981 Nilekani helped found Infosys, an information technology company that pioneered outsourcing and helped turned India into the world’s IT back office. In 1999, he was part of a government-appointed task force trying to upgrade the infrastructure and services in Bengaluru, then emerging as India’s tech capital. But Nilekani was at the time leery of being viewed as just another techno-optimist. “I didn’t want to be seen as naive enough to believe that tech could solve everything,” he says.

Nilekani holds a device to one eye
Nilekani demonstrates the biometric technology at the heart of Aadhaar, the system he spearheaded that provides a unique digital identity number to all Indians.
PALLAVA BAGLA/CORBIS/GETTY IMAGES

Seeing the scope of the problem changed his mind—sclerotic bureaucracy, endemic corruption, and financial exclusion were intractable without technological solutions. In 2008 Nilekani published a book, Imagining India: The Idea of a Renewed Nation. It was a manifesto for an India that could leapfrog into a networked future.

And it got him a job. At the time more than half the births in the country were not recorded, and up to 400 million Indians had no official identity document. Manmohan Singh, the prime minister, asked Nilekani to put into action an ill-defined plan to create a national identity card.

Nilekani’s team made a still-controversial decision to rely on biometrics. A system based on people’s fingerprints and retina scans meant nobody could sign up twice, and nobody had to carry paperwork. In terms of execution, it was like trying to achieve industrialization but skip a steam era. Deployment required a monumental data collection effort, as well as new infrastructure that could compare each new enrollment against hundreds of millions of existing records in seconds. At its peak, the Unique Identification Authority of India (UIDAI), the agency responsible for administering Aadhaar, was registering more than a million new users a day. That happened with a technical team of just about 50 developers, and in the end cost slightly less than half a billion dollars.

Buoyed by their success, Nilekani and his allies started casting around for other problems they could solve using the same digitize-the-real-world playbook. “We built more and more layers of capability,” Nilekani says, “and then this became a wider-ranging idea. More grandiose.”

While other countries were building digital backbones with full state control (as in China) or in public-private partnerships that favored profit-seeking corporate approaches (as in the US), Nilekani thought India needed something else. He wanted critical technologies in areas like identity, payments, and data sharing to be open and interoperable, not monopolized by either the state or private industry. So the tools that make up DPI use open standards and open APIs, meaning that anyone can plug into the system. No single company or institution controls access—no walled gardens.

A contested legacy

Of course, another way to look at putting financial and government services and records into giant databases is that it’s a massive risk to personal liberty. Aadhaar, in particular, has faced criticism from privacy advocates concerned about the potential for surveillance. Several high-profile data breaches of Aadhaar records held by government entities have shaken confidence in the system, most recently in 2023, when security researchers found hackers selling the records of more than 800 million Indians on the dark web.

Technically, this shouldn’t matter—an Aadhaar number ought to be useless without biometric or SMS-based authentication. It’s “a myth that this random number is a very powerful number,” says Sharma, the onetime co-lead of UIDAI. “I don’t have any example where somebody’s Aadhaar disclosure would have harmed somebody.” 

One problem is that in everyday use, Aadhaar users often bypass the biometric authentication system. To ensure that people use a genuine address at registration, Aadhaar administrators give people their numbers on an official-looking document. Indians co-opted this paperwork as a proof of identity on its own. And since the document—Indians even call it an “Aadhaar card”—doesn’t have an expiration date, it’s possible for people to get multiple valid cards with different details by changing their address or date of birth. That’s quite a loophole. In 2018 an NGO report found that 67% of people using Aadhaar to open a bank account relied on this verification document rather than digital authentication. That report was the last time anyone published data on the problem, so nobody knows how bad it is today. “Everybody’s living on anecdotes,” says Kiran Jonnalagadda, an anti-Aadhaar activist.

In other cases, flaws in Aadhaar’s biometric technology have caused people to be denied essential government services. The government downplays these risks, but again, it’s impossible to tell how serious the problem is because the UIDAI won’t disclose numbers. “There needs to be a much more honest acknowledgment, documentation, and then an examination of how those exclusions can be mitigated,” says Apar Gupta, director of the Internet Freedom Foundation.

Beyond the potential for fraud, it’s also true that the free and interoperable tools haven’t reached all the people who might find them useful, especially among India’s rural and poorer populations. Nilekani’s hopes for openness haven’t fully come to pass. Big e-commerce companies still dominate, and retail sales on ONDC have been dropping steadily since 2024, when financial incentives to participate began to taper off. The digital payments and government documentation services have hundreds of millions of users, numbers most global technology companies would love to see—but in a country as large as India, that leaves a lot of people out.

Going global

The usually calm Nilekani bristles at that criticism; he has heard it before. Detractors overlook the dysfunction that preceded these efforts, he says, and he remains convinced that technology was the only way forward. “How do you move a country of 1.4 billion people?” he asks. “There’s no other way you can fix it.”

The proof is self-evident, he says. Indians have opened more than 500 million basic bank accounts using Aadhaar; before it came into use, millions of those people had been completely unbanked. Earlier this year, India’s Unified Payments Interface overtook Visa as the world’s largest real-time payments system. “There is no way Aadhaar could have worked but for the fact that people needed this thing,” Nilekani says. “There’s no way payments would have worked without people needing it. So the voice of the people—they’re voting with their feet.”

A street vendor in Kolkata displays a QR code that lets him get paid via India’s Unified Payments Interface, part of the digital public infrastructure Nilekani helped build. The Reserve Bank of India says more than 657 million people used the system in the financial year 2024–2025.
DEBAJYOTI CHAKRABORTY/NURPHOTO/GETTY IMAGES

That need might be present in countries beyond India. “Many countries don’t have a proper birth registration system. Many countries don’t have a payment system. Many countries don’t have a way for data to be leveraged,” Nilekani says. “So this is a very powerful idea.” It seems to be spreading. Foreign governments regularly send delegations to Bengaluru to study India’s DPI tools. The World Bank and the United Nations have tried to introduce the concept to other developing countries equally eager to bring their economies into the digital age. The Gates Foundation has established projects to promote digital infrastructure, and Nilekani has set up and funded a network of think tanks, research institutes, and other NGOs aimed at, as he says, “propagating the gospel.”

Still, he admits he might not live to see DPI go global. “There are two races,” Nilekani says. “My personal race against time and India’s race against time.” He worries that the economic potential of its vast young population—the so-called demographic dividend—could turn into a demographic disaster. Despite rapid growth, gains have been uneven. Youth unemployment remains stubbornly high—a particularly volatile problem in a large and economically turbulent country. 

“Maybe I’m a junkie,” he says. “Why the hell am I doing all this? I think I need it. I think I need to keep curious and alive and looking at the future.” But that’s the thing about building the future: It never quite arrives.

Edd Gent is a journalist based in Bengaluru, India.

LLMs contain a LOT of parameters. But what’s a parameter?

MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.

I am writing this because one of my editors woke up in the middle of the night and scribbled on a bedside notepad: “What is a parameter?” Unlike a lot of thoughts that hit at 4 a.m., it’s a really good question—one that goes right to the heart of how large language models work. And I’m not just saying that because he’s my boss. (Hi, Boss!)

A large language model’s parameters are often said to be the dials and levers that control how it behaves. Think of a planet-size pinball machine that sends its balls pinging from one end to the other via billions of paddles and bumpers set just so. Tweak those settings and the balls will behave in a different way.  

OpenAI’s GPT-3, released in 2020, had 175 billion parameters. Google DeepMind’s latest LLM, Gemini 3, may have at least a trillion—some think it’s probably more like 7 trillion—but the company isn’t saying. (With competition now fierce, AI firms no longer share information about how their models are built.)

But the basics of what parameters are and how they make LLMs do the remarkable things that they do are the same across different models. Ever wondered what makes an LLM really tick—what’s behind the colorful pinball-machine metaphors? Let’s dive in.  

What is a parameter?

Think back to middle school algebra, like 2a + b. Those letters are parameters: Assign them values and you get a result. In math or coding, parameters are used to set limits or determine output. The parameters inside LLMs work in a similar way, just on a mind-boggling scale. 

How are they assigned their values?

Short answer: an algorithm. When a model is trained, each parameter is set to a random value. The training process then involves an iterative series of calculations (known as training steps) that update those values. In the early stages of training, a model will make errors. The training algorithm looks at each error and goes back through the model, tweaking the value of each of the model’s many parameters so that next time that error is smaller. This happens over and over again until the model behaves in the way its makers want it to. At that point, training stops and the values of the model’s parameters are fixed.

Sounds straightforward …

In theory! In practice, because LLMs are trained on so much data and contain so many parameters, training them requires a huge number of steps and an eye-watering amount of computation. During training, the 175 billion parameters inside a medium-size LLM like GPT-3 will each get updated tens of thousands of times. In total, that adds up to quadrillions (a number with 15 zeros) of individual calculations. That’s why training an LLM takes so much energy. We’re talking about thousands of specialized high-speed computers running nonstop for months.

Oof. What are all these parameters for, exactly?

There are three different types of parameters inside an LLM that get their values assigned through training: embeddings, weights, and biases. Let’s take each of those in turn.

Okay! So, what are embeddings?

An embedding is the mathematical representation of a word (or part of a word, known as a token) in an LLM’s vocabulary. An LLM’s vocabulary, which might contain up to a few hundred thousand unique tokens, is set by its designers before training starts. But there’s no meaning attached to those words. That comes during training.  

When a model is trained, each word in its vocabulary is assigned a numerical value that captures the meaning of that word in relation to all the other words, based on how the word appears in countless examples across the model’s training data.

Each word gets replaced by a kind of code?

Yeah. But there’s a bit more to it. The numerical value—the embedding—that represents each word is in fact a list of numbers, with each number in the list representing a different facet of meaning that the model has extracted from its training data. The length of this list of numbers is another thing that LLM designers can specify before an LLM is trained. A common size is 4,096.

Every word inside an LLM is represented by a list of 4,096 numbers?  

Yup, that’s an embedding. And each of those numbers is tweaked during training. An LLM with embeddings that are 4,096 numbers long is said to have 4,096 dimensions.

Why 4,096?

It might look like a strange number. But LLMs (like anything that runs on a computer chip) work best with powers of two—2, 4, 8, 16, 32, 64, and so on. LLM engineers have found that 4,096 is a power of two that hits a sweet spot between capability and efficiency. Models with fewer dimensions are less capable; models with more dimensions are too expensive or slow to train and run. 

Using more numbers allows the LLM to capture very fine-grained information about how a word is used in many different contexts, what subtle connotations it might have, how it relates to other words, and so on.

Back in February, OpenAI released GPT-4.5, the firm’s largest LLM yet (some estimates have put its parameter count at more than 10 trillion). Nick Ryder, a research scientist at OpenAI who worked on the model, told me at the time that bigger models can work with extra information, like emotional cues, such as when a speaker’s words signal hostility: “All of these subtle patterns that come through a human conversation—those are the bits that these larger and larger models will pick up on.”

The upshot is that all the words inside an LLM get encoded into a high-dimensional space. Picture thousands of words floating in the air around you. Words that are closer together have similar meanings. For example, “table” and “chair” will be closer to each other than they are to “astronaut,” which is close to “moon” and “Musk.” Way off in the distance you can see “prestidigitation.” It’s a little like that, but instead of being related to each other across three dimensions, the words inside an LLM are related across 4,096 dimensions.

Yikes.

It’s dizzying stuff. In effect, an LLM compresses the entire internet into a single monumental mathematical structure that encodes an unfathomable amount of interconnected information. It’s both why LLMs can do astonishing things and why they’re impossible to fully understand.    

Okay. So that’s embeddings. What about weights?

A weight is a parameter that represents the strength of a connection between different parts of a model—and one of the most common types of dial for tuning a model’s behavior. Weights are used when an LLM processes text.

When an LLM reads a sentence (or a book chapter), it first looks up the embeddings for all the words and then passes those embeddings through a series of neural networks, known as transformers, that are designed to process sequences of data (like text) all at once. Every word in the sentence gets processed in relation to every other word.

This is where weights come in. An embedding represents the meaning of a word without context. When a word appears in a specific sentence, transformers use weights to process the meaning of that word in that new context. (In practice, this involves multiplying each embedding by the weights for all other words.)

And biases?

Biases are another type of dial that complement the effects of the weights. Weights set the thresholds at which different parts of a model fire (and thus pass data on to the next part). Biases are used to adjust those thresholds so that an embedding can trigger activity even when its value is low. (Biases are values that are added to an embedding rather than multiplied with it.) 

By shifting the thresholds at which parts of a model fire, biases allow the model to pick up information that might otherwise be missed. Imagine you’re trying to hear what somebody is saying in a noisy room. Weights would amplify the loudest voices the most; biases are like a knob on a listening device that pushes quieter voices up in the mix. 

Here’s the TL;DR: Weights and biases are two different ways that an LLM extracts as much information as it can out of the text it is given. And both types of parameters are adjusted over and over again during training to make sure they do this. 

Okay. What about neurons? Are they a type of parameter too? 

No, neurons are more a way to organize all this math—containers for the weights and biases, strung together by a web of pathways between them. It’s all very loosely inspired by biological neurons inside animal brains, with signals from one neuron triggering new signals from the next and so on. 

Each neuron in a model holds a single bias and weights for every one of the model’s dimensions. In other words, if a model has 4,096 dimensions—and therefore its embeddings are lists of 4,096 numbers—then each of the neurons in that model will hold one bias and 4,096 weights. 

Neurons are arranged in layers. In most LLMs, each neuron in one layer is connected to every neuron in the layer above. A 175-billion-parameter model like GPT-3 might have around 100 layers with a few tens of thousands of neurons in each layer. And each neuron is running tens of thousands of computations at a time. 

Dizzy again. That’s a lot of math.

That’s a lot of math.

And how does all of that fit together? How does an LLM take a bunch of words and decide what words to give back?

When an LLM processes a piece of text, the numerical representation of that text—the embedding—gets passed through multiple layers of the model. In each layer, the value of the embedding (that list of 4,096 numbers) gets updated many times by a series of computations involving the model’s weights and biases (attached to the neurons) until it gets to the final layer.

The idea is that all the meaning and nuance and context of that input text is captured by the final value of the embedding after it has gone through a mind-boggling series of computations. That value is then used to calculate the next word that the LLM should spit out. 

It won’t be a surprise that this is more complicated than it sounds: The model in fact calculates, for every word in its vocabulary, how likely that word is to come next and ranks the results. It then picks the top word. (Kind of. See below …) 

That word is appended to the previous block of text, and the whole process repeats until the LLM calculates that the most likely next word to spit out is one that signals the end of its output. 

That’s it?  

Sure. Well …

Go on.

LLM designers can also specify a handful of other parameters, known as hyperparameters. The main ones are called temperature, top-p, and top-k.

You’re making this up.

Temperature is a parameter that acts as a kind of creativity dial. It influences the model’s choice of what word comes next. I just said that the model ranks the words in its vocabulary and picks the top one. But the temperature parameter can be used to push the model to choose the most probable next word, making its output more factual and relevant, or a less probable word, making the output more surprising and less robotic. 

Top-p and top-k are two more dials that control the model’s choice of next words. They are settings that force the model to pick a word at random from a pool of most probable words instead of the top word. These parameters affect how the model comes across—quirky and creative versus trustworthy and dull.   

One last question! There has been a lot of buzz about small models that can outperform big models. How does a small model do more with fewer parameters?

That’s one of the hottest questions in AI right now. There are a lot of different ways it can happen. Researchers have found that the amount of training data makes a huge difference. First you need to make sure the model sees enough data: An LLM trained on too little text won’t make the most of all its parameters, and a smaller model trained on the same amount of data could outperform it. 

Another trick researchers have hit on is overtraining. Showing models far more data than previously thought necessary seems to make them perform better. The result is that a small model trained on a lot of data can outperform a larger model trained on less data. Take Meta’s Llama LLMs. The 70-billion-parameter Llama 2 was trained on around 2 trillion words of text; the 8-billion-parameter Llama 3 was trained on around 15 trillion words of text. The far smaller Llama 3 is the better model. 

A third technique, known as distillation, uses a larger model to train a smaller one. The smaller model is trained not only on the raw training data but also on the outputs of the larger model’s internal computations. The idea is that the hard-won lessons encoded in the parameters of the larger model trickle down into the parameters of the smaller model, giving it a boost. 

In fact, the days of single monolithic models may be over. Even the largest models on the market, like OpenAI’s GPT-5 and Google DeepMind’s Gemini 3, can be thought of as several small models in a trench coat. Using a technique called “mixture of experts,” large models can turn on just the parts of themselves (the “experts”) that are required to process a specific piece of text. This combines the abilities of a large model with the speed and lower power consumption of a small one.

But that’s not the end of it. Researchers are still figuring out ways to get the most out of a model’s parameters. As the gains from straight-up scaling tail off, jacking up the number of parameters no longer seems to make the difference it once did. It’s not so much how many you have, but what you do with them.

Can I see one?

You want to see a parameter? Knock yourself out: Here’s an embedding.

hello

The Download: war in Europe, and the company that wants to cool the planet

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

Europe’s drone-filled vision for the future of war

Last spring, 3,000 British soldiers deployed an invisible automated intelligence network, known as a “digital targeting web,” as part of a NATO exercise called Hedgehog in the damp forests of Estonia’s eastern territories.

The system had been cobbled together over the course of four months—an astonishing pace for weapons development, which is usually measured in years. Its purpose is to connect everything that looks for targets—“sensors,” in military lingo—and everything that fires on them (“shooters”) to a single, shared wireless electronic brain.

Eighty years after total war last transformed the continent, the Hedgehog tests signal a brutal new calculus of European defense. But leaning too much on this new mathematics of warfare could be a risky bet. Read the full story.

—Arthur Holland Michel

This story is from the next print issue of MIT Technology Review magazine. If you haven’t already, subscribe now to receive it once it lands.

MIT Technology Review Narrated: How one controversial startup hopes to cool the planet

Stardust Solutions believes that it can solve climate change—for a price.

The Israel-based geoengineering startup has said it expects nations will soon pay it more than a billion dollars a year to launch specially equipped aircraft into the stratosphere. Once they’ve reached the necessary altitude, those planes will disperse particles engineered to reflect away enough sunlight to cool down the planet, purportedly without causing environmental side effects. 

But numerous solar geoengineering researchers are skeptical that Stardust will line up the customers it needs to carry out a global deployment in the next decade. They’re also highly critical of the idea of a private company setting the global temperature for us.

This is our latest story to be turned into a MIT Technology Review Narrated podcast, which we’re publishing each week on Spotify and Apple Podcasts. Just navigate to MIT Technology Review Narrated on either platform, and follow us to get all our new content as it’s released.

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 Amazon has been accused of listing products without retailers’ consent
Small shop owners claim Amazon’s AI tool sold their goods without their permission. (Bloomberg $)
+ It also listed products the shops didn’t actually have in stock. (CNBC)
+ A new feature called “Shop Direct” appears to be to blame. (Insider $)

2 Data centers are a political issue 
Opposition to them is uniting communities across the political divide. (WP $)
+ Power-grid operators have suggested the centers power down at certain times. (WSJ $)
+ The data center boom in the desert. (MIT Technology Review)

3 Things are looking up for the nuclear power industry
The Trump administration is pumping money into it—but success is not guaranteed. (NYT $)
+ Why the grid relies on nuclear reactors in the winter. (MIT Technology Review)

4 A new form of climate modelling pins blame on specific companies
It may not be too long until we see the first case of how attribution science holds up in court. (New Scientist $)
+ Google, Amazon and the problem with Big Tech’s climate claims. (MIT Technology Review)

5 Meta has paused the launch of its Ray-Ban smartglasses 🕶
They’re just too darn popular, apparently. (Engadget)
+ Europe and Canada will just have to wait. (Gizmodo)
+ It’s blaming supply shortages and “unprecedented” demand. (Insider $)

6 Sperm contains information about a father’s fitness and diet
New research is shedding light on how we think about heredity. (Quanta Magazine)

7 Meta is selling online gambling ads in countries where it’s illegal
It’s ignoring local laws across Asia and the Middle East. (Rest of World)

8 AI isn’t always trying to steal your job
Sometimes it makes your toy robot a better companion. (The Verge)
+ How cuddly robots could change dementia care. (MIT Technology Review)

9 How to lock down a job at one of tech’s biggest companies
You’re more likely to be accepted into Harvard, apparently. (Fast Company $)

10 Millennials are falling out of love with the internet
Is a better future still possible? (Vox)
+ How to fix the internet. (MIT Technology Review)

Quote of the day

“I want to keep up with the latest doom.”

—Author Margaret Atwood explains why she doomscrolls to Wired.

One more thing

Inside the decades-long fight over Yahoo’s misdeeds in China

When you think of Big Tech these days, Yahoo is probably not top of mind. But for Chinese dissident Xu Wanping, the company still looms large—and has for nearly two decades.

In 2005, Xu was arrested for signing online petitions relating to anti-Japanese protests. He didn’t use his real name, but he did use his Yahoo email address. Yahoo China violated its users’ trust—providing information on certain email accounts to Chinese law enforcement, which in turn allowed the government to identify and arrest some users.

Xu was one of them; he would serve nine years in prison. Now, he and five other Chinese former political prisoners are suing Yahoo and a slate of co-defendants—not because of the company’s information-sharing (which was the focus of an earlier lawsuit filed by other plaintiffs), but rather because of what came after. Read the full story.

—Eileen Guo

We can still have nice things

A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)

+ It’s time to celebrate the life and legacy of Cecilia Giménez Zueco, the legendary Spanish amateur painter whose botched fresco restoration reached viral fame in 2012.
+ If you’re a sci-fi literature fan, there’s plenty of new releases to look forward to in 2026.
+ Last week’s wolf supermoon was a sight to behold.
+ This Mississippi restaurant is putting its giant lazy Susan to good use.

Deploying a hybrid approach to Web3 in the AI era

When the concept of “Web 3.0” first emerged about a decade ago the idea was clear: Create a more user-controlled internet that lets you do everything you can now, except without servers or intermediaries to manage the flow of information.

Where Web2, which emerged in the early 2000s, relies on centralized systems to store data and supply compute, all owned—and monetized by—a handful of global conglomerates, Web3 turns that structure on its head. Instead, data and compute are decentralized through technologies like blockchain and peer-to-peer networks.

What was once a futuristic concept is quickly becoming a more concrete reality, even at a time when Web2 still dominates. Six out of ten Fortune 500 companies are exploring blockchain-based solutions, most taking a hybrid approach that combines traditional Web2 business models and infrastructure with the decentralized technologies and principles of Web3.

Popular use cases include cloud services, supply chain management, and, most notably financial services. In fact, at one point, the daily volume of transactions processed on decentralized finance exchanges exceeded $10 billion.

Gaining a Web3 edge

Among the advantages of Web3 for the enterprise are greater ownership and control of sensitive data, says Erman Tjiputra, founder and CEO of the AIOZ Network, which is building infrastructure for Web3, powered by decentralized physical infrastructure networks (DePIN), blockchain-based systems that govern physical infrastructure assets.

More cost-effective compute is another benefit, as is enhanced security and privacy as the cyberattack landscape grows more hostile, he adds. And it could even help protect companies from outages caused by a single point of failure, which can lead to downtime, data loss, and revenue deficits.

But perhaps the most exciting opportunity, says Tjiputra, is the ability to build and scale AI reliably and affordably. By leveraging a people-powered internet infrastructure, companies can far more easily access—and contribute to—shared resource like bandwidth, storage, and processing power to run AI inference, train models, and store data. All while using familiar developer tooling and open, usage-based incentives.

“We’re in a compute crunch where requirements are insatiable, and Web3 creates this ability to benefit while contributing,” explains Tjiputra.

In 2025, AIOZ Network launched a distributed compute platform and marketplace where developers and enterprises can access and monetize AI assets, and run AI inference or training on AIOZ Network’s more than 300,000 contributing devices. The model allows companies to move away from opaque datasets and models and scale flexibly, without centralized lock in.

Overcoming Web3 deployment challenges

Despite the promise, it is still early days for Web3, and core systemic challenges are leaving senior leadership and developers hesitant about its applicability at scale.

One hurdle is a lack of interoperability. The current fragmentation of blockchain networks creates a segregated ecosystem that makes it challenging to transfer assets or data between platforms. This often complicates transactions and introduces new security risks due to the reliance on mechanisms such as cross-chain bridges. These are tools that allow asset transfers between platforms but which have been shown to be vulnerable to targeted attacks.

“We have countless blockchains running on different protocols and consensus models,” says Tjiputra. “These blockchains need to work with each other so applications can communicate regardless of which chain they are on. This makes interoperability fundamental.”

Regulatory uncertainty is also a challenge. Outdated legal frameworks can sit at odds with decentralized infrastructures, especially when it comes to compliance with data protection and anti-money laundering regulations.

“Enterprises care about verifiability and compliance as much as innovation, so we need frameworks where on-chain transparency strengthens accountability instead of adding friction,” Tjiputra says.

And this is compounded by user experience (UX) challenges, says Tjiputra. “The biggest setback in Web3 today is UX,” he says. “For example, in Web2, if I forget my bank username or password, I can still contact the bank, log in and access my assets. The trade-off in Web3 is that, should that key be compromised or lost, we lose access to those assets. So, key recovery is a real problem.”

Building a bridge to Web3

Although such systemic challenges won’t be solved overnight, by leveraging DePIN networks, enterprises can bridge the gap between Web2 and Web3, without making a wholesale switch. This can minimize risk while harnessing much of the potential.

AIOZ Network’s own ecosystem includes capacity for media streaming, AI compute, and distributed storage that can be plugged into an existing Web2 tech stack. “You don’t need to go full Web3,” says Tjiputra. “You can start by plugging distributed storage into your workflow, test it, measure it, and see the benefits firsthand.”

The AIOZ Storage solution, for example, offers scalable distributed object storage by leveraging the global network of contributor devices on AIOZ DePIN. It is also compatible with existing storage systems or commonly used web application programming interfaces (APIs).

“Say we have a programmer or developer who uses Amazon S3 Storage or REST APIs, then all they need to do is just repoint the endpoints,” explains Tjiputra. “That’s it. It’s the same tools, it’s really simple. Even with media, with a single one-stop shop, developers can do transcoding and streaming with a simple REST API.”

Built on Cosmos, a network of hundreds of different blockchains that can communicate with each other, and a standardized framework enabled by Ethereum Virtual Machine (EVM), AIOZ Network has also prioritized interoperability. “Applications shouldn’t care which chain they’re on. Developers should target APIs without worrying about consensus mechanisms. That’s why we built on Cosmos and EVM—interoperability first.”

This hybrid model, which allows enterprises to use both Web2 and Web3 advantages in tandem, underpins what Tjiputra sees as the longer-term ambition for the much-hyped next iteration of the internet.

“Our vision is a truly peer-to-peer foundation for a people-powered internet, one that minimizes single points of failure through multi-region, multi-operator design,” says Tjiputra. “By distributing compute and storage across contributors, we gain both cost efficiency and end-to-end security by default.

“Ideally, we want to evolve the internet toward a more people-powered model, but we’re not there yet. We’re still at the starting point and growing.”

Indeed, Web3 isn’t quite snapping at the heels of the world’s Web2 giants, but its commercial advantages in an era of AI have become much harder to ignore. And with DePIN bridging the gap, enterprises and developers can step into that potential while keeping one foot on surer ground.

To learn more from AIOZ Network, you can read the AIOZ Network Vision Paper.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

This content was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.

New Ecommerce Tools: January 7, 2026

This week’s installment of new products and services for merchants includes marketing and advertising platforms, livestream tools, pop-up and form builders, fulfillment networks, AI voice agents, agentic commerce, and reverse logistics.

Got an ecommerce product release? Email updates@practicalecommerce.com.

New Tools for Merchants

Orca launches LiveMax to book a shoppable livestream in minutes. Orca, a livestream and social commerce provider, has launched LiveMax, a self-serve tool that empowers brands and retailers to book and execute shoppable livestreams on TikTok Shop and Amazon Live. According to Orca, LiveMax enables any brand to schedule a produced livestream quickly. Orca’s production resources include professional hosts and producers.

Home page of Orca

Orca

PayPal Ads launches Transaction Graph Insights and Measurement. PayPal Ads has launched its Transaction Graph Insights and Measurement Program, providing merchants and advertisers with a view into shopper behavior, campaign effectiveness, and data-driven recommendations. The tools help understand cross-merchant, cross-surface shopper journeys to deliver brand-specific recommendations and independent campaign validation with third-party partners.

Slingwave brings AI-powered unified measurement to ecommerce.  Slingwave has unveiled its AI-native marketing platform for ecommerce and direct-to-consumer brands. The system combines marketing mix modeling, agile marketing attribution, and experimentation with an intelligence layer and customized models that run millions of scenarios to deliver a clear plan for optimizing spend. According to Slingwave, the platform learns with every campaign, ensuring recommendations continuously improve.

Getsitecontrol updates widget builder for pop-ups and forms. Getsitecontrol, an email marketing platform for ecommerce, has released a redesigned widget editor that offers enhanced visual control when designing website pop-ups, forms, and teasers. The editor introduces a visual element tree that displays the complete structure of each widget in a sidebar. Getsitecontrol now allows users to fine-tune every visual aspect of their widgets, including margins, paddings, alignment, sizes, and colors. The result, says Getsitecontrol, is professional widgets that adapt to any screen size.

ReturnPro launches Shopify app. ReturnPro, a provider of returns management and reverse logistics, has launched its Returns Portal App on the Shopify App Store. The app combines returns initiation with a connected reverse supply chain and recommerce ecosystem. Shopify merchants gain access to ReturnPro’s infrastructure, including more than 1,000 partner drop-off locations. Merchants can resell refurbished inventory through their Shopify storefronts or distribute products across ReturnPro’s network of integrated marketplaces, creating secondary revenue streams and reducing write-offs.

Home page of ReturnPro

ReturnPro

Stord acquires Shipwire to expand its fulfillment network. Stord, a logistics provider for pre-purchase, checkout, delivery, and returns, has acquired Shipwire, a subsidiary of Ceva Logistics. Stord says the acquisition continues its expansion of fulfillment networks by adding 12 locations, strengthening its presence in Europe, and maintaining access to Ceva’s global network of warehouses through Shipwire’s existing logistics agreements. Ceva manages 120 million square feet of warehouse space worldwide.

Amazon launches Alexa+ for users to chat with its assistant. Amazon has launched an Alexa+ website that lets select users chat with its assistant via their browser. Users can access Alexa.com to get quick answers, explore complex topics, create content, and more. Alexa.com combines information with real-world actions, offering integrations across devices for shopping, home control, cooking, and entertainment, per Amazon. Customers with early access to Alexa+ can visit Alexa.com while logged into their Amazon account and start chatting.

ITTRackNap launches marketplace and subscription commerce platform. ITTRackNap, an AI-powered marketplace and subscription automation platform for cloud and technology providers, announced its U.S. launch. The platform enables managed service providers, telecommunications and connectivity providers, and technology distributors to launch and scale cloud and digital commerce faster and cost-effectively. RackNap streamlines and lowers the cost of channel back-office operations through native integrations with hyperscalers and portals, including Microsoft, Amazon Web Services, Google, and Acronis.

PubMatic launches AgenticOS for agent-to-agent advertising. PubMatic, an ad tech company, has launched AgenticOS, an operating system to orchestrate autonomous, agent-to-agent advertising across digital environments. AgenticOS deploys a three-layer framework to plan, transact, and optimize programmatic advertising: (i) an Nvidia-powered infrastructure layer, (ii) an application layer with embedded agentic capabilities to interpret intent through protocols such as the Ad Context and Model Context, and (iii) a transaction layer that connects agentic decisioning to PubMatic’s Activate buying platform.

Home page of PubMatic

PubMatic

eBay introduces credit notes for U.S. seller fees and tax reversals. eBay is issuing separate credit notes for all seller fees, charges, and tax reversals in the U.S. A credit note reduces or cancels an invoice. Each credit note will show the reduced amounts and reference to the original invoice. According to eBay, the update improves transparency and helps match charges with reversals.

Cloudhands launches cross-model AI platform. Cloudhands, a marketplace for AI tools, has announced a new unified platform that lets users move among leading models such as OpenAI, Anthropic, and Google while keeping their conversation history, documents, tasks, and creative work connected. Interested users can join the waitlist for the platform, which will launch early this year, per Cloudhands.

xAI launches Grok Business and Grok Enterprise. xAI, the chatbot natively integrated into X, has launched Grok Business and Grok Enterprise, two new tiers providing access to Grok 3, Grok 4, and Grok 4 Heavy. Grok Business offers a self-serve process for small-to-medium teams. For larger organizations, Grok Enterprise includes Grok Business plus Custom Single Sign-On, Directory Sync, and audit and security controls.

VoAgents launches enterprise voice AI platform for customer conversations. VoAgents, a provider of enterprise voice tools, has launched voice AI agents capable of handling inbound and outbound calls. The platform’s self-learning capability means voice agents improve with every interaction. Core platform features include customizable voice personalities and workflows tailored to brand requirements, calendar and customer-management integrations, real-time call recordings and transcripts, outbound campaign management, and more. VoAgents offers access to all leading language models, including OpenAI and Anthropic.

Home page of VoAgents

VoAgents

Most Major News Publishers Block AI Training & Retrieval Bots via @sejournal, @MattGSouthern

Most top news publishers block AI training bots via robots.txt, but they’re also blocking the retrieval bots that determine whether sites appear in AI-generated answers.

BuzzStream analyzed the robots.txt files of 100 top news sites across the US and UK and found 79% block at least one training bot. More notably, 71% also block at least one retrieval or live search bot.

Training bots gather content to build AI models, while retrieval bots fetch content in real time when users ask questions. Sites blocking retrieval bots may not appear when AI tools try to cite sources, even if the underlying model was trained on their content.

What The Data Shows

BuzzStream examined the top 50 news sites in each market based on SimilarWeb traffic share, then deduplicated the list. The study grouped bots into three categories: training, retrieval/live search, and indexing.

Training Bot Blocks

Among training bots, Common Crawl’s CCBot was the most frequently blocked at 75%, followed by Anthropic-ai at 72%, ClaudeBot at 69%, and GPTBot at 62%.

Google-Extended, which trains Gemini, was the least blocked training bot at 46% overall. US publishers blocked it at 58%, nearly double the 29% rate among UK publishers.

Harry Clarkson-Bennett, SEO Director at The Telegraph, told BuzzStream:

“Publishers are blocking AI bots using the robots.txt because there’s almost no value exchange. LLMs are not designed to send referral traffic and publishers (still!) need traffic to survive.”

Retrieval Bot Blocks

The study found 71% of sites block at least one retrieval or live search bot.

Claude-Web was blocked by 66% of sites, while OpenAI’s OAI-SearchBot, which powers ChatGPT’s live search, was blocked by 49%. ChatGPT-User was blocked by 40%.

Perplexity-User, which handles user-initiated retrieval requests, was the least blocked at 17%.

Indexing Blocks

PerplexityBot, which Perplexity uses to index pages for its search corpus, was blocked by 67% of sites.

Only 14% of sites blocked all AI bots tracked in the study, while 18% blocked none.

The Enforcement Gap

The study acknowledges that robots.txt is a directive, not a barrier, and bots can ignore it.

We covered this enforcement gap when Google’s Gary Illyes confirmed robots.txt can’t prevent unauthorized access. It functions more like a “please keep out” sign than a locked door.

Clarkson-Bennett raised the same point in BuzzStream’s report:

“The robots.txt file is a directive. It’s like a sign that says please keep out, but doesn’t stop a disobedient or maliciously wired robot. Lots of them flagrantly ignore these directives.”

Cloudflare documented that Perplexity used stealth crawling behavior to bypass robots.txt restrictions. The company rotated IP addresses, changed ASNs, and spoofed its user agent to appear as a browser.

Cloudflare delisted Perplexity as a verified bot and now actively blocks it. Perplexity disputed Cloudflare’s claims and published a response.

For publishers serious about blocking AI crawlers, CDN-level blocking or bot fingerprinting may be necessary beyond robots.txt directives.

Why This Matters

The retrieval-blocking numbers warrant attention here. In addition to opting out of AI training, many publishers are opting out of the citation and discovery layer that AI search tools use to surface sources.

OpenAI separates its crawlers by function: GPTBot gathers training data, while OAI-SearchBot powers live search in ChatGPT. Blocking one doesn’t block the other. Perplexity makes a similar distinction between PerplexityBot for indexing and Perplexity-User for retrieval.

These blocking choices affect where AI tools can pull citations from. If a site blocks retrieval bots, it may not appear when users ask AI assistants for sourced answers, even if the model already contains that site’s content from training.

The Google-Extended pattern is worth watching. US publishers block it at nearly twice the UK rate, though whether that reflects different risk calculations around Gemini’s growth or different business relationships with Google isn’t clear from the data.

Looking Ahead

The robots.txt method has limits, and sites that want to block AI crawlers may find CDN-level restrictions more effective than robots.txt alone.

Cloudflare’s Year in Review found GPTBot, ClaudeBot, and CCBot had the highest number of full disallow directives across top domains. The report also noted that most publishers use partial blocks for Googlebot and Bingbot rather than full blocks, reflecting the dual role Google’s crawler plays in search indexing and AI training.

For those tracking AI visibility, the retrieval bot category is what to watch. Training blocks affect future models, while retrieval blocks affect whether your content shows up in AI answers right now.


Featured Image: Kitinut Jinapuck/Shutterstock

Google’s Mueller Weighs In On SEO vs GEO Debate via @sejournal, @MattGSouthern

Google Search Advocate John Mueller says businesses that rely on referral traffic should think about how AI tools fit into the picture.

Mueller responded to a Reddit thread asking whether SEO is still enough or whether practitioners need to start considering GEO, a term some in the industry use for optimizing visibility in AI-powered answer engines like ChatGPT, Gemini, and Perplexity.

“If you have an online business that makes money from referred traffic, it’s definitely a good idea to consider the full picture, and prioritize accordingly,” Mueller wrote.

What Mueller Said

Mueller didn’t endorse or reject the GEO terminology. He framed the question in terms of practical business decisions rather than new optimization techniques.

“What you call it doesn’t matter, but ‘AI’ is not going away, but thinking about how your site’s value works in a world where ‘AI’ is available is worth the time,” he wrote.

He also pushed back on treating AI visibility as a universal priority. Mueller suggested practitioners look at their own data first.

Mueller added:

“Also, be realistic and look at actual usage metrics and understand your audience (what % is using ‘AI’? what % is using Facebook? what does it mean for where you spend your time?).”

Why This Matters

I’ve been tracking Mueller’s public statements for years, and this one lands differently than the usual “it depends” responses he’s known for. He’s reframing the GEO question as a resource allocation problem rather than a terminology debate.

The GEO conversation has picked up steam over the past year as AI answer engines started sending measurable referral traffic. I’ve covered the citation studies, the traffic analyses, and the research comparing Google rankings to LLM citations. What’s been missing is a clear signal from Google: is this a distinct discipline, or just rebranded SEO?

Mueller’s answer is consistent with what Google said at Search Central Live, when Gary Illyes emphasized that AI features share infrastructure with traditional Search. The message from both is that you probably don’t need a separate framework, but you do need to understand how discovery is changing.

What I find more useful is his emphasis on checking your own numbers. Current data shows ChatGPT referrals at roughly 0.19% of traffic for the average site. AI assistants combined still drive less than 1% for most publishers. That’s growing, but it’s not yet a reason to reorganize your entire strategy.

The industry has a habit of chasing trends that apply to some sites but not others. Mueller’s pushing back on that pattern. Look at what percentage of your audience actually uses AI tools before reallocating resources toward them.

Looking Ahead

The GEO terminology will likely stick, regardless of Google’s stance. Mueller’s framing puts the decision back on individual businesses to measure their own audience behavior.

For practitioners, this means the homework is in your analytics. If AI referrals are showing up in your traffic sources, they’re worth understanding. If they’re not, you have other priorities.


Featured Image: Roman Samborskyi/Shutterstock

Google’s Mueller Explains ‘Page Indexed Without Content’ Error via @sejournal, @MattGSouthern

Google Search Advocate John Mueller responded to a question about the “Page Indexed without content” error in Search Console, explaining the issue typically stems from server or CDN blocking rather than JavaScript.

The exchange took place on Reddit after a user reported their homepage dropped from position 1 to position 15 following the error’s appearance.

What’s Happening?

Mueller clarified a common misconception about the cause of “Page Indexed without content” in Search Console.

Mueller wrote:

“Usually this means your server / CDN is blocking Google from receiving any content. This isn’t related to anything JavaScript. It’s usually a fairly low level block, sometimes based on Googlebot’s IP address, so it’ll probably be impossible to test from outside of the Search Console testing tools.”

The Reddit user had already attempted several diagnostic steps. They ran curl commands to fetch the page as Googlebot, checked for JavaScript blocking, and tested with Google’s Rich Results Test. Desktop inspection tools returned “Something went wrong” errors while mobile tools worked normally.

Mueller noted that standard external testing methods won’t catch these blocks.

He added:

“Also, this would mean that pages from your site will start dropping out of the index (soon, or already), so it’s a good idea to treat this as something urgent.”

The affected site uses Webflow as its CMS and Cloudflare as its CDN. The user reported the homepage had been indexing normally with no recent changes to the site.

Why This Matters

I’ve covered this type of problem repeatedly over the years. CDN and server configurations can inadvertently block Googlebot without affecting regular users or standard testing tools. The blocks often target specific IP ranges, which means curl tests and third-party crawlers won’t reproduce the problem.

I covered when Google first added “indexed without content” to the Index Coverage report. Google’s help documentation at the time noted the status means “for some reason Google could not read the content” and specified “this is not a case of robots.txt blocking.” The underlying cause is almost always something lower in the stack.

The Cloudflare detail caught my attention. I reported on a similar pattern when Mueller advised a site owner whose crawling stopped across multiple domains simultaneously. All affected sites used Cloudflare, and Mueller pointed to “shared infrastructure” as the likely culprit. The pattern here looks familiar.

More recently, I covered a Cloudflare outage in November that triggered 5xx spikes affecting crawling. That was a widespread incident. This case appears to be something more targeted, likely a bot protection rule or firewall setting that treats Googlebot’s IP addresses differently from other traffic.

Search Console’s URL Inspection tool and Live URL test remain the primary ways to identify these blocks. When those tools return errors while external tests pass, server-level blocking becomes the likely cause. Mueller made a similar point in August when advising on crawl rate drops, suggesting site owners “double-check what actually happened” and verify “if it was a CDN that actually blocked Googlebot.”

Looking Ahead

If you’re seeing the “Page Indexed without content” error, check the CDN and server configurations for rules that affect Googlebot’s IP ranges. Google publishes its crawler IP addresses, which can help identify whether security rules are targeting them.

The Search Console URL Inspection tool is the most reliable way to see what Google receives when crawling a page. External testing tools won’t catch IP-based blocks that only affect Google’s infrastructure.

For Cloudflare users specifically, check bot management settings, firewall rules, and any IP-based access controls. The configuration may have changed through automatic updates or new default settings rather than manual changes.

Why Global Search Misalignment Is An Engineering Feature And A Business Bug via @sejournal, @billhunt

Google’s AI Overviews (AIO) represent a fundamental architectural shift in search. Retrieval has moved from a localized ranking-and-serving model, designed to return the most appropriate regional URL, to a semantic synthesis model, designed to assemble the most complete and defensible explanation of a topic.

This shift has introduced a new and increasingly visible failure mode: geographic leakage, where AI Overviews cite international or out-of-market sources for queries with clear local or commercial relevance.

This behavior is not the result of broken geo-targeting, misconfigured hreflang, or poor international SEO hygiene. It is the predictable outcome of systems designed to resolve ambiguity through semantic expansion, not contextual narrowing. When a query is ambiguous, AI Overviews prioritize explanatory completeness across all plausible interpretations. Sources that resolve any sub-facet with greater clarity, specificity, or freshness gain disproportionate influence – regardless of whether they are commercially usable or geographically appropriate for the user.

From an engineering perspective, this is a technical success. The system reduces hallucination risk, maximizes factual coverage, and surfaces diverse perspectives. From a business and user perspective, however, it exposes a structural gap: AI Overviews have no native concept of commercial harm. The system does not evaluate whether a cited source can be acted upon, purchased from, or legally used in the user’s market.

This article reframes geographic leakage as a feature-bug duality inherent to generative search. It explains why established mechanisms such as hreflang struggle in AI-driven experiences, identifies ambiguity and semantic normalization as force multipliers in misalignment, and outlines a Generative Engine Optimization (GEO) framework to help organizations adapt in the generative era.

The Engineering Perspective: A Feature Of Robust Retrieval

From an AI engineering standpoint, selecting an international source for an AI Overview is not an error. It is the intended outcome of a system optimized for factual grounding, semantic recall, and hallucination prevention.

1. Query Fan-Out And Technical Precision

AI Overviews employ a query fan-out mechanism that decomposes a single user prompt into multiple parallel sub-queries. Each sub-query explores a different facet of the topic – definitions, mechanics, constraints, legality, role-specific usage, or comparative attributes.

The unit of competition in this system is no longer the page or the domain. It is the fact-chunk. If a particular source contains a paragraph or explanation that is more explicit, more extractable, or more clearly structured for a specific sub-query, it may be selected as a high-confidence informational anchor – even if it is not the best overall page for the user.

2. Cross-Language Information Retrieval (CLIR)

The appearance of English summaries sourced from foreign-language pages is a direct result of Cross-Language Information Retrieval.

Modern LLMs are natively multilingual. They do not “translate” pages as a discrete step. Instead, they normalize content from different languages into a shared semantic space and synthesize responses based on learned facts rather than visible snippets. As a result, language differences no longer serve as a natural boundary in retrieval decisions.

Semantic Retrieval Vs. Ranking Logic: A Structural Disconnect

The technical disconnect observed in AI Overviews, where an out-of-market page is cited despite the presence of a fully localized equivalent, stems from a fundamental conflict between search ranking logic and LLM retrieval logic.

Traditional Google Search is designed around serving. Signals such as IP location, language, and hreflang act as strong directives once relevance has been established, determining which regional URL should be shown to the user.

Generative systems are designed around retrieval and grounding. In Retrieval-Augmented Generation pipelines, these same signals are frequently treated as secondary hints, or ignored entirely, when they conflict with higher-confidence semantic matches discovered during fan-out retrieval.

Once a specific URL has been selected as the source of truth for a given fact, downstream geographic logic has limited ability to override that choice.

The Vector Identity Problem: When Markets Collapse Into Meaning

At the core of this behavior is a vector identity problem.

In modern LLM architectures, content is represented as numerical vectors encoding semantic meaning. When two pages contain substantively identical content, even if they serve different markets, they are often normalized into the same or near-identical semantic vector.

From the model’s perspective, these pages are interchangeable expressions of the same underlying entity or concept. Market-specific constraints such as shipping eligibility, currency, or checkout availability are not semantic properties of the text itself; they are metadata properties of the URL.

During the grounding phase, the AI selects sources from a pool of high-confidence semantic matches. If one regional version was crawled more recently, rendered more cleanly, or expressed the concept more explicitly, it can be selected without evaluating whether it is commercially usable for the searcher.

Freshness As A Semantic Multiplier

Freshness amplifies this effect. Retrieval-Augmented Generation systems often treat recency as a proxy for accuracy. When semantic representations are already normalized across languages and markets, even a minor update to one regional page can unintentionally elevate it above otherwise equivalent localized versions.

Importantly, this does not require a substantive difference in content. A change in phrasing, the addition of a clarifying sentence, or a more explicit explanation can tip the balance. Freshness, therefore, acts as a multiplier on semantic dominance, not as a neutral ranking signal.

Ambiguity As A Force Multiplier In Generative Retrieval

One of the most significant, and least understood, drivers of geographic leakage is query ambiguity.

In traditional search, ambiguity was often resolved late in the process, at the ranking or serving layer, using contextual signals such as user location, language, device, and historical behavior. Users were trained to trust that Google would infer intent and localize results accordingly.

Generative retrieval systems respond to ambiguity very differently. Rather than forcing early intent resolution, ambiguity triggers semantic expansion. The system explores all plausible interpretations in parallel, with the explicit goal of maximizing explanatory completeness.

This is an intentional design choice. It reduces the risk of omission and improves answer defensibility. However, it introduces a new failure mode: as the system optimizes for completeness, it becomes increasingly willing to violate commercial and geographic constraints that were previously enforced downstream.

In ambiguous queries, the system is no longer asking, “Which result is most appropriate for this user?”

It is asking, “Which sources most completely resolve the space of possible meanings?”

Why Correct Hreflang Is Overridden

The presence of a correctly implemented hreflang cluster does not guarantee regional preference in AI Overviews because hreflang operates at a different layer of the system.

Hreflang was designed for a post-retrieval substitution model. Once a relevant page is identified, the appropriate regional variant is served. In AI Overviews, relevance is resolved upstream during fan-out and semantic retrieval.

When fan-out sub-queries focus on definitions, mechanics, legality, or role-specific usage, the system prioritizes informational density over transactional alignment. If an international or home-market page provides the “first best answer” for a specific sub-query, that page is retrieved immediately as a grounding source.

Unless a localized version provides a technically superior answer for the same semantic branch, it is simply not considered.

In short, hreflang can influence which URL is served. It cannot influence which URL is retrieved, and in AI Overviews, retrieval is where the decision is effectively made.

The Diversity Mandate: The Programmatic Driver Of Leakage

AI Overviews are explicitly designed to surface a broader and more diverse set of sources than traditional top 10 search results.

To satisfy this requirement, the system evaluates URLs, not business entities, as distinct sources. International subfolders or country-specific paths are therefore treated as independent candidates, even when they represent the same brand and product.

Once a primary brand URL has been selected, the diversity filter may actively seek an alternative URL to populate additional source cards. This creates a form of ghost diversity, where the system appears to surface multiple perspectives while effectively referencing the same entity through different market endpoints.

The Business Perspective: A Commercial Bug

The failures described below are not due to misconfigured geo-targeting or incomplete localization. They are the predictable downstream consequence of a system optimized to resolve ambiguity through semantic completeness rather than commercial utility.

1. The Commercial Blind Spot

From a business standpoint, the goal of search is to facilitate action. AI Overviews, however, do not evaluate whether a cited source can be acted upon. They have no native concept of commercial harm.

When users are directed to out-of-market destinations, conversion probability collapses. These dead-end outcomes are invisible to the system’s evaluation loop and therefore incur no corrective penalty.

2. Geographic Signal Invalidation

Signals that once governed regional relevance – IP location, language, currency, and hreflang – were designed for ranking and serving. In generative synthesis, they function as weak hints that are frequently overridden by higher-confidence semantic matches selected upstream.

3. Zero-Click Amplification

AI Overviews occupy the most prominent position on the SERP. As organic real estate shrinks and zero-click behavior increases, the few cited sources receive disproportionate attention. When those citations are geographically misaligned, opportunity loss is amplified.

The Generative Search Technical Audit Process

To adapt, organizations must move beyond traditional visibility optimization towards what we would now call Generative Engine Optimization (GEO).

  1. Semantic Parity: Ensure absolute parity at the fact-chunk level across markets. Minor asymmetries can create unintended retrieval advantages.
  2. Retrieval-Aware Structuring: Structure content into atomic, extractable blocks aligned to likely fan-out branches.
  3. Utility Signal Reinforcement: Provide explicit machine-readable indicators of market validity and availability to reinforce constraints the AI does not infer reliably on its own.

Conclusion: Where The Feature Becomes The Bug

Geographic leakage is not a regression in search quality. It is the natural outcome of search transitioning from transactional routing to informational synthesis.

From an engineering perspective, AI Overviews are functioning exactly as designed. Ambiguity triggers expansion. Completeness is prioritized. Semantic confidence wins.

From a business and user perspective, the same behavior exposes a structural blind spot. The system cannot distinguish between factually correct and consumer-engagable information.

This is the defining tension of generative search: A feature designed to ensure completeness becomes a bug when completeness overrides utility.

Until generative systems incorporate stronger notions of market validity and actionability, organizations must adapt defensively. In the AI era, visibility is no longer won by ranking alone. It is earned by ensuring that the most complete version of the truth is also the most usable one.

More Resources:


Featured Image: Roman Samborskyi/Shutterstock

How Search Engines Tailor Results To Individual Users & How Brands Should Manage It

How many times have you seen different SERP layouts and results across markets?

No two people see the same search results, as per Google’s own documentation. No two users receive identical outputs from AI platforms either, even when using the same prompt. In a time of information overload, this raises an important question for global marketers: How do we manage and leverage personalized search experiences across multiple markets?

Today, clarity and transparency matter more than ever. Users have countless choices and distractions, so they expect experiences that feel relevant, trustworthy, and aligned with their needs in the moment. Personalization is now central to how potential customers discover, evaluate, and engage with brands.

Search engines have been personalizing results for years based on language, search behavior, device type, and technical elements such as hreflang. With the quick evolution of generative artificial intelligence (AI), personalization has expanded into summarized answers on AI platforms and hyper-personalized experiences that depend on internal data flows and processes.

This shift forces marketers to rethink how they measure visibility and business impact. According to McKinsey, 76% of users feel frustrated when experiences are not personalized, which shows how closely relevance and user satisfaction are linked.

At the same time, long-tail discovery increasingly happens outside of search engines, particularly on platforms like TikTok. Statista reports that 78% of global internet users now research brands and products on social media.

All of this is happening while most users know little about how search engines or AI systems operate.

Regardless of where people search, the implications extend far beyond algorithms. Personalization affects how teams collaborate, how data moves across departments, and how global organizations define success.

This article explores what personalization means today and how global brands can turn it into a competitive advantage.

From SERPs To AI Summaries

Search engines no longer return lists of blue links alone or People Also Ask (PAA). They now provide summarized information in AI Overviews and AI Mode, currently for informational queries.

Google often surfaces AI summaries first and URLs second, while continuously testing different layouts for mobile and desktop, as shown below.

Screenshot from search for [what is a nepo baby], Google, December 2025

Google’s Search Labs experiments, including features such as Preferred Sources, show how layouts and summaries change based on context, trust signals, and behavioral patterns.

Large language models (LLMs) add another layer. They adjust responses based on user context, intent, and sometimes whether the user has a free or paid account. Because users rarely get exactly what they need on the first attempt, they re-prompt the AI, creating iterative conversations where each instruction or prompt influences the next.

What prompts users to click through to a source or research it on search engines, whether it is curiosity, uncertainty, boredom, a call-to-action, or the model stating it does not know, is still unclear. Understanding this behavior will soon be as important as traditional click-through rate (CTR) analysis.

For global brands, the challenge is not simply keeping up with technology. It’s maintaining a consistent brand voice and value exchange across channels and markets when every user sees a different interpretation of the brand. Trust is now as important as visibility.

This landscape increases the importance of market research, segmentation, cultural insights, and competitive analysis. It also raises concerns about echo chambers, search inequality, and the barriers brands face when entering new markets or reaching new audiences.

Meanwhile, the long tail continues to shift to platforms like TikTok, where discovery works very differently from traditional search. And as enthusiasm for AI cools, many professionals believe we have entered the “trough of disillusionment” stage described by Jackie Fenn’s technology adoption lifecycle.

What Personalization Means Today

In marketing, personalization refers to tailoring content, offers, and experiences based on available data.

In search, it describes how search engines customize results and SERP features for individual users using signals such as:

  • Data patterns.
  • Inferred interests.
  • Location.
  • Search behavior.
  • Device type.
  • Language.
  • AI-driven memory (which is discussed below).

The goal of search engines is to provide relevant results and keep users engaged, especially as people now search across multiple channels and AI platforms. As a result of this, two people searching the same query rarely see identical results. For example:

  • A cuisine enthusiast searching for [apples] may see food-related content.
  • A tech-oriented user may see Apple product news.

SERP features can also vary across markets and profiles. People Also Ask (PAA) questions and filters may differ by region, language, or click behavior, and may not appear at all. For example, the query “vote of no confidence” displays different filters and different top results in Spain and the UK, and PAA does not appear in the UK version.

AI platforms push this further with session-based memory. Platforms like AI Mode, Gemini, ChatGPT, and Copilot handle context in a way that makes users feel there are real conversations, with each prompt influencing the next. In some cases, results from earlier responses may also be surfaced.

A human-in-the-loop (HITL) approach is essential to evaluate, monitor, and correct outputs before using them.

How Personalization Technically Works

Personalization operates across several layers. Understanding these helps marketers see where influence is possible.

1. SERP Features And Layout

Google and Bing adapt their layouts based on history, device type, user engagement, and market signals. Featured Snippets, PAA modules, videos, forums, or Top Stories may appear or disappear depending on behavior and intent.

2. AI Overviews, AI Mode, And Bing Copilot

AI platforms can:

  • Summarize content from multiple URLs.
  • Adapt tone and depth based on user behavior.
  • Personalize follow-up suggestions.
  • Integrate patterns learnt within the session or even previous sessions.

Visibility now includes being referenced in AI summaries. Current patterns show this depends on:

  • Clear site and URL structure.
  • Factual accuracy.
  • Strong entity signals.
  • Online credibility.
  • Fresh, easily interpreted content.

3. Structured Data And Entity Consistency

When algorithms understand a brand, they can personalize results more accurately. Schema markup helps avoid entity drift, where regional websites are mistaken for separate brands.

Bing uses Microsoft Graph to connect brand data with the Microsoft ecosystem, extending the influence of structured data.

4. Context Windows And AI Memory

LLMs simulate “memory” using context windows, which is the amount of information they can consider at once. This is measured in tokens, which represent words or parts of words. It is what makes conversations feel continuous.

This has some important implications:

  • Semantic consistency matters.
  • Tone should be unified across markets.
  • Messaging needs to be coherent across content formats.

Once an AI system associates a brand with a specific theme, that context can persist for a while, although it is unclear how long for. This is probably why LLMs favor fresh content as a way to reinforce authority.

5. Recommenders

In ecommerce and content-heavy sites, recommenders show personalized suggestions based on behavior. This reduces friction and increases time on site.

Benefits Of Personalization

When personalization works, users and brands can benefit from:

  • Reduced user friction.
  • Increased user satisfaction.
  • Improved conversion rates.
  • Stronger engagement.
  • Higher CTR.

This can positively influence the customer lifetime value. However, these benefits rely on consistent and trustworthy experiences across channels.

Potential Drawbacks

Alongside the benefits, personalization brings some challenges that marketers need to be aware of. These are not reasons to avoid personalization, but important considerations when planning global strategies. Consider:

  • Filter bubbles reduce exposure to diverse viewpoints and competing brands.
  • Privacy concerns increase as platforms rely on more behavioral and demographic data.
  • Reduced result diversity makes it harder for new or smaller brands to appear.
  • Global templates lose effectiveness when markets expect local nuance.

This means that brands using the same template or unified content across markets for globalization lose even more effectiveness in markets, as cultural nuance, context, or different user motivations are expected. Furthermore, purchase journeys vary across markets. Hence, the effectiveness of hyper-personalization.

It is probably more important than ever that brands spend time researching and planning to gain or maintain visibility in global markets, as well as strengthening their brand perception.

Managing Personalization Across Teams And Channels

At the moment, LLMs tend to favor strong, clearly structured brands and websites. If a brand is not well understood online, it is less likely to be referenced in AI summaries.

Successful digital and SEO projects rely on strong internal processes. When teams work in isolation, inconsistencies appear in data, content, and technical implementation, which then surface as inconsistencies in personalized search.

Common issues include:

  • Weak global alignment.
  • Translations that miss local relevance.
  • Conflicting schema markup.
  • Local pages ranking for the wrong intent.
  • Important local keywords being ignored.

Below is a framework to help organizations manage personalization across markets and channels.

1. Shared Objectives And Understanding Across Teams

Many search or marketing challenges can be prevented by building a shared understanding across teams of:

  • Business and project goals.
  • Issues across markets.
  • Search developments across markets.
  • Audience segmentation.
  • Integrated insights across all channels.
  • Data flows that connect global and local teams.
  • AI developments.

2. Strengthen The Technical Elements Of Your Website

Reinforce the technical elements of your website so that it is easy for search engines and LLMs to understand your brand across markets to avoid entity drift:

  • Website structure.
  • Schema markup on the appropriate sections.
  • Strong on-page structure.
  • Strong internal linking.
  • Appropriate hreflang.

3. Optimize For Content Clusters And User Intent, Not Keywords

Structure is everything. Organizing content into clusters helps users and search engines understand the website clearly, which supports personalization.

4. Use First-Party Data To Personalize On-Site Experiences

Internal search and logged-in user experiences are important to understand your users and build user journeys based on behavior. This helps with content relevance and stronger intent signals.

First-party data can support:

  • Personalized product recommendations.
  • Dynamic filters.
  • Auto-suggestions based on browsing behavior.

5. Maintain Cross-Channel Consistency

A coherent experience supports stronger personalization and prevents fragmented journeys, and search is only one personalized environment. Tone, structure, messaging, and data should remain consistent across:

  • Social platforms.
  • Email.
  • Mobile apps.
  • Websites and on-site search.

Clear and consistent USPs should be visible everywhere.

6. Strengthen Your Brand Perception

With so much online competition, brands whose work is being referenced positively across the internet. It is the old PR: Focus on your strengths and publish well-researched work, with stats that are useful to your target users.

Conclusion: Turning Personalization Into An Advantage

Conway’s Law matters more than ever. The idea that organizations design systems that mirror their own communication structures is highly visible in search today. If teams operate in silos, those silos often show up in fragmented content, inconsistent signals, and mixed user experiences. Personalization then amplifies these gaps even further by not being cited on AI platforms or the wrong information being spread.

Understanding how personalization works and how it shapes visibility, trust, and user behavior helps brands deliver experiences that feel coherent rather than confusing.

Success is no longer just about optimizing for Google. It is about understanding how people search, how AI interprets and summarizes content, how brands are referenced across the web, and how teams collaborate across channels to present a unified message.

Where every search result is unique, the brands that succeed will be the ones that coordinate, connect, and communicate clearly, both internally and across global markets, to help strengthen the perception of their brand.

More Resources:


Featured Image: Master1305/Shutterstock