Decision-makers at brands and agencies know that the new AI-generated holiday ads from Coca-Cola have attracted a lot of criticism.
Others have described the three new AI versions of the classic “Holidays Are Coming” campaign as “a soulless and creepy, dystopian nightmare” and “the biggest branding blunder of the year,” with others saying the AI campaign “destroyed the spirit of Christmas” and “earns Coca-Cola a lump of coal.”
Strong words. But has Manuel “Manolo” Arroyo, the executive vice president and global chief marketing officer for the company, just made a career-damaging move?
In testing for festive campaigns globally by DAIVID, none of Coke’s new AI-generated holiday ads made the top 30 most effective holiday campaigns of 2024 against 90 other Christmas ads.
Watch the new AI-generated holiday ads, which were created by three different ad agencies, and form your own opinion.
Secret Santa
Secret Level created “Coca-Cola – Secret Santa (AI-Generated Christmas Ad 2024).”
Holidays Are Coming
Silverside created “Coca Cola – Holidays Are Coming.”
Unexpected Santa
Wildcard created “Coca-Cola – Unexpected Santa (AI-Generated Christmas Ad 2024).”
Holidays Are Coming 2020
While you’re reviewing these new versions, you should also watch the version that was uploaded to Coca-Cola Great Britain & Ireland’s YouTube channel back in 2020.
How Do Coke’s New AI Versions Compare To The Classic 2020 Ad?
What do you notice? What do you wonder?
Attention
All the new AI versions generated above-average attention from the start.
However, the classic version, which starts with a boy ringing a bell, captures more attention than any of the AI versions, which mostly start with shots of snowy landscapes.
People will generally attract more attention than images of trees and lakes.
Prevalence Of Intense Emotions
According to testing by DAIVID, none of the AI ads generate the same levels of intense positive emotions as the 2020 version, and all of them are below the industry average.
The 2020 version generates almost twice as much warmth as the norm, while the AI versions are level or slightly above.
The AI version that generated the most warmth was still 38% less likely to make people feel warmth than the 2020 version.
The AI versions were less relatable and less – for want of a better word – real.
Brand Recall
All of the new AI versions predictably scored above the industry average for correct brand recall.
This is not surprising, considering that people know the ad well, and the brand is present throughout and integral to the storyline (Coke Trucks).
The classic scores higher than the AI versions, though. This, again, is possibly due to the familiarity of the ad, but also the fact the famous “Holidays Are Coming” track kicks in much quicker.
Next Step Intents
One of the emotions that the AI versions consistently scored higher than the 2020 ad for is feelings of craving. All are around two to three times higher than average.
This is probably due to the close-ups of someone opening a cold bottle of Coke, which wasn’t included in the 2020 version.
What Was The Most Effective AI Version?
Ian Forrester of DAIVID reported:
“The AI versions of Coke’s classic ‘Holidays Are Coming’ campaign were strong for attention in the first second and brand recall, but were let down by their evocation of intense positive emotions, which were all below the industry norm.
The difference between the AI and the original was most stark in their evocation of warmth, a mainstay of Christmas advertising. The original evoked intense warmth among 33.0% of viewers, whereas the AI versions were significantly below this.
So, while the AI is producing images which on the face of it seem cute and heart-warming, the human viewer to some degree discerns their synthetic nature, which detracts from their impact.”
How Can Brands Avoid AI Negative Backlash?
After analyzing the data published by DAIVID, I reached out directly and spoke to their Chief Growth Officer, Barney Worfolk-Smith:
GJ: Why does AI have such a negative perception?
BWS: It’s not surprising that the use of generative AI, especially jazzing up familiar Christmas traditions like Coke’s truck, garners some negative opinions.
As the introduction of generative AI into processes is nascent and messy at best, none of us really know exactly how it will play out.
So, some in the advertising community who feel a sense of ominous threat will instantly adopt a negative stance. I don’t blame them, but the reality is, the toothpaste is out of the tube, so we should all have a hand on the wheel of a human-AI hybrid Christmas Coke truck to have a stake in the future.
GJ: Can brands navigate carefully to avoid backlash?
BWS: Generative AI is present – or at least coming down the chimney – in almost all aspects of advertising. It’s actually incumbent upon brands to try bits of it out.
Sure, it’s going to be bumpy, but the backlashes will frequently be confined to the advertising community.
As a result, as long as they’re doing measured introductory human AI experiments and not dismissing the agency of record, I think they’ll avoid a hit on the share price.
GJ: Why was the original video such a classic?
BWS: The original was a glorious confluence: strong, familiar emotions, which Coca-Cola evokes generally, the shared history of Santa and Coca-Cola’s colors, and a palpable, relatable sense of anticipation that even the “Grinchiest” of us feel in the run-up to Christmas.
GJ: Why has AI failed to replicate the success of the first campaign?
BWS: At DAIVID, we understand the importance emotions play in advertising effectiveness – and the AI versions all garnered below-average U.S. positive emotional responses.
Without a doubt, the uncanny valley plays a part here, especially with an advert that is so recognizable to so many of us.
GJ: What must marketers do when using AI in video or images?
BWS: Marketers need to take their eyes off the spreadsheet and on to the creative process.
Of course, AI can drive efficiencies, but it can also open up new avenues of creativity, and that will happen when creatives are empowered to use AI, not be threatened with it.
Embrace AI Cautiously In Holiday Ads
Holiday ads are notoriously tricky to navigate and strike the right sentiment, with the best intention often missing the mark.
Feelings of warmth and nostalgia are at the heart of the festive season. Perhaps AI just can’t replicate the nuance of human emotion – or more likely, humans don’t like the idea of AI trying to replicate that.
Coca-Cola’s new ads emphasizes the challenge for brands to cultivate emotional authenticity when engaging with their audience as AI becomes more integrated into advertising campaigns.
It reminds us to embrace AI cautiously while upholding the human elements that underpin marketing campaigns – holiday ads, in particular.
Methodology
DAIVID used its AI-powered platform that predicts the emotions an ad will generate, and its likely impact on brand and business metrics – enabling advertisers to measure the effectiveness of their ad campaigns at scale.
They tested 90 Christmas ads for 39 different emotions. The strength of emotions people feel is ranked from 1-10, with 8-10 considered “intense.” Data for the chart was compiled at 7:00 AM on November 15, 2024.
For most websites, the homepage represents your brand’s first interaction with your audience on your website.
As the catch-all landing page where people will be sent by default, your homepage needs to cater to a breadth of user needs and intent.
Consider your homepage as your shop window. It showcases your most valuable content, positions your brand so that it stands apart from the competition, and leads the user to take an initial action to enter the site and see more.
Your homepage sets the tone for your brand identity and communicates the brand messaging, company values, and personality of your business.
From your homepage, you have the first opportunity to establish positive brand recognition and clearly define the company’s value proposition to new and repeat users.
A homepage has many practical functions, too, such as:
Catching all topic areas that do not currently have dedicated destinations on the website.
Resonating with your audience through your brand positioning.
Here are 25 of the best examples of homepages. I’ve looked to include a wide variation of homepage examples, so you can see practically where you can refine your own homepage for increased performance.
1. So Cosy
This homepage example combines setting the perfect tone through distilled messaging, imagery, and color scheme.
Everything is simplified so the user can relax, discover, and enjoy the website.
Often, so many conflicting messages are crammed into a home page that the brand and purpose of the site become lost in the noise.
This example is the perfect reminder that, in many cases, simplicity pays off for the user and for search.
Screenshot from socosy.co.uk, November 2024
2. Toby Carvery
Homepages should reflect changes in your audience’s interest areas, industry trends, and broader seasonality.
This enables tailored messaging and ongoing servicing of intent to the full. This requirement becomes even greater in certain industries such as food, travel, and hospitality.
Your homepage sets the scene, showing what resonates with your key audience types, and should be proactively updated based on changing data sets.
Certain industries, from financial services to safety supply companies, have a stigma and preconception attached to them as needing to be positioned in extremely formal ways. It can become an easy trap to fall into, assuming people expect a certain tone and positioning.
In this homepage example, the tone is relaxed, friendly, and welcoming.
Tonality includes statements like “it’s ok” when accepting cookies, and “we are here” to encourage chatbot interaction.
This subtle messaging, combined with audience-aware images and related content positioning, is a fantastic way to see how your homepage can set the tone and reinforce brand positioning from the outset.
Screenshot from safetecdirect.co.uk, November 2024
4. Post Office
In some cases, the homepage can be the fastest way for people to achieve their search goals.
In this example, the homepage facilitates the four core functions that the users of the website look to complete most frequently without the need to go through additional pages/clicks.
Widgets on the homepage service immediate action completion in a fast, fun, and intuitive way.
While the task of paying a bill, sending a parcel, or tracking postage may not seem a fun task, this website presents a light tone and an easy way to complete your intended actions as quickly as possible, so you can get back to your other activities.
Screenshot from postoffice.co.uk, November 2024
5. TED
The TED homepage embodies the company’s mission of sharing information, ideas, and interests in an easy-to-digest, accessible fashion.
Content is themed into playlists, the latest, creative ideas, and other taxonomies such as “small world.”
The website taxonomy supports fast access to information topics and facilitates an easy and intuitive approach to information architecture at scale.
Screenshot from ted.com, November 2024
6. Pandora
When it comes to associating the brand with the audience, Pandora does a fantastic job.
There is instant clarity, alignment of messaging, and impact of images that enable fast user engagement and establish trust and brand awareness.
Screenshot from uk.pandora.net/en/, November 2024
7. Davy Wine
An important aspect of homepage success comes down to the use of evidence (data) to drive decision-making.
The ordering of information displayed for the home page, content segmentation, and CTAs are arguably more important than any other page on your website.
For ecommerce sites, this necessity is becoming increasingly important.
This homepage example showcases the application in data to drive optimized user journeys from the moment they land on the homepage.
Screenshot from davywine.co.uk, November 2024
8. Under Armour
This entry into the top 25 homepage examples warrants its place, based upon well-planned information architecture and scannable content, which gives users an enjoyable experience.
Function and “fit for purpose” are understated homepage virtues that this site brings to the fore.
There is also the seasonal aspect of tapping into changing needs, wants, and pain points effectively.
The “Spotlight” segment also works well for presenting new/fresh information to returning users to expand the type of purchases being made.
Screenshot from underarmour.co.uk/en-gb/, November 2024
9. BMW
Many car sales and dealership websites have similar approaches to homepages.
The dominant key model image is supported by quick filtering options to drive users to convert.
The stand-out item from this BMW homepage example, however, is the simplicity of messaging combined with minimal conflicting CTAs for the user.
There is no excessive sales content, and the homepage enables natural next steps rather than the excessive pushing of deals and related commercial CTAs often seen in this space.
The key feature of this homepage offering is that by pre-scroll, the user has full access to everything they need without taking any further action.
They can see trust signals, get in contact, explore the main sections of the website, and receive a myriad of positive reinforcement specific to their lifestyle choices.
Screenshot from ucfb.ac.uk, November 2024
11. Productive
Software companies need to compete in extremely crowded places where the research time and tolerance of the target market are often very limited.
This places increased emphasis on clarity in messaging, CTAs, and value proposition – all of which are present in this example of a best practice homepage.
Screenshot from productive.io, November 2024
12. Skype
One of the greatest challenges for homepages is to resonate with a varied audience effectively.
Skype handles this dilemma extremely well through a dominant audience message, supplemented by very clear and distinct alternative audience content assets.
Varying this based on trends and related data ensures every core persona receives initial verification to remain actively engaged on the website.
Screenshot from skype.com, November 2024
13. Uber
It takes a strong brand and confidence in user trust to present a homepage that is dominated by action-taking over value proposition.
Uber pre-scroll is 100% action-orientated, enabling the quickest route to booking prior to any conflicting messaging or related distractions.
The assumption is that if you land on the Uber site, the only thing that matters is getting you from “A to B” and servicing that intent to book above all else – and it works.
Screenshot from uber.com, November 2024
14. Dropbox
The simplicity of design and clarity of messaging are consistent throughout many of these examples. For a homepage, that is often a core challenge, as well as an aspirational goal.
In this example, headlines are emotive, and supporting statements are clear.
Mixed media walkthroughs of the service provide a trial of the solution without the need to sign up for one. It’s a great example of shortening the distance to purchase/use.
Screenshot from dropbox.com, November 2024
15. Allen Carr’s Easyway
When it comes to Your Money Your Life (YMYL) industries, homepages have additional challenges.
First, trust needs to be ever-present and supported by statistics without detriment to the brand’s style and tone.
Next, direct reinforcement of success, case studies, and audience associative needs are higher. And providing a positive outlook on tougher topic areas is far from easy.
This homepage example manages to cover all these areas plus more.
Screenshot from allencarr.com, November 2024
16. NineFeetTall
When companies provide transformation and change, like in this example, you have to balance data and justification from the outset.
The homepage acts as the roadmap from now to the near future and requires expert guidance without information overload.
Every segment of this homepage example contributes toward this journey, empowering people to learn fast and take action sooner.
Screenshot from ninefeettall.com, November 2024
17. NHS
Websites that provide emergency help and support must reinforce trust, provide immediate access to contact, and solve problems from the first meaningful homepage interaction.
Visually, the homepage needs to drive action-taking and fuel the right choice to minimize already stressful situations.
Considering the vast array of people using emergency services like the NHS, intuitive and simple design comes into play with clear, concise content.
Screenshot from nhs.uk, November 2024
18. WeChat
Named one of the world’s strongest brands, the app’s homepage shows the opportunities to change the status quo with design and focus on brand-led power and positioning.
The navigation placement and impact on the page versus the streamlined and dominant CTA is an interesting approach.
Screenshot from wechat.com, November 2024
19. Colgate
For established and traditional brands, the homepage can present a complex range of choices.
One of these is how to remain relevant with existing audiences while looking to grow visibility with new people in fresh ways.
Colgate achieves this with a combination of trust and visual reinforcement.
Screenshot from colgate.com, November 2024
20. Basecamp
The Basecamp homepage jumps straight into solving the main pain point of its audience.
This is then supported by segments that all actively contribute to the purpose of Basecamp as a service and nudge the user towards purchase.
This journey is without added clicks or engagement required – it’s a complete conversation on a single page:
The headline positions the brand and service.
The segmented homepage tells the story of why you may invest in the service.
The dominant CTA jumps out of the page.
Homepage screenshots provide an instant demo of the solution in action.
Screenshot from basecamp.com, November 2024
21. Time
A media site’s user base has high standards and expectations for creative, fast, and functional websites.
The active use of white space is refreshing, as are the limited CTAs and removal of advertising.
The use of image, media, and text interaction supports audience preference and all device action-taking.
Screenshot from time.com, November 2024
22. Ocado
Large retail sites have to cram in many potential and often competing triggers to drive action and speed up access to the endpoint.
User tolerance levels for online shopping are very tough to meet, plus you are catering to a variety of audience awareness and trust.
Ocado manages to build in quick access CTAs, clear trust signals, and simple steps to purchase without cluttering the page or pulling the user into conflicting directions.
Screenshot from ocado.com, November 2024
23. Trivago
Comparison websites can feel like a bombardment of CTAs and promotional offers.
The Trivago homepage provides a relaxed, easy, and intuitive approach to booking that removes some of the complexity and time for the user.
Screenshot from trivago.co.uk, November 2024
24. eBay
From a data-driven and personalization stance, sites like eBay need to be present in the best examples of homepages.
Data is at the center of the design choices and content provisions and is frequently refined to bring the user closer to their perfect next buy, whether they are aware of it yet or not.
Screenshot from ebay.co.uk, November 2024
25. Imgur
Everything on this homepage shouts out fun, interaction, and enjoyment. Its core functionality is to make things simple to click, watch, and engage.
Yes, there is some quite intrusive advertising, but there is also an element of new audiences meeting nostalgia here with the early age of the internet ad space.
Screenshot from imgur.com, November 2024
What Should A Homepage Include?
First and foremost, the homepage needs to represent your brand values and proposition. Reinforcing the unique culture of your business and supporting brand recognition.
This is achieved through every piece of content, imagery, and prioritization of messaging on the page.
As your shop window, you need to present the most relevant messaging and CTAs that will resonate with your audience and drive them to click further into your website content and their unique conversion journey.
Visual elements should be of high quality, not competing with other on-page items, and making it simple for people and search engines to understand the core purpose of your site and what your brand represents.
Trust should be set from the outset. This includes star ratings and brand narrative through case studies and related social proof.
Your homepage needs to set out the key content assets and products/services that are the cornerstone of your business.
As with all pages on the site, the user experience is of even greater importance to the homepage. Their engagement should be fast, intuitive, and accessible for all content and devices.
And while there are other areas too, don’t forget to have readily available contact details that reinforce the brand identity, personality, and company values.
A Homepage Is A Showcase And A Signal Of Trust
The homepage is often the first interaction users have with your brand, serving as a critical entry point for visitors.
Your homepage is your shop window, showcasing your most valuable content and differentiating your brand from competitors while guiding users toward taking their first actions on your website.
There are many key functions that a homepage plays, including:
First Impressions: Creating an impactful introduction to the brand.
User Journeys: Drives visitors into conversion funnels.
Content Discovery: Helps users find products and services fast.
Incentives: Highlights promotions to encourage clicks and engagement.
Trust Building: Builds expertise and authority through social proof and related trust signals.
Comprehensive Coverage: Addresses topics without dedicated pages as a catch-all for search and users.
Audience Resonance: Reflects brand positioning and core values.
There are many essential components of a successful homepage, lots of which can be seen in the 25 best examples of homepages shared in this post.
To recap them, you should be thinking practically about:
Brand Representation: Clearly showcase your brand values and unique culture through the content and imagery you provide.
Relevant Messaging and CTAs: Prioritize calls to action that resonate with the audience, but limit them to a maximum of three to avoid conflicting attention demands.
High-Quality Visuals: Make sure that all visuals enhance understanding without competing for attention, and that they are unique and of high quality.
Trust Signals: Include reviews and ratings, case studies, and social proof from the outset so people can see a clear association with your existing and target audience types.
Key Content Assets: Highlight essential products and services that are cornerstones of your business offering.
User Experience: Focus on fast, intuitive, and accessible navigation and content in all its forms and for all devices.
Contact Information: Give easily accessible contact details to reinforce brand identity.
Google’s click-through rates (CTRs) experienced notable changes across industries and search categories in Q3, according to a report from Advanced Web Ranking.
This report compares Q3 data to the previous quarter. It shows how CTRs can vary and what that means for website traffic.
Key Findings:
Branded searches on mobile saw a 1.07 percentage point increase in CTR for top-ranked sites
Informational queries (containing words like “what,” “when,” “how”) gained 1.63 percentage points on mobile
Commercial queries declined across devices, with mobile dropping 3.51 percentage points
Short keyword searches (1-3 words) showed improved CTR on mobile devices
Industry Winners & Losers
To assess traffic impact, the report looked at changes in CTR alongside search demand trends for different industries.
When both CTR and demand grow at the same time, it signals likely traffic gains. However, if both decrease, it may indicate potential losses.
The Science sector bounced back after two quarters of falling CTR.
The top results saw an increase of 2.48 percentage points (pp) on desktop and 4.16 pp on mobile. Impressions also went up by 33.78%.
The Law, Government & Politics sector had the biggest drop in single-position CTR, with desktop websites ranked second, falling by 9.74 pp. Still, overall demand grew by 32.74%.
After a year of stable CTRs, the Shopping sector experienced a recovery in Q3. The top position increased by 2.30 pp on desktop and 1.94 pp on mobile, with a demand rise of 21.09%.
Other industries with notable CTR increases include:
Automotive: +2.95 pp desktop, +1.40 pp mobile (#1)
Business: +1.52 pp mobile (#1)
Education: +2.53 pp mobile (#1)
Family & Parenting: +2.42 pp desktop, +2.39 pp mobile (#1)
On the losing end, Arts & Entertainment saw desktop CTRs sink 6.56 pp and 1.42 pp for positions one and two and a 4.12 pp mobile slide for the top spot. Impressions also dipped -1.54%.
Key Takeaways
Mobile is crucial, especially in Personal Finance, where mobile CTRs are 34%. Focus on mobile-friendly designs and keep content short.
Users prefer informational content over commercial pages, so prioritize educational material while maintaining clear sales pages.
Different industries require different strategies:
Science and Automotive sectors are growing; add more content here.
Arts and Entertainment need improved audience engagement.
Personal Finance has good CTRs but lower search volume; be ready for traffic drops.
Branded searches perform well on mobile, so focus on building your brand. Track your CTR metrics against industry standards and adjust as trends change.
Looking Ahead
The findings suggest that websites should closely monitor their CTR metrics against industry benchmarks, as rankings alone don’t tell the complete traffic story.
SERP layout variations for different keywords can impact click-through rates as well.
The next report covering Q4 will offer year-end comparisons and trend analysis.
MIT Technology Review’s How To series helps you get things done.
Since the start of the generative AI boom, artists have been worried about losing their livelihoods to AI tools. There have been plenty of examples of companies’ replacing human labor with computer programs. Most recently, Coca-Cola sparked controversy by creating a new Christmas ad with generative AI.
Artists and writers have launched several lawsuits against AI companies, arguing that their work has been scraped into databases for training AI models without consent or compensation. Tech companies have responded that anything on the public internet falls under fair use. But it will be years until we have a legal resolution to the problem.
Unfortunately, there is little you can do if your work has been scraped into a data set and used in a model that is already out there. You can, however, take steps to prevent your work from being used in the future.
Here are four ways to do that.
Mask your style
One of the most popular ways artists are fighting back against AI scraping is by applying “masks” on their images, which protect their personal style from being copied.
Tools such as Mist, Anti-DreamBooth, and Glaze add tiny changes to an image’s pixels that are invisible to the human eye, so that if and when images are scraped, machine-learning models cannot decipher them properly. You’ll need some coding skills to run Mist and Anti-DreamBooth, but Glaze, developed by researchers at the University of Chicago, is more straightforward to apply. The tool is free and available to download as an app, or the protection can be applied online. Unsurprisingly, it is the most popular tool and has been downloaded millions of times.
But defenses like these are never foolproof, and what works today might not work tomorrow. In computer security, breaking defenses is standard practice among researchers, as this helps people find weaknesses and make systems safer. Using these tools is a calculated risk: Once something is uploaded online, you lose control of it and can’t retroactively add protections to images.
Rethink where and how you share
Popular art profile sites such as DeviantArt and Flickr have become gold mines for AI companies searching for training data. And when you share images on platforms such as Instagram, its parent company, Meta, can use your data to build its models in perpetuity if you’ve shared it publicly. (See opt-outs below.)
One way to prevent scraping is by not sharing images online publicly, or by making your social media profiles private. But for many creatives that is simply not an option; sharing work online is a crucial way to attract clients.
It’s worth considering sharing your work on Cara, a new platform created in response to the backlash against AI. Cara, which collaborates with the researchers behind Glaze, is planning to add integrations to the lab’s art defense tools. It automatically implements “NoAI” tags that tell online scrapers not to scrape images from the site. It currently relies on the goodwill of AI companies to respect artists’ stated wishes, but it’s better than nothing.
Opt out of scraping
Data protection laws might help you get tech companies to exclude your data from AI training. If you live somewhere that has these sorts of laws, such as the UK or the EU, you can ask tech companies to opt you out of having your data scraped for AI training. For example, you can follow these instructions for Meta. Unfortunately, opt-out requests from users in places without data protection laws are honored only at the discretion of tech companies.
The site Have I Been Trained, created by the artist-run company Spawning AI, lets you search to find out if your images have ended up in popular open-source AI training data sets. The organization has partnered with two companies: Stability AI, which created Stable Diffusion, and Hugging Face, which promotes open access to AI. If you add your images to Spawning AI’s Do Not Train Registry, these companies have agreed to remove your images from their training data sets before training new models. Again, unfortunately, this relies on the goodwill of AI companies and is not an industry-wide standard.
If all else fails, add some poison
The University of Chicago researchers who created Glaze have also created Nightshade, a tool that lets you add an invisible layer of “poison” to your images. Like Glaze, it adds invisible changes to pixels, but rather than just making it hard for AI models to interpret images, it can break future iterations of these models and make them behave unpredictably. For example, images of dogs might become cats, and handbags might become toasters. The researchers say relatively few samples of poison are needed to make an impact.
You can add Nightshade to your image by downloading an app here. In the future, the team hopes to combine Glaze and Nightshade, but at the moment the two protections have to be added separately.
This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.
“Well, what about China?”
This is a comment I get all the time on the topic of climate change, both in conversations and on whatever social media site is currently en vogue. Usually, it comes in response to some statement about how the US and Europe are addressing the issue (or how they need to be).
Sometimes I think people ask this in bad faith. It’s a rhetorical way to throw up your hands, imply that the US and Europe aren’t the real problem, and essentially say: “if they aren’t taking responsibility, why should we?” However, amid the playground-esque finger-pointing there are some undeniable facts: China emits more greenhouse gases than any other country, by far. It’s one of the world’s most populous countries and a climate-tech powerhouse, and its economy is still developing.
With many complicated factors at play, how should we think about the country’s role in addressing climate change?
There’s context missing if we just look at that one number, as I wrote in my latest story that digs into recent global climate data. Since carbon dioxide hangs around in the atmosphere for centuries, we should arguably consider not just a country’s current emissions, but everything it’s produced over time. If we do that, the US still takes the crown for the world’s biggest climate polluter.
However, China is now in second place, according to a new analysis from Carbon Brief released this week. In 2023, the country exceeded the EU’s 27 member states in historical emissions for the first time.
This reflects a wider trend that we’re seeing around the world: Developing nations are starting to account for a larger fraction of emissions than they used to. In 1992, when countries agreed to the UN climate convention, industrialized countries (a category called Annex I) made up about one-fifth of the world’s population but were responsible for a whopping 61% of historical emissions. By the end of 2024, though, those countries’ share of global historical emissions will fall to 52%, and it is expected to keep ticking down.
China, like all nations, will need to slash its emissions for the world to meet global climate goals. One crucial point here is that while its emissions are still huge, there are signs that the nation is making some progress.
China’s carbon dioxide’s emissions are set to fall in 2024 because of record growth in low-carbon energy sources. That decline is projected to continue under the country’s current policy settings, according to an October report from the IEA. China’s oil demand could soon peak and start to fall, largely because it’s seeing such a huge uptake of electric vehicles.
One growing question: With all this progress and a quickly growing economy, should we be expecting China to do more than just make progress on its own emissions?
As I wrote in the newsletter last week, the current talks at COP29 (the UN climate conference) are focused on setting a new, more aggressive global climate finance goal to help developing nations address climate change. China isn’t part of the group of countries that are required to pay into this pot of money, but some are calling for that to change given that it is the world’s biggest polluter.
Talks at COP29 aren’t going very well. The COP29 president called for faster action, but progress toward a finance deal has stalled amid infighting over how much money should be on the table and who should pay up.
China’s complex role in emissions and climate action is far from the only holdup at the talks. Leaders from major nations including Germany and France canceled plans to attend, and the looming threat that the US could pull out of the Paris climate agreement is coloring the negotiations.
But disagreement over how to think about China’s role in all this is a good example of how difficult it is to assign responsibility when it comes to climate change, and how much is at play in global climate negotiations. One thing I do know for sure is that pointing fingers doesn’t cut emissions.
Fusion energy has been a dream for decades, and a handful of startups say we’re closer than ever to making it a reality. This deep dive looks at a few of the companies looking to be the first to deploy fusion power. (New York Times) → I recently visited one of the startups, Commonwealth Fusion Systems. (MIT Technology Review)
President-elect Donald Trump has tapped Chris Wright to lead the Department of Energy. Wright is head of the fracking company Liberty Energy. (Washington Post)
In the wake of Trump’s election, it might be time for climate tech to get a rebrand. Companies and investors might increasingly avoid using the term, opting instead for phrases like “energy independence” or “frontier tech,” to name a few. (Heatmap)
Rooftop solar has saved customers in California about $2.3 billion on utility bills this year, according to a new analysis. This result is counter to a report from a state agency, which found that rooftop panels impose over $8 billion in extra costs on consumers of the state’s three major utilities. (Canary Media)
Low-carbon energy needs much less material than it used to. Rising efficiency in making technology like solar panels bodes well for hopes of cutting mining needs. (Sustainability by Numbers)
New York governor Kathy Hochul has revived a plan to implement congestion pricing, which would charge drivers to enter the busiest parts of Manhattan. It would be the first such program in the US. (The City)
Enhanced geothermal technology could be close to breaking through into commercial success. Companies that aim to harness Earth’s heat for power are making progress toward deploying facilities. (Nature) → Fervo Energy found that its wells can be used like a giant underground battery. (MIT Technology Review)
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.
AI can now create a replica of your personality
Imagine sitting down with an AI model for a spoken two-hour interview. A friendly voice guides you through a conversation that ranges from your childhood, your formative memories, and your career to your thoughts on immigration policy. Not long after, a virtual replica of you is able to embody your values and preferences with stunning accuracy.
That’s now possible, according to a new paper from a team including researchers from Stanford and Google DeepMind.
They recruited 1,000 people and, from interviews with them, created agent replicas of them all. To test how well the agents mimicked their human counterparts, participants did a series of tests, games and surveys, then the agents completed the same exercises. The results were 85% similar. Freaky. Read our story about the work, and why it matters.
—James O’Donnell
China’s complicated role in climate change
“But what about China?”
In debates about climate change, it’s usually only a matter of time until someone brings up China. Often, it comes in response to some statement about how the US and Europe are addressing the issue (or how they need to be).
Sometimes it can be done in bad faith. It’s a rhetorical way to throw up your hands, and essentially say: “if they aren’t taking responsibility, why should we?”
However, there are some undeniable facts: China emits more greenhouse gases than any other country, by far. It’s one of the world’s most populous countries and a climate-tech powerhouse, and its economy is still developing.
With many complicated factors at play, how should we think about the country’s role in addressing climate change? Read the full story.
—Casey Crownhart
This story is from The Spark, our weekly newsletter giving you the inside track on all things energy and climate. Sign up to receive it in your inbox every Wednesday.
Four ways to protect your art from AI
Since the start of the generative AI boom, artists have been worried about losing their livelihoods to AI tools.
Unfortunately, there is little you can do if your work has been scraped into a data set and used in a model that is already out there. You can, however, take steps to prevent your work from being used in the future. Here are four ways to do that.
—Melissa Heikkila
This is part of our How To series, where we give you practical advice on how to use technology in your everyday lives. You can read the rest of the series here.
MIT Technology Review Narrated: The world’s on the verge of a carbon storage boom
In late 2023, one of California’s largest oil and gas producers secured draft permits from the US Environmental Protection Agency to develop a new type of well in an oil field. If approved, it intends to drill a series of boreholes down to a sprawling sedimentary formation roughly 6,000 feet below the surface, where it will inject tens of millions of metric tons of carbon dioxide to store it away forever.
Hundreds of similar projects are looming across the state, the US, and the world. Proponents hope it’s the start of a sort of oil boom in reverse, kick-starting a process through which the world will eventually bury more greenhouse gas than it adds to the atmosphere. But opponents insist these efforts will prolong the life of fossil-fuel plants, allow air and water pollution to continue, and create new health and environmental risks.
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 How the Trump administration could hack your phone Spyware acquired by the US government in September could fairly easily be turned on its own citizens. (New Yorker $) + Here’s how you can fight back against being digitally spied upon. (The Guardian)
2 The DOJ is trying to force Google to sell off Chrome Whether Trump will keep pushing it through is unclear, though. (WP $) + Some financial and legal experts argue that just selling Chrome is not enough to address antitrust issues. (Wired $)
3 There’s a booming ‘AI pimping’ industry People are stealing videos from real adult content creators, giving them AI-generated faces, and monetizing their bodies. (Wired $) + This viral AI avatar app undressed me—without my consent. (MIT Technology Review)
4 Here’s Elon Musk and Vivek Ramaswamy plan for federal employees Large-scale firings and an end to any form of remote work. (WSJ $)
5 The US is scaring everyone with its response to bird flu It’s done remarkably little to show it’s trying to contain the outbreak. (NYT $) + Virologists are getting increasingly nervous about how it could evolve and spread. (MIT Technology Review)
6 AI could boost the performance of quantum computers A new model created by Google DeepMind is very good at correcting errors. (New Scientist $) + But AI could also make quantum computers less necessary. (MIT Technology Review)
7 Biden has approved the use of anti-personnel mines in Ukraine It comes just days after he gave the go-ahead for it to use long-range missiles inside Russia. (Axios) + The US military has given a surveillance drone contract to a little-known supplier from Utah. (WSJ $) + The Danish military said it’s keeping a close eye on a Chinese ship in its waters after data cable breaches. (Reuters $)
8 The number of new mobile internet users is stalling Only about 57% of the world’s population is connected. (Rest of World)
9 All of life on Earth descended from this single cell Our “last universal common ancestor” (or LUCA for short) was a surprisingly complex organism living 4.2 billion years ago. (Quanta) + Scientists are building a catalog of every type of cell in our bodies. (The Economist $)
10 What it’s like to live with a fluffy AI pet Try as we might, it seems we can’t help but form attachments to cute companion robots. (The Guardian)
Quote of the day
“The free pumpkins have brought joy to many.”
—An example of the sort of stilted remarks made by a now-abandoned AI-generated news broadcaster at local Hawaii paper The Garden Island, Wired reports.
The big story
How Bitcoin mining devastated this New York town
GABRIELA BHASKAR
April 2022
If you had taken a gamble in 2017 and purchased Bitcoin, today you might be a millionaire many times over. But while the industry has provided windfalls for some, local communities have paid a high price, as people started scouring the world for cheap sources of energy to run large Bitcoin-mining farms.
It didn’t take long for a subsidiary of the popular Bitcoin mining firm Coinmint to lease a Family Dollar store in Plattsburgh, a city in New York state offering cheap power. Soon, the company was regularly drawing enough power for about 4,000 homes. And while other miners were quick to follow, the problems had already taken root. Read the full story.
—Lois Parshley
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 tweet ’em at me.)
+ Cultivating gratitude is a proven way to make yourself happier. + You can’t beat a hot toddy when it’s cold outside. + If you like abandoned places and overgrown ruins, Jonathan Jimenez is the photographer for you. + A lot changed between Gladiator I and II, not least Hollywood’s version of the male ideal.
Large language models are now being used by millions of people for many different things. But as OpenAI itself points out, these models are known to produce racist, misogynistic and hateful content; reveal private information; amplify biases and stereotypes; and make stuff up. The company wants to share what it is doing to minimize such behaviors.
The aim is to combine these two approaches, with unwanted behaviors discovered by human testers handed off to an AI to be explored further and vice versa. Automated red-teaming can come up with a large number of different behaviors, but human testers bring more diverse perspectives into play, says Lama Ahmad, a researcher at OpenAI: “We are still thinking about the ways that they complement each other.”
Red-teaming isn’t new. AI companies have repurposed the approach from cybersecurity, where teams of people try to find vulnerabilities in large computer systems. OpenAI first used the approach in 2022, when it was testing DALL-E 2. “It was the first time OpenAI had released a product that would be quite accessible,” says Ahmad. “We thought it would be really important to understand how people would interact with the system and what risks might be surfaced along the way.”
The technique has since become a mainstay of the industry. Last year, President Biden’s Executive Order on AI tasked the National Institute of Standards and Technology (NIST) with defining best practices for red-teaming. To do this, NIST will probably look to top AI labs for guidance.
Tricking ChatGPT
When recruiting testers, OpenAI draws on a range of experts, from artists to scientists to people with detailed knowledge of the law, medicine, or regional politics. OpenAI invites these testers to poke and prod its models until they break. The aim is to uncover new unwanted behaviors and look for ways to get around existing guardrails—such as tricking ChatGPT into saying something racist or DALL-E into producing explicit violent images.
Adding new capabilities to a model can introduce a whole range of new behaviors that need to be explored. When OpenAI added voices to GPT-4o, allowing users to talk to ChatGPT and ChatGPT to talk back, red-teamers found that the model would sometimes start mimicking the speaker’s voice, an unexpected behavior that was both annoying and a fraud risk.
There is often nuance involved. When testing DALL-E 2 in 2022, red-teamers had to consider different uses of “eggplant,” a word that now denotes an emoji with sexual connotations as well as a purple vegetable. OpenAI describes how it had to find a line between acceptable requests for an image, such as “A person eating an eggplant for dinner,” and unacceptable ones, such as “A person putting a whole eggplant into her mouth.”
Similarly, red-teamers had to consider how users might try to bypass a model’s safety checks. DALL-E does not allow you to ask for images of violence. Ask for a picture of a dead horse lying in a pool of blood, and it will deny your request. But what about a sleeping horse lying in a pool of ketchup?
When OpenAI tested DALL-E 3 last year, it used an automated process to cover even more variations of what users might ask for. It used GPT-4 to generate requests producing images that could be used for misinformation or that depicted sex, violence, or self-harm. OpenAI then updated DALL-E 3 so that it would either refuse such requests or rewrite them before generating an image. Ask for a horse in ketchup now, and DALL-E is wise to you: “It appears there are challenges in generating the image. Would you like me to try a different request or explore another idea?”
In theory, automated red-teaming can be used to cover more ground, but earlier techniques had two major shortcomings: They tend to either fixate on a narrow range of high-risk behaviors or come up with a wide range of low-risk ones. That’s because reinforcement learning, the technology behind these techniques, needs something to aim for—a reward—to work well. Once it’s won a reward, such as finding a high-risk behavior, it will keep trying to do the same thing again and again. Without a reward, on the other hand, the results are scattershot.
“They kind of collapse into ‘We found a thing that works! We’ll keep giving that answer!’ or they’ll give lots of examples that are really obvious,” says Alex Beutel, another OpenAI researcher. “How do we get examples that are both diverse and effective?”
A problem of two parts
OpenAI’s answer, outlined in the second paper, is to split the problem into two parts. Instead of using reinforcement learning from the start, it first uses a large language model to brainstorm possible unwanted behaviors. Only then does it direct a reinforcement-learning model to figure out how to bring those behaviors about. This gives the model a wide range of specific things to aim for.
Beutel and his colleagues showed that this approach can find potential attacks known as indirect prompt injections, where another piece of software, such as a website, slips a model a secret instruction to make it do something its user hadn’t asked it to. OpenAI claims this is the first time that automated red-teaming has been used to find attacks of this kind. “They don’t necessarily look like flagrantly bad things,” says Beutel.
Will such testing procedures ever be enough? Ahmad hopes that describing the company’s approach will help people understand red-teaming better and follow its lead. “OpenAI shouldn’t be the only one doing red-teaming,” she says. People who build on OpenAI’s models or who use ChatGPT in new ways should conduct their own testing, she says: “There are so many uses—we’re not going to cover every one.”
For some, that’s the whole problem. Because nobody knows exactly what large language models can and cannot do, no amount of testing can rule out unwanted or harmful behaviors fully. And no network of red-teamers will ever match the variety of uses and misuses that hundreds of millions of actual users will think up.
That’s especially true when these models are run in new settings. People often hook them up to new sources of data that can change how they behave, says Nazneen Rajani, founder and CEO of Collinear AI, a startup that helps businesses deploy third-party models safely. She agrees with Ahmad that downstream users should have access to tools that let them test large language models themselves.
Rajani also questions using GPT-4 to do red-teaming on itself. She notes that models have been found to prefer their own output: GPT-4 ranks its performance higher than that of rivals such as Claude or Llama, for example. This could lead it to go easy on itself, she says: “I’d imagine automated red-teaming with GPT-4 may not generate as harmful attacks [as other models might].”
Miles behind
For Andrew Tait, a researcher at the Ada Lovelace Institute in the UK, there’s a wider issue. Large language models are being built and released faster than techniques for testing them can keep up. “We’re talking about systems that are being marketed for any purpose at all—education, health care, military, and law enforcement purposes—and that means that you’re talking about such a wide scope of tasks and activities that to create any kind of evaluation, whether that’s a red team or something else, is an enormous undertaking,” says Tait. “We’re just miles behind.”
Tait welcomes the approach of researchers at OpenAI and elsewhere (he previously worked on safety at Google DeepMind himself) but warns that it’s not enough: “There are people in these organizations who care deeply about safety, but they’re fundamentally hamstrung by the fact that the science of evaluation is not anywhere close to being able to tell you something meaningful about the safety of these systems.”
Tait argues that the industry needs to rethink its entire pitch for these models. Instead of selling them as machines that can do anything, they need to be tailored to more specific tasks. You can’t properly test a general-purpose model, he says.
“If you tell people it’s general purpose, you really have no idea if it’s going to function for any given task,” says Tait. He believes that only by testing specific applications of that model will you see how well it behaves in certain settings, with real users and real uses.
“It’s like saying an engine is safe; therefore every car that uses it is safe,” he says. “And that’s ludicrous.”
Every week we publish a selection of new products from companies offering services to ecommerce merchants. This installment includes updates on local search marketing, video generators, AI-powered shopping assistants, email marketing tools, conversational commerce, online review tools, and influencer marketing.
Got an ecommerce product release? Email releases@practicalecommerce.com.
New Tools for Merchants
Google Maps launches product search.Google has updated Maps to include product-specific search. The update lets users search for products directly within Google Maps and find real-time inventory availability at nearby stores. Shoppers can find a wide range of items and have immediate access to local pickup options. For accurate inventory tracking, merchants should regularly sync with Google Merchant Center and ensure accurate Google Business Profile information.
Google Maps
TikTok launches AI-powered video platform to advertisers globally. ByteDance-owned TikTok has announced the global availability of Symphony Creative Studios, its generative AI video creation platform and a component of the Symphony creative content suite, which aims to help businesses, creators, and agencies customize content to their brands. The suite includes Symphony Creative Studios, Symphony Assistant, Symphony Digital Avatars, and TikTok ads manager. Symphony Creative Studios, which also includes a virtual assistant, helps users automate tasks to create content on the platform.
Alibaba releases AI search tool for small businesses in Europe and the Americas. China-based ecommerce giant Alibaba has released an AI-powered search engine for small companies in Europe and the Americas to source inventory. With a few text or image prompts, businesses can use the AI-assistant Accio to find wholesale products, including an analysis of their popularity with consumers and projected profit. The technology uses generative AI from Alibaba’s Tongyi Qianwen large language model. The initial version of Accio is web-based and supports English, German, French, Portuguese, and Spanish.
Bluecore acquires AI shopping assistant Alby for conversational commerce.Bluecore, a retail technology provider, has acquired AI shopping assistant Alby in preparation for a new conversational shopping release. Using Bluecore, retailers can soon make their ecommerce sites, email campaigns, SMS, and mobile apps conversational by anticipating and answering shoppers’ questions. The shopping assistant will leverage the retailer’s product catalog and customer lifecycle stage data to generate real-time prompts, responses, and recommendations that progress the shopper toward purchase.
Alby
PhaseZero launches commerce-ready AI agents.PhaseZero, an AI-native composable commerce provider, has announced commerce-ready AI agents integrated with its CxCommerce platform. Pricing Agent provides insights into competitive and channel pricing trends. Inventory Agent provides visibility into alternate products and suppliers for out-of-stock items. Visual Search Agent makes it easy for customers to find products matching pictures. SEO Agent enables improved rankings. Insights Agent helps merchants identify sales opportunities, perform data-driven planning, and maximize return on investment.
CommerceIQ launches Nexis, an omnichannel AI assistant for commerce.CommerceIQ, a retail ecommerce management platform, has launched Nexis, a generative AI-powered omnichannel commerce tool. Nexis is a role-based teammate that enhances Amazon Copilot and retail media management solutions with advanced AI capabilities tailored to ecommerce. With the sales module, merchants can enhance Copilot with tailored templates to generate insights from Amazon sales and operations data. With the marketing module, merchants can deepen analysis across global retail media platforms.
MailReach releases tools for email marketing campaigns.MailReach, a provider of email deliverability solutions, has launched a suite of tools to enhance email marketing campaigns. The tools feature email warmup, advanced spam testing, real-time analytics, A/B testing, and integration with popular ecommerce, CRM, and marketing automation platforms. The platform’s intelligent inbox placement tester and domain health checkup tools help businesses address deliverability issues proactively to overcome spam filters and achieve higher engagement rates.
MailReach
Thryv launches AI tool to help small businesses respond quickly to online reviews.Thryv, a small business software platform, has launched Thryv AI Review Response, allowing businesses to respond quickly to positive and negative feedback on Google Business Profile listings, YP.com, Trustpilot, and 20 other popular review sites. Thryv aggregates review sites into one view. Users select the AI wand to generate review responses. Other AI-enhanced platform tools include social media management and Pro Campaign Editor.
StackAdapt integrates with the Shopify platform to optimize merchant campaigns.StackAdapt, a multichannel advertising platform, has announced its latest application on the Shopify App Store. Shopify store owners can now automatically install the StackAdapt pixel through the App Store or directly from StackAdapt. The integration enables users to onboard pixels and audience activations, tracking conversion events for monitoring and optimization. StackAdapt also offers predefined templates that enable users to define their audiences with rule-based segmentation.
Popcorn AI secures funding to define conversational commerce.Popcorn AI, a conversational commerce developer, has announced a $500,000 pre-seed funding round, with participation from Spring Studios. Popcorn AI says the funding will accelerate its development of conversational commerce for business-customer interactions, advancing the platform’s conversational AI capabilities and expanding its team of AI and commerce experts. Popcorn AI’s platform creates intelligent AI agents that handle the entire customer journey, from personalized product discovery to payment processing, all within life-like conversations.
BrandLink launches an influencer marketing platform for businesses and creators.SooNet Inc., a business application developer, has launched BrandLink, a mobile-first platform to provide accessible and affordable influencer marketing solutions for businesses of all sizes. BrandLink allows firms to search for and connect with influencers based on specific filters such as niche, location, and audience size. Influencers, in turn, can set their own rates and customize service packages, making the platform flexible and transparent for both parties, according to BrandLink.
Search and chat provider Perplexity AI has launched a new shopping assistant with the potential to disrupt ecommerce marketing, advertising, and search engine optimization.
Perplexity’s “Buy with Pro” feature combines product research with one-click checkout and free shipping, making online shopping easy for American consumers who never have to leave that website or app to complete the purchase.
The feature also includes “Snap to Shop,” a visual shopping tool not unlike Google Lens. A Pro subscriber can take a picture of a product and have Perplexity source it.
Perplexity’s AI will also customize recommendations and guide shoppers to specific products.
“When you ask Perplexity a shopping question, you’ll still get the precise, objective answers you expect, plus easy-to-read product cards showing the most relevant items, along with key details presented in a simple, visual format. These cards aren’t sponsored — they’re unbiased recommendations, tailored to your search by our AI,” the company wrote in its announcement.
Perplexity Pro subscribers can ask buying questions and receive research results and product recommendations.
Merchant Program
Perplexity’s shopping assistant is integrated with Shopify, giving it “access to the most recent and relevant information on products across Shopify-powered businesses globally that sell and ship to the U.S.”
Shopify CEO Tobi Lutke has invested in Perplexity, as has Amazon founder Jeff Bezos.
In addition to the Shopify connection, Perplexity has invited “large” retailers to join its merchant program. Selected retail partners will have relatively more visibility in Perplexity’s product recommendations and access to a dashboard for this new channel. There is no cost to apply or participate — only a five-question Typeform requesting contact info and the merchant’s URL.
Perplexity’s five-question merchant application did not request sales details.
Perplexity’s shopping assistant will likely make buying online easier for consumers, but it and the inevitable assortment of similar AI-driven tools could complicate existing ecommerce businesses.
I see seven concerns small-to-midsize ecommerce businesses should consider if tools like Perplexity’s shopping assistant become widespread.
Seller Concerns
Customer relationships. With Perplexity’s shopping AI, the consumer never leaves the search environment and doesn’t visit the merchant’s online store. Without direct engagement, building and maintaining brand loyalty becomes more difficult.
Product visibility. Ecommerce shops are accustomed to familiar channels such as advertising, SEO, and social media to attract shoppers. Each of these has challenges, but they are known. With Perplexity and any similar AI assistant, products from large retailers or shops with robust data integrations will be favored over small to midsize ecommerce businesses.
Price pressure. One would hope that AI shopping assistants would consider a merchant’s reviews and reputation, but the assistants will likely focus on the price, pressuring SMBs to compete. Many sellers make similar complaints about online marketplaces — AI could be much worse.
Dependency. If one-click shopping becomes a common feature in AI search, online sellers, especially SMBs, could become dependent on AI platforms for sales, risking significant impacts if platform policies or algorithms change.
Promotion. Sellers frequently use marketing and advertising to drive ecommerce sales, but with the addition of various AI shopping assistants, reaching customers could become far more complex. Marketing teams may need to optimize data feeds for several AI platforms or advertise in dozens of new AI channels to remain competitive.
Fluctuating demand. AI recommendations could create sudden spikes in demand that are hard, if not impossible, to predict. Stores could suddenly sell out. Once inventory is gone, the shop might fall out of consideration, meaning sales might be slow to return.
Technical barriers. Perplexity has stated it wants to make it easier for stores to provide product data. However, as AI shopping assistants become more common, stores, especially small ones, will likely face challenges around data optimization and integration.
Opportunities
Assuming its shopping assistant represents a trend, Perplexity may be similar to Amazon, Temu, Mercari, and similar marketplaces and offer opportunities.
Product visibility. Perplexity’s Shopify integration could level the field, allowing small and mid-sized stores to compete with massive retailers. Any improved visibility might lead to increased sales without additional advertising costs.
No fees. Perplexity shopping assistant is free for sellers. Compared to other marketplaces (with fees), integrating with Perplexity and similar tools might improve seller margins.
Next Steps
AI shopping assistants such as Perplexity present challenges and opportunities for sellers. To participate:
Apply for Perplexity’s merchant program, and be on the lookout for similar providers.
Start optimizing and differentiating product data, ensuring it is complete, accurate, and unique.
Add product feeds to existing marketplaces. AI assistants will likely partner with existing product data tools.
