Microsoft CEO, Google Engineer Deflect AI Quality Complaints via @sejournal, @MattGSouthern

Within a week of each other, Microsoft CEO Satya Nadella and Jaana Dogan, a Principal Engineer working on Google’s Gemini API, posted comments about AI criticism that shared a theme. Both redirected attention away from whether AI output is “good” or “bad” and toward how people are reacting to the technology.

Nadella published “Looking Ahead to 2026” on his personal blog, writing that the industry needs to “get beyond the arguments of slop vs sophistication.”

Days later, Dogan posted on X that “people are only anti new tech when they are burned out from trying new tech.”

The timing coincides with Merriam-Webster naming “slop” its Word of the Year. For publishers, these statements can land less like reassurance and more like a request to stop focusing on quality.

Nadella Urges A Different Framing Than “AI Slop”

Nadella’s post argues that the conversation should move past the “slop” label and focus on how AI fits into human life and work. He characterizes AI as “cognitive amplifier tools” and believes that 2026 is the year in which AI must “prove its value in the real world.”

He writes: “We need to get beyond the arguments of slop vs sophistication,” and calls for “a new equilibrium” that accounts for humans having these tools. In the same section, he also calls it “the product design question we need to debate and answer,” which makes the point less about ending debate and more about steering it toward product integration and outcomes.

Dogan’s “Burnout” Framing Came Days After A Claude Code Post

Dogan’s post framed anti-AI sentiment as burnout from trying new technology. The line was blunt: “People are only anti new tech when they are burned out from trying new tech. It’s understandable.”

People are only anti new tech when they are burned out from trying new tech. It’s understandable.

— Jaana Dogan ヤナ ドガン (@rakyll) January 4, 2026

A few days earlier, Dogan had posted about using Claude Code to build a working prototype from a description of distributed agent orchestrators. She wrote that the tool produced something in about an hour that matched patterns her team had been building for roughly a year, adding: “In 2023, I believed these current capabilities were still five years away.”

I’m not joking and this isn’t funny. We have been trying to build distributed agent orchestrators at Google since last year. There are various options, not everyone is aligned… I gave Claude Code a description of the problem, it generated what we built last year in an hour.

— Jaana Dogan ヤナ ドガン (@rakyll) January 2, 2026

Replies to the “burnout” post pushed back on Dogan. Many responses pointed to forced integrations, costs, privacy concerns, and tools that feel less reliable within everyday workflows.

Dogan is a Principal Engineer on Google’s Gemini API and is not speaking as an official representative of Google policy.

The Standards Platforms Enforce On Publishers Still Matter

I’ve written E-E-A-T guides for Search Engine Journal for years. Those pieces reflected Google’s long-running expectation that publishers demonstrate experience, expertise, and trust, especially for “Your Money or Your Life” topics like health, finance, and legal content.

That’s why the current disconnect lands so sharply for publishers. Platforms have quality standards for ranking and visibility, while AI products increasingly present information directly to users with citations that can be difficult to evaluate at a glance.

When Google executives have been asked about declining click-through rates, the public framing has included “quality clicks” rather than addressing the volume loss publishers measure on their side.

What The Traffic Data Shows

Pew Research Center tracked 68,879 real Google searches. When AI Overviews appeared, only 8% of users clicked any link, compared to 15% when AI summaries did not appear. That works out to a 46.7% drop.

Publishers can be told the remaining clicks are higher intent, but volume still matters. It’s what drives ad impressions, subscriptions, and affiliate revenue.

Separately, Similarweb data indicates that the share of news-related searches that resulted in no click-through to news sites rose from 56% to 69%.

The crawl-to-referral imbalance adds another layer. Cloudflare has estimated Google Search at about a 14:1 crawl-to-referral ratio, compared with far higher ratios for OpenAI (around 1,700:1) and Anthropic (73,000:1).

Publishers have long operated on an implicit trade where they allow crawling in exchange for distribution and traffic. Many now argue that AI features weaken that trade because content can be used to answer questions without the same level of referral back to the open web.

Why This Matters

These posts from Nadella and Dogan help show how the AI quality debate may get handled in 2026.

When people are urged to move past “slop vs sophistication” or describe criticism as burnout, the conversation can drift away from accuracy, reliability, and the economic impact on publishers.

We see clear signs of traffic declines, and the crawling-to-referral ratios are also measurable. The economic impact is real.

Looking Ahead

Keep an eye out for more messaging that frames AI criticism as a user issue rather than a product- and economics-related issue.

I’m eager to see whether these companies make any changes to their product design in response to user feedback.

Featured Image: Jack_the_sparow/Shutterstock

Microsoft Explains How Duplicate Content Affects AI Search Visibility via @sejournal, @MattGSouthern

Microsoft has shared new guidance on duplicate content that’s aimed at AI-powered search.

The post on the Bing Webmaster Blog discusses which URL serves as the “source page” for AI answers when several similar URLs exist.

Microsoft describes how “near-duplicate” pages can end up grouped together for AI systems, and how that grouping can influence which URL gets pulled into AI summaries.

How AI Systems Handle Duplicates

Fabrice Canel and Krishna Madhavan, Principal Product Managers at Microsoft AI, wrote:

“LLMs group near-duplicate URLs into a single cluster and then choose one page to represent the set. If the differences between pages are minimal, the model may select a version that is outdated or not the one you intended to highlight.”

If multiple pages are interchangeable, the representative page might be an older campaign URL, a parameter version, or a regional page you didn’t mean to promote.

Microsoft also notes that many LLM experiences are grounded in search indexes. If the index is muddied by duplicates, that same ambiguity can show up downstream in AI answers.

How Duplicates Can Reduce AI Visibility

Microsoft lays out several ways duplication can get in the way.

One is intent clarity. If multiple pages cover the same topic with nearly identical copy, titles, and metadata, it’s harder to tell which URL best fits a query. Even when the “right” page is indexed, the signals are split across lookalikes.

Another is representation. If the pages are clustered, you’re effectively competing with yourself for which version stands in for the group.

Microsoft also draws a line between real page differentiation and cosmetic variants. A set of pages can make sense when each one satisfies a distinct need. But when pages differ only by minor edits, they may not carry enough unique signals for AI systems to treat them as separate candidates.

Finally, Microsoft links duplication to update lag. If crawlers spend time revisiting redundant URLs, changes to the page you actually care about can take longer to show up in systems that rely on fresh index signals.

Categories Of Duplicate Content Microsoft Highlights

The guidance calls out a few repeat offenders.

Syndication is one. When the same article appears across sites, identical copies can make it harder to identify the original. Microsoft recommends asking partners to use canonical tags that point to the original URL and to use excerpts instead of full reprints when possible.

Campaign pages are another. If you’re spinning up multiple versions targeting the same intent and differing only slightly, Microsoft recommends choosing a primary page that collects links and engagement, then using canonical tags for the variants and consolidating older pages that no longer serve a distinct purpose.

Localization comes up in the same way. Nearly identical regional pages can look like duplicates unless they include meaningful differences. Microsoft suggests localizing with changes that actually matter, such as terminology, examples, regulations, or product details.

Then there are technical duplicates. The guidance lists common causes such as URL parameters, HTTP and HTTPS versions, uppercase and lowercase URLs, trailing slashes, printer-friendly versions, and publicly accessible staging pages.

The Role Of IndexNow

Microsoft points to IndexNow as a way to shorten the cleanup cycle after consolidating URLs.

When you merge pages, change canonicals, or remove duplicates, IndexNow can help participating search engines discover those changes sooner. Microsoft links that faster discovery to fewer outdated URLs lingering in results, and fewer cases where an older duplicate becomes the page that’s used in AI answers.

Microsoft’s Core Principle

Canel and Madhavan wrote:

“When you reduce overlapping pages and allow one authoritative version to carry your signals, search engines can more confidently understand your intent and choose the right URL to represent your content.”

The message is consolidation first, technical signals second. Canonicals, redirects, hreflang, and IndexNow help, but they work best when you’re not maintaining a long tail of near-identical pages.

Why This Matters

Duplicate content isn’t a penalty by itself. The downside is weaker visibility when signals are diluted, and intent is unclear.

Syndicated articles can keep outranking the original if canonicals are missing or inconsistent. Campaign variants can cannibalize each other if the “differences” are mostly cosmetic. Regional pages can blend together if they don’t clearly serve different needs.

Routine audits can help you catch overlap early. Microsoft points to Bing Webmaster Tools as a way to spot patterns such as identical titles and other duplication indicators.

Looking Ahead

As AI answers become a more common entry point, the “which URL represents this topic” problem becomes harder to ignore.

Cleaning up near-duplicates can influence which version of your content gets surfaced when an AI system needs a single page to ground an answer.

Sam Altman Explains OpenAI’s Bet On Profitability via @sejournal, @martinibuster

In an interview with the Big Technology Podcast, Sam Altman seemed to struggle answering the tough questions about OpenAI’s path to profitability.

At about the 36 minute mark the interviewer asked the big question about revenues and spending. Sam Altman said OpenAI’s losses are tied to continued increases in training costs while revenue is growing. He said the company would be profitable much earlier if it were not continuing to grow its training spend so aggressively.

Altman said concern about OpenAI’s spending would be reasonable only if the company reached a point where it had large amounts of computing it could not monetize profitably.

The interviewer asked:

“Let’s, let’s talk about numbers since you brought it up. Revenue’s growing, compute spend is growing, but compute spend still outpaces revenue growth. I think the numbers that have been reported are OpenAI is supposed to lose something like 120 billion between now and 2028, 29, where you’re going to become profitable.

So talk a little bit about like, how does that change? Where does the turn happen?”

Sam Altman responded:

“I mean, as revenue grows and as inference becomes a larger and larger part of the fleet, it eventually subsumes the training expense. So that’s the plan. Spend a lot of money training, but make more and more.

If we weren’t continuing to grow our training costs by so much, we would be profitable way, way earlier. But the bet we’re making is to invest very aggressively in training these big models.”

At this point the interviewer pressed Altman harder about the path to profitability, this time mentioning the spending commitment of $1.4 trillion dollars versus the $20 billion dollars in revenue. This was not a softball question.

The interviewer pushed back:

“I think it would be great just to lay it out for everyone once and for all how those numbers are gonna work.”

Sam Altman’s first attempt to answer seemed to stumble in a word salad kind of way: 

“It’s very hard to like really, I find that one thing I certainly can’t do it and very few people I’ve ever met can do it.

You know, you can like, you have good intuition for a lot of mathematical things in your head, but exponential growth is usually very hard for people to do a good quick mental framework on.

Like for whatever reason, there were a lot of things that evolution needed us to be able to do well with math in our heads. Modeling exponential growth doesn’t seem to be one of them.”

Altman then regained his footing with a more coherent answer:

“The thing we believe is that we can stay on a very steep growth curve of revenue for quite a while. And everything we see right now continues to indicate that we cannot do it if we don’t have the compute.

Again, we’re so compute constrained, and it hits the revenue line so hard that I think if we get to a point where we have like a lot of compute sitting around that we can’t monetize on a profitable per unit of compute basis, it’d be very reasonable to say, okay, this is like a little, how’s this all going to work?

But we’ve penciled this out a bunch of ways. We will of course also get more efficient on like a flops per dollar basis, as you know, all of the work we’ve been doing to make compute cheaper comes to pass.

But we see this consumer growth, we see this enterprise growth. There’s a whole bunch of new kinds of businesses that, that we haven’t even launched yet, but will. But compute is really the lifeblood that enables all of this.

We have always been in a compute deficit. It has always constrained what we’re able to do.

I unfortunately think that will always be the case, but I wish it were less the case, and I’d like to get it to be less of the case over time, because I think there’s so many great products and services that we can deliver, and it’ll be a great business.”

The interviewer then sought to clarify the answer, asking:

“And then your expectation is through things like this enterprise push, through things like people being willing to pay for ChatGPT through the API, OpenAI will be able to grow revenue enough to pay for it with revenue.”

Sam Altman responded:

“Yeah, that is the plan.”

Altman’s comments define a specific threshold for evaluating whether OpenAI’s spending is a problem. He points to unused or unmonetizable computing power as the point at which concern would be justified, rather than current losses or large capital commitments.

In his explanation, the limiting factor is not willingness to pay, but how much computing capacity OpenAI can bring online and use. The follow-up question makes that explicit, and Altman’s confirmation makes clear that the company is relying on revenue growth from consumer use, enterprise adoption, and additional products to cover its costs over time.

Altman’s path to profitability rests on a simple bet: that OpenAI can keep finding buyers for its computing as fast as it can build it. Eventually, that bet either keeps winning or the chips run out.

Watch the interview starting at about the 36 minute mark:

Featured Image/Screenshot