In the span of a few weeks, three separate groups of Google employees said versions of the same thing in three separate venues: people are no longer typing search queries the way they used to, and Google’s own systems are having to work harder because of it. Liz Reid talked about it on a Bloomberg podcast. Robby Stein and Mike Torres announced a product change that put it into practice inside Chrome. And Martin Splitt and Nikola Todorovic discussed it plainly on Google’s own Search Off The Record podcast.
Taken separately, each of these is a minor news item. Taken together, they describe a single, coherent shift in how search actually works in 2026 — and what it means for anyone who depends on organic visibility to run a business.
Google Admits People Never Wanted To Type Keywords
Google Search VP Liz Reid made an admission on the Bloomberg Odd Lots podcast that’s worth sitting with. For years, she said, people typing something like “restaurants New York” into Google actually had a far more specific need in mind — a place for five people, not too pricey, with a vegan option, that could accommodate kids. They just had no way to type that and expect a useful answer, so they compressed it down to two or three keywords and hoped Google’s ranking systems would guess the rest.
That compression is the entire reason keyword research became the foundation of SEO. Marketers learned to reverse-engineer what searchers actually meant by a short phrase, because the search box itself couldn’t handle anything more specific. Reid’s point is that AI Overviews and AI Mode remove that constraint. Users increasingly type the actual, messy, specific version of their need and expect Google to do the translation work that keyword-guessing used to require.
The catch, which Reid was candid about, is that this makes Google’s job harder, not easier. A complex, one-off natural language query can’t be cached and reused the way “restaurants New York” can, because a meaningful share of these queries are asked in a slightly different way by every person who asks them. Google’s response is something most SEOs have already run into without necessarily understanding the mechanism: query fan-out, where one long, specific prompt gets broken into several smaller, classic-search-style queries behind the scenes, each one run separately, with the results synthesized back into a single answer.
We’ve covered the mechanics of this fan-out process in detail in our Google AI Mode guide — AI Mode alone can fire off up to 16 of these sub-searches for a single user prompt. The practical implication is the one Reid’s comments make explicit: optimizing for long-tail phrasing as if it were one target keyword misses the point. Your content still needs to win the individual fan-out queries it might get matched against, which are often shorter and more specific than the original question, not longer.
Chrome’s AI Mode Update Is a Stress Test, Not a Verdict
A few weeks after Reid’s comments, Google Search VP Robby Stein and Chrome VP Mike Torres announced a Chrome update that puts this shift into a product. AI Mode in Chrome now lets users browse a webpage and keep chatting with Google’s AI side by side, without switching tabs or losing their place. It sounds like a minor convenience feature. It’s really a signal about where Google thinks the click belongs in the search journey.
Search Engine Journal contributor Greg Jarboe argued that this changes the sequence of discovery. A user can now get an AI-generated answer first, stay inside that AI layer, and open publisher pages only to verify or go deeper — meaning the click increasingly happens after trust is already established, not before. The scale of the shift behind that argument is significant: Ahrefs found that AI Overviews now correlate with a 58% reduction in click-through rate for top-ranking pages, roughly double the decline measured just a year earlier, and separate research from Index Exchange found publishers saw ad opportunity declines averaging 14% year-over-year through 2025.
None of that means SEO stopped working. It means weak SEO stopped being able to hide. Rand Fishkin’s analysis of roughly 400 websites that held up through what he called the industry’s zero-click reckoning found five shared traits among the survivors: a genuinely unique product or service, the ability to help users complete a task rather than just read about it, proprietary data or assets nobody else has, tight topical focus instead of covering everything loosely, and a brand strong enough that people seek it out by name. Fishkin’s blunter point was that no amount of technical polish saves a site whose entire business model is something AI can simply summarize and disintermediate.
Google Says This Is a New Wave, Not a Temporary Spike
The third data point removes any doubt that this is a passing trend. On Google’s own Search Off The Record podcast, Search engineering director Nikola Todorovic told colleague Martin Splitt directly that current search behavior represents a genuine step change, not a gradual evolution of what came before. His reasoning: users are actively discovering, in real time, that search can now handle questions they wouldn’t previously have bothered typing in — and that discovery process is itself driving average query length upward.
That framing matters because it reframes what’s often described as “SEO getting harder” as something closer to “the addressable set of queries getting larger.” Classic search hasn’t gone away — Todorovic and Splitt were both clear that organic retrieval still sits underneath every AI-generated answer, just moved into the background of the experience rather than the foreground. Content still has to be retrievable by classic search mechanics first, and only then does it get a chance to be useful inside a synthesized AI answer.
What This Actually Means If You Run a Website
Put the three announcements together and the practical picture is fairly clear, even if the underlying mechanics are new. Query volume and complexity are both increasing, which is a growth opportunity, not just a threat. But the content that captures that growth has to satisfy two separate layers at once: it needs to be structured and specific enough to win the smaller fan-out queries a complex question gets broken into, and it needs to contain something — a data point, a firsthand result, a specific detail — that a synthesized answer can’t fully substitute for a visit to the source.
This is also exactly why search intent research hasn’t become less important, it’s become more load-bearing. We’ve written before about matching content to what the reader is actually trying to accomplish rather than to a keyword string, and Reid’s comments are effectively Google confirming that this is now closer to how the underlying system actually works, not just good editorial advice. Pages that were built to rank for a phrase, without addressing the fuller need behind it, are the ones most exposed by this shift.
It’s worth checking your own analytics with this lens before assuming a traffic dip is an algorithm problem. We’ve covered this diagnostic step in more depth in Google Search Ranking Volatility 2026 — a page that’s losing clicks but holding impressions may not be losing rankings at all. It may be getting summarized inside an AI answer instead of clicked through, which shows up in your data very differently than a real ranking drop, and needs a different response.
Not sure whether your traffic changes are an AI Overview effect, a ranking issue, or something else entirely? Get in touch — this kind of diagnosis is exactly what we help clients work through before recommending any changes.
Sources referenced: Search Engine Journal, Google Search Central (blog.google), Ahrefs, Index Exchange, and SparkToro.
Sanjeev Kumar is a digital marketing expert with over 14 years of experience in SEO, PPC, content marketing, and online growth strategies. He specializes in search engine optimization, AI-driven marketing, and digital strategy for businesses and agencies worldwide.
