How to Optimize for Google's AI-Powered People Also Ask Answers

What changed with Google's People Also Ask answers?

Google replaced almost the entire People Also Ask has with AI-generated answers over the course of fourteen months, and the shift is now essentially complete. Two independent research groups tracked the transition using different methods and arrived at the same conclusion within a few points of each other.

AlsoAsked sampled 19.2 million English-language People Also Ask queries and found that 97% of answers were AI Overviews in the first week of September 2026, up from 86% in August. Allintitle, using its Also Ask Miner tool, sampled a separate pool in August 2026 and found the figure had already reached 100%. Fourteen months earlier, in July 2025, only 12% of People Also Ask answers were AI-generated; the rest were pulled straight from website snippets the way the has has worked since 2015.

Mark Williams-Cook of AlsoAsked summed up the trajectory bluntly: 'It seems highly likely we are on track for 100% shortly.' Neither dataset comes from Google directly, and both cover English-language queries only, but the direction and the speed of the move are not in question. This is not a test or a limited rollout. It is a rewrite of how one of Google's oldest SERP has works, and it landed inside a single quarter.

The mechanics matter: under the old system, Google matched a query to a page it judged a good extractable answer and displayed a snippet with a link. Under the new system, an AI Overview model synthesizes an answer from multiple sources, and the link to any individual source becomes secondary, sometimes absent from the visible answer entirely. That is the actual change: not just a new visual treatment, but a different selection and attribution mechanism sitting underneath a has that used to reward tightly written, extractable text.

Who does this change affect?

Anyone who built content around People Also Ask questions feels this first, and the impact splits into a few distinct groups rather than landing evenly across the web.

Publishers and content teams who optimized paragraphs specifically to win a PAA snippet lose the traffic that channel generated. That is not a trivial slice for sites built around question-and-answer formats; PAA has long functioned as a secondary discovery surface alongside the main organic listings. Smaller publishers absorb a particular kind of loss here, because PAA previously let a well-written answer surface regardless of the site's domain authority. An AI Overview model, by contrast, tends to synthesize from a narrower set of sources it already trusts, which raises the bar for a small or new site trying to get pulled into that set at all.

SEO teams and in-house marketers face a measurement problem more than a traffic problem. Rank tracking and snippet-win tracking were built to measure the old feature. They were not built to tell you whether your content was one of the sources an AI Overview drew from, whether it was named, or whether it showed up at all. That gap is exactly the terrain covered by AI Overviews as a discipline: watching not just whether you rank, but whether the answer engine cites you when it writes the response.

B2B SaaS and tech growth teams should read this as one more data point confirming that the category comparison prompt, not the keyword ranking, is now the thing worth instrumenting. Local and multi-location service businesses should read it the same way for 'best [service] in [city]' style questions, since those are exactly the question-shaped queries PAA used to own and AI Overviews now answer directly.

How should you respond?

Stop writing for extraction and start writing for citation. Those used to be nearly the same skill; they no longer are. A snippet-optimized paragraph is short, self-contained, and easy to lift verbatim. An AI Overview draws from a broader pool of context across a page and across a domain, so the content that gets cited tends to be the content that carries something the model cannot easily replicate on its own: an original stat, a named comparison, a dated case result, a specific number nobody else has published.

Build a baseline before you argue about impact. Sample your own question-shaped queries and record what's actually showing up, on a fixed schedule, before drawing conclusions from memory or from a single bad week. Relevant Audience's recommended practice, taking monthly screenshots of your target question queries and logging them alongside impressions and position in Search Console, works as a starting discipline even if you outgrow the manual version of it quickly.

Track citation, not just rank. The headline number to watch going forward is not where a page sits in the traditional results; it is how often an AI answer names you when it responds to a question in your category. Rankings get you found, but citations get you chosen, and there is no page two in an AI answer, so a page that ranks fourth but never gets cited produces less business value than a page that ranks eighth and gets named in the response.

Diversify beyond any single SERP feature. PAA's near-total conversion to AI Overviews shows that individual SERP real estate can be redefined overnight by the platform that owns it. Treat every engine as its own surface. ChatGPT, Perplexity, Gemini, Copilot and Google's AI Overviews each draw from different sources and weight freshness and authority differently, so a strategy tuned only to one of them is a strategy tuned to disappear when that one changes again.

Finally, treat this as confirmation rather than alarm. The publishers losing the most here are the ones who had nothing beyond a well-optimized snippet. The ones absorbing this change without much disruption are the ones already producing content dense enough, current enough, and specific enough to be worth quoting, whether the quoting happens in a snippet, an AI Overview, or a chat answer synthesized somewhere else entirely.

Sources

Source: Search Engine Roundtable

97% of People Also Ask answers were AI Overviews in the first week of September 2026, up from 86% in August and about 12% in July 2025 Search Engine Roundtable, 2026-09-09

Source: Relevant Audience

Recommended tracking methods for question-shaped queries after the PAA to AI Overviews shift, including monthly baseline sampling Relevant Audience, 2026-09-11

Source: Auspia

Timeline and impact analysis of the People Also Ask to AI Overviews transition, including effects on publishers and SEO measurement Auspia, 2026-09-01