How to Optimize for AI Citations in Google's People Also Ask

What changed in Google's People Also Ask results?

People Also Ask, or PAA, appears to have changed from a has that expanded a website snippet into one that frequently displays an AI-generated answer. Auspia reported on September 9, 2026 that two third-party tracking sources saw the transition accelerate through 2026, with one dataset reporting 97% AI Overview penetration in the first week of September.

That is a material change in what publishers are competing for. The old PAA model offered a visible source page when a user expanded a related question. The emerging model answers the question inside Google first. A supporting link may still create a route to the source, but the answer itself is no longer simply a passage lifted from one page.

The evidence needs careful handling. Google has not published an official percentage for AI-generated PAA answers or an announcement defining this specific transition. The before-and-after figures come from third-party observation, not a Google product report. Treat the figures as a strong signal about the search interface, then validate the queries that matter to your own business.

Google's current documentation does confirm the broader direction: AI Overviews surface relevant links, use Search eligibility as their baseline, and can vary from classic results. That means a PAA change does not create a separate technical eligibility system. It changes where the answer is assembled and how a page may earn visibility within it.

1. Before: PAA commonly expanded a website-sourced snippet and link.
2. After: third-party tracking reported AI Overview answers at 97% of sampled PAA results in early September 2026.
3. What to do: audit query-level PAA visibility and track supporting links inside AI answers, not only snippet ownership.

How large is the before-and-after shift?

The reported shift is from 12% AI-generated PAA answers in July 2025 to 97% in the first week of September 2026. That is an 85-percentage-point change over roughly fourteen months, based on the figures summarized by Auspia from Allintitle and AlsoAsked data.

The September figure comes with an important scope note. According to Auspia, Mark Williams-Cook analyzed 19.2 million English-language PAA queries made in 2026, and found 97% AI Overviews in the first week of September. Auspia also reported that Allintitle's Also Ask Miner data showed 100% AI-generated PAA answers beginning in August 2026. These are separate third-party datasets, not Google-confirmed global totals.

For editorial and growth teams, the operating conclusion is clearer than the exact decimal point. PAA should no longer be managed as a dependable extractive-snippet surface. A content program that celebrates a PAA placement without checking the current answer design may be reporting a visibility type that no longer sends the same kind of click or attribution.

Do not replace one weak metric with another. A screenshot of an AI answer is evidence of presence, not evidence of commercial impact. Record the query, locale, device, date, visible supporting links, destination URL, and downstream performance. That creates a repeatable observation set instead of a vague claim that a brand appeared in AI search.

1. July 2025: 12% AI-generated PAA answers, according to the third-party tracking summarized by Auspia.
2. August 2026: 86% AI-generated PAA answers, according to the same report.
3. Early September 2026: 97% AI Overview answers in the reported AlsoAsked sample.
4. Response: preserve dated query evidence because interfaces and citations can change.

Who does the People Also Ask change affect most?

The change affects publishers that treated PAA as a discoverability channel, especially teams targeting informational and long-tail questions. A page could previously earn an expanded answer even when it was not one of the highest-ranking blue links for the root query. If the interface now answers through an AI Overview, the publisher has to earn inclusion as a supporting source rather than depend on a directly extracted snippet.

B2B SaaS teams feel this when category education, implementation guidance, and comparison content lose an easy path from a question to a product-aware page. Local and service businesses face a related issue when question results about a service or area are answered before a searcher reaches a provider page. The exposure risk is not identical across queries, so it should be measured by intent and market rather than assumed from one result.

Smaller sites should not conclude that visibility is impossible. Google's documentation says AI has can surface relevant links and has opportunities for more types of sites to appear. But eligibility is not a guarantee of serving, and pages must first be indexed and eligible to show a normal Search snippet.

The practical dividing line is operational maturity. Teams that know their priority questions, source pages, and AI-answer appearances can respond quickly. Teams that only monitor rank positions may miss a change in how users receive the answer. Rankings still matter, but citations and supporting links increasingly show whether a brand is present at the moment a user gets an answer.

1. Publishers with question-led informational content need to re-baseline PAA traffic and appearance data.
2. B2B teams should test category, alternative, integration, and implementation questions.
3. Local businesses should sample service-and-location questions by market.
4. Every team should distinguish search impressions, supporting-link presence, visits, and conversions.

Should you still optimize pages around questions?

Yes, pages should still answer real questions clearly, but not because a rigid PAA format guarantees extraction. Google says there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode. The durable work is the same foundation: crawlable pages, indexability, useful content, and information that serves the searcher's task.

Question-led structure remains useful because it makes scope visible to readers and helps a page cover the subtopics that arise around a decision. It should not become a factory for shallow FAQ pages. An answer that merely restates the query gives an AI system little reason to use the page as a supporting source. A page should define the problem, make its evidence inspectable, and explain conditions that change the answer.

Build around a specific information need. For a software category, that may mean explaining selection criteria, implementation limits, pricing mechanics, or migration tradeoffs. For a service business, it may mean explaining service boundaries, local availability, timing, and what a customer should prepare. Use direct answers first, then add detail that helps a person evaluate the answer.

Avoid a false binary between SEO and citation work. Google explicitly says its existing SEO best practices remain relevant for AI features. The shift in PAA is a reason to improve content quality and measurement, not a reason to add unproven AI markup or hide ordinary Search fundamentals behind a new label.

1. Keep question-shaped headings when they match the reader's real decision.
2. Put the answer in the first sentence, then show the evidence or conditions behind it.
3. Use original expertise, dated sources, and clear page ownership where relevant.
4. Do not create special AI files or schema solely to chase AI Overview inclusion.

How should you measure AI citations after the change?

Measure AI citation visibility at the query level, because aggregate Search Console data cannot tell you whether a specific PAA answer cited a specific page. Google's documentation says traffic from AI has is included in the overall Web search type in Search Console. That makes Search Console essential for trend analysis, but insufficient on its own for identifying AI Overview supporting-link presence.

Create a monitored question set. Include the commercial category questions that shape buyer consideration, adjacent educational questions, comparison questions, and local intent questions where applicable. Capture each result on a defined schedule with country, language, device context, date, whether PAA appeared, whether the answer was AI-generated, which domains were visibly linked, and which URL was used.

Then join those observations to performance data. Search Console can show changes in clicks and impressions for pages and queries. Analytics can show engagement and conversions after a visit. Neither should be treated as proof that one AI Overview caused one conversion, but together they provide a better operating picture than rank tracking alone.

This is where Citation Share becomes useful as a management metric. Define the relevant answer set before measuring. Then calculate the percentage of those answers that visibly cite your domain. Keep it separate from Citation Count per day, Answer Presence, and Share of Voice. Each metric answers a different question, and combining them carelessly makes change look larger or smaller than it is.

1. Define a stable question universe before calculating Citation Share.
2. Store query, market, date, answer type, cited domains, and cited destination URLs.
3. Use Search Console for overall Web trends and analytics for post-click outcomes.
4. Review changes by topic cluster, not only as a single sitewide total.

What content changes are justified now?

The justified changes are practical and evidence-led: make important pages indexable, keep facts current, answer the main question directly, and cover the decision context that a thin snippet cannot. Google states that a page must be indexed and eligible to show a Search snippet to be eligible as a supporting link in AI Overviews or AI Mode. Start with that prerequisite before investing in new editorial formats.

Refresh pages that currently own or historically owned PAA visibility, but do not erase their useful structure. Check whether the answer is still accurate, whether dates and sources are visible, whether the page explains exceptions, and whether internal navigation helps readers reach the deeper material. Where a claim depends on a number, cite the primary source and include its date.

Build coverage around question clusters rather than publishing isolated answers. An isolated page can answer one query. A connected set of pages can demonstrate depth across the definitions, methods, comparisons, constraints, and use cases that surround the topic. That is a stronger editorial product for readers and a more resilient asset when Google changes an interface.

Do not claim that any content pattern forces a citation. Google's guidance is direct: technical compliance and good practices do not guarantee crawling, indexing, or serving. The goal is not to game the model. It is to give Google and readers a reliable page worth surfacing when the question requires supporting evidence.

1. Audit historic PAA pages for indexability, stale claims, weak sourcing, and thin explanations.
2. Prioritize pages that support high-intent category and comparison questions.
3. Link related pages so readers can move from a short answer to verification and depth.
4. Keep a change log so future citation movement can be compared with actual editorial updates.

What should publishers avoid doing in response?

Publishers should avoid treating the reported PAA shift as permission to abandon Search fundamentals. Google says no special optimization or additional technical requirement is needed for AI Overviews. A rushed program built around supposed secret signals is more likely to create clutter than durable visibility.

Do not publish unsupported claims about traffic losses from PAA. The current report identifies an interface shift, but Google has not supplied aggregate traffic-impact data for it. A business may lose visits on one question, gain supporting-link visibility on another, or see no measurable change. Make the case with your own baseline and post-change data.

Avoid confusing inclusion with endorsement. A source link can be useful evidence of Answer Presence, but it does not prove that Google recommends the business, that the answer is stable, or that a conversion followed. Those are distinct observations requiring different evidence.

Finally, do not overreact by blocking content without understanding the consequence. Google documents controls such as nosnippet, data-nosnippet, max-snippet, and noindex for limiting how content appears in Search AI features. Those controls can affect ordinary Search visibility too. Any restriction should begin with a documented policy decision, not panic over a changing result layout.

1. Do not buy or implement unproven AI Overview markup.
2. Do not report a sitewide traffic effect without a dated baseline.
3. Do not equate one visible link with demand generation.
4. Do not change preview or indexing controls before documenting the expected tradeoff.

What should your next 30 days look like?

The next 30 days should establish a baseline and improve the pages that matter, rather than chase a universal PAA tactic. Begin with a short list of questions that map to revenue, product consideration, or local demand. Record what the result looks like today and retain the evidence. That gives the team a way to distinguish a real visibility change from normal result volatility.

In week two, audit the source pages most relevant to those questions. Fix technical eligibility issues, update dated facts, strengthen source transparency, and make the core answer easy to find. In week three, fill material gaps where the monitored questions expose missing coverage. In week four, compare the new answer observations and Search Console trend data with the baseline.

The change also calls for a reporting change. Replace a simple PAA-win count with a brief view of citation presence, cited destinations, competitor share, and post-click outcomes where available. This moves the conversation from a disappearing search has to the broader question that matters: when Google answers the category question, is your evidence among the sources it shows?

That is the useful response to a fast-moving interface. Keep publishing work grounded in reader needs and verifiable sources. Keep technical eligibility clean. Watch the actual answer surfaces. The brands that can see their citation position will make better decisions than the brands waiting for an old SERP has to return.

1. Days 1 to 7: define priority questions and capture a dated baseline.
2. Days 8 to 14: audit technical eligibility and source quality on relevant pages.
3. Days 15 to 21: close substantive coverage gaps by topic cluster.
4. Days 22 to 30: compare answer presence, supporting links, search trends, and conversions.

Sources

Source: AuspiaAI

Third-party tracking reported PAA answers moving from 12% AI-generated in July 2025 to 97% in early September 2026, while Google had not publicly confirmed the transition. AuspiaAI, 2026-09-09

Source: Google Search Central

Google says standard SEO best practices remain relevant for AI has and that no additional requirements or special optimizations are needed for AI Overviews or AI Mode. Google Search Central, 2025-12-10

Source: Google Search Central

Google says a page must be indexed and eligible to show a Search snippet to be eligible as a supporting link in AI Overviews or AI Mode, and inclusion is not guaranteed. Google Search Central, 2025-12-10

Source: Google Search Central

Google says AI-has traffic is included in Search Console's overall Web search type and recommends using Search Console and analytics to evaluate site performance. Google Search Central, 2025-12-10