Every article about the death of SEO cites the same number: AI Overviews reduce organic click-through rates by 35%. Some cite 47%. The variation depends on methodology, but the direction is consistent. When Google puts an AI-generated answer at the top of the results page, fewer people click through to the websites below it.

That finding is real. What is missing from almost every analysis is the denominator. AI Overviews do not appear on every search. They appear on roughly 30% of queries, and the actual trigger rate varies wildly by category. On the other 70%, the SERP looks exactly like it did in 2023.

The difference between “AI Overviews cut clicks by 35%” and “AI Overviews cut clicks by 35% on 30% of queries” is the difference between a crisis and a headwind. Most of the industry is responding to the crisis version.

Where AI Overviews Actually Show Up

Google’s own data and third-party tracking from Semrush, Authoritas, and Advanced Web Ranking converge on a range: AI Overviews trigger on 13% to 30% of all queries, depending on market, device, and time period. The lower end reflects conservative counting (only full AI Overview panels). The upper end includes partial AI-generated snippets and expandable summaries.

The trigger rate is not random. It follows intent type with predictable logic.

Informational queries trigger the most. “How does photosynthesis work.” “What causes inflation.” “Symptoms of a torn ACL.” Google’s models are most confident generating summaries for factual, well-sourced topics with clear consensus answers. About 80% of AI Overview appearances are on informational queries.

Transactional queries almost never trigger. “Buy running shoes.” “Book a hotel in Barcelona.” “Best price iPhone 16.” Google knows these users want to browse products, compare prices, and make purchases. An AI summary cannot complete a transaction. Showing one would frustrate the user and reduce ad revenue. Google has no incentive to cannibalize its own shopping ads.

Navigational queries do not trigger. “Gmail login.” “Amazon.” “Netflix.” The user wants a specific destination. An AI Overview would be a speed bump, not a service.

Complex informational queries trigger inconsistently. “Should I use React or Vue in 2026.” “What are the trade-offs of microservices.” “How to structure a Series A pitch deck.” These require nuance, opinion, and context that AI Overviews handle poorly. Google is cautious about generating answers that could be wrong in ways that matter. The trigger rate on these queries is low and declining as Google learns where its summaries create more complaints than satisfaction.

The Featured Snippet Problem Nobody Mentions

Here is the number that changes the entire narrative: in 80% of cases where AI Overviews now appear, Google was previously showing a Featured Snippet.

Featured Snippets have been eating organic clicks since 2014. They pull a paragraph from a website, display it at the top of the SERP (“Position Zero”), and give the user an answer without clicking. Studies from Ahrefs, Moz, and SEMrush have consistently shown that Featured Snippets reduce click-through to the source page by 15-25% compared to a standard #1 organic result.

AI Overviews replaced Featured Snippets in most of their query territory. The incremental damage, the new clicks lost beyond what Featured Snippets were already capturing, is the difference between a 35% CTR reduction and a 15-25% CTR reduction that was already baked in. That net new impact is somewhere between 10% and 20% on the 30% of queries where AI Overviews appear.

Run the math across all queries: a 10-20% CTR reduction on 30% of queries produces a 3-6% reduction in total organic click volume. That range is consistent with the 2.5% overall decline measured by the Graphite/Similarweb 40,000-site study. The numbers align. The story is coherent. It is just a much smaller story than “AI Overviews are destroying SEO.”

The Category Split

Not every website feels the 2.5% average equally. That average contains a distribution with clear winners and losers.

Categories where AI Overviews hit hardest (10%+ organic traffic decline): news sites, health information, cooking/recipes, and entertainment schedules. The common thread: content that can be fully consumed as a paragraph. A recipe is a list of ingredients and steps. A symptom checker is a bulleted list. An entertainment schedule is a table. AI Overviews excel at extracting and presenting this content, and users have little reason to click through for more.

Categories that grew organic traffic: shopping, clothing, marketplaces, and financial services. The common thread: the user needs to do something the SERP cannot do for them. Try on clothes virtually. Compare 47 product listings. Read the fine print on a mortgage. Fill out an application. These actions require the destination site. AI Overviews cannot substitute for them.

The implication for content strategy is precise. If your content can be fully consumed as a snippet, AI Overviews are a direct threat. If your content is the beginning of a journey that requires engagement (buying, applying, exploring, deciding between complex options), AI Overviews are irrelevant to your traffic.

Google Is Also Learning What Not to Show

AI Overviews launched with some spectacular failures. Generated answers that recommended eating rocks. Medical advice that was dangerously wrong. Legal interpretations that were fabricated. Google responded by narrowing the trigger criteria, especially for YMYL (Your Money or Your Life) queries where bad answers carry real consequences.

The trigger rate has actually declined in some categories since the initial rollout. Health queries, which saw aggressive AI Overview placement in mid-2025, now trigger less frequently after user feedback and regulatory attention. Legal queries trigger rarely. Financial queries are restricted.

Google’s December 2025 core update explicitly emphasized “AI content quality and people-first signals.” The update pushed some AI Overview appearances down or removed them in categories where user engagement data showed dissatisfaction. Google is optimizing AI Overviews for user satisfaction, not maximum coverage. When the feature hurts engagement (users bouncing back to refine their query), Google pulls it back.

This means the 30% trigger rate is not a floor that will inevitably rise to 100%. It is a ceiling that Google is actively managing based on where the feature actually works. Some query categories will see higher trigger rates over time. Others will see lower rates. The blanket assumption that AI Overviews will eventually cover all queries misreads Google’s incentive structure. Google makes money when users are satisfied and keep searching. AI Overviews that frustrate users reduce session length and ad revenue.

What Publishers Should Actually Worry About

The real concern is not AI Overviews as a technology. It is the content model they make obsolete.

For a decade, a viable SEO strategy was: find a high-volume informational keyword, write a 1,000-word article that answers it, optimize the title and headers, build some links, rank on page one, collect traffic. Millions of pages were built this way. Many of them were useful. Many were padding around a paragraph of actual information.

AI Overviews deliver that paragraph directly. The 1,000-word wrapper is no longer needed. For publishers who built their traffic on that model, the decline is real and structural. It is not coming back.

But the 2.36 trillion clicks still flowing from Google to external websites every year are going somewhere. They are going to content that cannot be summarized in a paragraph. Original reporting. First-hand product testing. Expert analysis with a point of view. Interactive tools. Deep technical documentation. Content where the destination is the value, not a wrapper around an extractable answer.

AI Overviews did not kill SEO. They killed the kind of SEO that was already on borrowed time.