AI search vs SEO: what is changing and what is not

AI Overviews and AI Mode put a written answer above, or instead of, the usual list of results. This post answers one question: what has measurably changed in search, and what hasn't? It sticks to what Google says and what studies measured, with the method and the limits of each. If you want our forecast instead, read what we think will happen over the next two to three years.

What Google says about AI Overviews and AI Mode

Google's documentation on AI features (updated December 10, 2025) is short and plain. It says there are no additional requirements to appear in AI Overviews or AI Mode, and no other special optimizations are necessary. To be shown as a supporting link, a page must be indexed and eligible to appear in Google Search with a snippet.

The same page explains how the answers are built. Both features may use what Google calls a "query fan-out" technique: the system issues several related searches across subtopics and data sources. Google says this lets it show a wider and more diverse set of links than a classic search. Clicks and impressions from these features are counted in the Web search type in Search Console, not in a separate bucket.

1. 1You askOften a longer, more specific question
2. 2Fan-outThe system runs several related searches
3. 3Ranking systemsCore Search systems retrieve pages
4. 4AnswerA written reply with supporting links

A newer Google guide, updated July 10, 2026, goes further. It says SEO best practices still apply because the AI features are rooted in the core ranking and quality systems. It lists things you don't need to do: create llms.txt or other AI-specific files, break your content into small pieces, or write in a special style for AI. It also says that seeking inauthentic "mentions" across the web isn't as helpful as it might seem, because other systems block spam. That's a warning about manufactured mentions. It doesn't say that real coverage on real sites is worthless. For what llms.txt does and doesn't do, see our guide to llms.txt.

On traffic, Google's head of Search, Liz Reid, wrote on August 6, 2025 that total organic click volume from Google Search to websites was "relatively stable" year over year, and that Google was sending slightly more "quality clicks" (clicks where the visitor doesn't quickly go back). The post gave no figure for "relatively stable". It's a statement about all websites added together, so it can be true while many individual sites lose clicks.

How often do AI answers appear?

Studies disagree here, mostly because they count different things.

The Pew Research Center tracked the browsing of 900 U.S. adults in March 2025. Of 68,879 unique Google searches, 12,593 produced an AI summary, or 18%. The share depended heavily on the search. Only 8% of one- or two-word searches got a summary, against 53% of searches with ten words or more and 60% of searches that began with a question word such as who, what, when or why.

Semrush counted keywords instead of people. In a study of more than 10 million keywords, it found AI Overviews on 6.49% of queries in January 2025, 24.61% in July 2025 and 15.69% in November 2025. Semrush itself calls that growth volatile: it rose, peaked in summer and then fell back.

Those two aren't in conflict. One counts what a sample of people searched, the other counts a keyword database, and the months differ. What they share is the pattern: longer, question-style searches get an AI answer far more often than short ones.

Share of Google searches that showed an AI summary

• One- or two-word searches8%
• Ten words or more53%
• Searches starting with a question word60%

That matters because the searches most exposed are the informational ones: "what is," "how do I," "why does." If most of your traffic comes from those, you're closer to the change than a business that mostly gets searches for its own name.

What happens to clicks when an AI answer shows up

Four sources, four different measurements.

Pew (July 22, 2025). On visits to a Google results page, people who saw an AI summary clicked a traditional result in 8% of visits. People who didn't see one clicked in 15% of visits. Clicks on a link inside the AI summary happened in 1% of visits to pages that had a summary. The session ended after the results page on 26% of pages with a summary, against 16% of pages with only traditional results. This is U.S. data from a small tracked sample, and Pew only looked at Google.

Ahrefs (published February 2026). Ahrefs looked at 300,000 keywords: 150,000 where an AI Overview showed and 150,000 informational keywords where none did. Using aggregated Search Console data on desktop click-through rate, it compared December 2023 with December 2025. It found that an AI Overview was associated with a position 1 click-through rate about 58% lower. An earlier Ahrefs study from April 2025 had found 34.5%. Ahrefs notes a limit: the keywords that showed an AI Overview in December 2025 hadn't shown one in December 2023, so the comparison is indirect.

Seer Interactive (September 2025 update). Seer used Search Console data from 42 client organizations: 3,119 search terms and 25.1 million organic impressions, from June 2024 to September 2025. In September 2025, organic click-through rate was 0.61% on queries with an AI Overview and 1.62% on queries without one. The same study found that click-through on queries without an AI Overview also fell, from 2.72% in June 2024 to 1.62% in September 2025. So something other than AI Overviews is also moving clicks.

SparkToro and Similarweb (2026). Using a Similarweb panel of browser data for January to April 2026 in the U.S., SparkToro found that 68.01% of Google searches ended without a click. A "click" there means any click: an organic link, an ad, or a Google property such as Maps or YouTube. That figure isn't about AI Overviews alone, and the study notes its browser-only data likely understates the true zero-click rate.

Why the click studies disagree

Put the four side by side and the differences are in the details, not in anyone's honesty.

Three things explain most of the gap. The unit differs: visits, click-through rate, or total clicks. The population differs: a panel of people, a set of keywords, or every site on Google. And the question differs: how one result performs, versus how many clicks exist in total. A result at position 1 can lose a large share of its clicks while the total stays level, if people run more searches overall. Google says people are searching more. We can't check that from the outside.

Our view: Treat the direction as established and the size as unknown. If you rely on informational searches, plan for fewer clicks. Don't plan around any one percentage, including the biggest ones.

How big is AI search compared with Google?

Smaller than the headlines suggest, and growing.

A SparkToro study using Datos panel data (published August 2025) found that 95% of Americans still use traditional search engines each month, and that over 20% are heavy users of AI tools, meaning ten or more uses a month. The panel misses mobile app use, so both numbers are rough.

For referrals, Ahrefs analyzed 81,947 websites in June 2025. AI assistants made up 0.25% of a site's traffic on average. Google sent about 210 times more traffic than the three biggest AI platforms combined. Ahrefs warns that some assistants don't pass referral data, so AI traffic is probably undercounted, and it describes AI traffic as marginal for most sites.

Revenue gives a third angle. Alphabet reported on July 22, 2026 that Google Search and other revenue for the second quarter of 2026 was $63.271 billion, up 17% on the year, and said its AI features are driving Search query growth. Revenue isn't clicks to your site, and it can't tell us whether publishers are better or worse off.

None of this says AI search is small for everyone. Traffic from AI chat is tiny on average. The influence of AI answers on what people believe and which brands they remember is a different thing, and these studies don't measure it.

What stays the same for SEO

Most of what Google lists is the basics. A page has to be crawlable, indexed and eligible to show with a snippet. The AI features retrieve pages through the same ranking and quality systems as normal search. Spam policies apply, including the policy on scaled content abuse for pages made with generative AI "without adding value for users." And Google keeps asking for the same thing it asked for before: unique, "non-commodity" content that goes beyond what a model could produce by remixing what's already online.

What stays and what shifts

Stays the same

• Pages must be indexed and snippet-eligible
• Core ranking and quality systems sit underneath
• Google still asks for original, helpful content
• Spam policies still apply

Shifting

• Fewer clicks on many answer-style searches
• Citations come from a wider set of pages
• Google reports AI impressions, not AI clicks

Links are still part of how pages are found. Google's link documentation (updated December 10, 2025) says it uses links as a signal when determining the relevancy of pages and to find new pages to crawl, and we cover what that means in the future of link building.

What changes for people who earn links and mentions

Here the evidence is thinner, and we want to be careful about what it says.

In December 2025, Ahrefs studied 75,000 domains with a Domain Rating above 40, using each domain's highest-volume keyword (at least 800 searches a month) and millions of AI responses. It measured Spearman correlations between different signals and a brand's AI visibility:

Ahrefs describes the correlation with the raw number of backlinks as very weak and gives no figure. It also states the limit itself: correlation isn't causation, and the patterns don't mean that improving a metric will boost AI visibility. Also note who was studied. These are established sites with a Domain Rating above 40, so the results may not hold for small or new brands. A separate Ahrefs test in June 2025 compared Ahrefs Rank, a link-based domain score, with the sites most mentioned in AI answers. The correlations were -0.12 for AI Overviews, 0.01 for ChatGPT and -0.34 for Perplexity. The author said the sample, the top 50 sites per system, was too small to rely on.

What we can fairly say: in these data, how often a brand is mentioned across the web moves with its AI visibility more strongly than link counts do, and nothing here shows cause. What we can't say: that mentions produce citations, or that links don't. Seer's data adds one more correlation: brands cited in an AI Overview had a higher organic click-through rate (0.70%) than brands that weren't cited (0.52%) in Q3 2025. Cited brands may simply be the stronger brands.

Our view: The practical lesson is to judge a placement by whether real readers see it and whether it suits your brand, not by an AI promise. We go deeper on this in brand mentions vs backlinks, and on how link text fits in anchor text in the AI era. If you're weighing paid placements, our guides to buying backlinks and rel attributes explain how Google treats them.

How to measure this without fooling yourself

Both Google and Bing now give site owners a first-party report on AI features, and both are limited.

Google's generative AI performance report in Search Console shows impressions: how often links to your pages were shown in AI Overviews and AI Mode. It doesn't show clicks or position. The usual 1,000-row limit applies, and recent data can be preliminary. Microsoft launched AI Performance in Bing Webmaster Tools as a public preview on February 10, 2026. It shows total citations, the average number of cited pages per day and "grounding queries," the phrases the AI used to retrieve your content. Microsoft says it doesn't show placement within an answer, ranking, authority or the role a page played.

For checking mentions yourself, a single screenshot proves little. SparkToro and Gumshoe ran 2,961 runs with 600 volunteers on 12 prompts across ChatGPT, Claude and Google AI Overviews, in a study published January 28, 2026. They found less than a 1 in 100 chance that the same list of brands appears in two answers to the same prompt. Their conclusion was that visibility as a percentage across many prompts, each run many times, is a reasonable measure, and that a single "ranking position in AI" isn't. Google adds that no third-party tool has access to its internal ranking or AI systems.

Our guides on measuring SEO when clicks fall and checking AI mentions yourself walk through a method. For how a link or mention shows up in other reporting, see measuring link building success.

What to do now, and what to ignore

Do:

• Check that your important pages are indexed, crawlable and snippet-eligible.
• Put your effort into pages that say something others don't: original data, firsthand experience, clear explanations for your market. Google calls this non-commodity content.
• Track impressions, clicks and branded searches together in Search Console, and read the new AI report for what it is: impressions.
• Earn links and mentions on real sites your readers use. Judge each by its audience and relevance first. If you want help finding those sites, we run manual outreach where you approve every site before anything is published.
• Re-run any citation checks several times before you draw a conclusion.

Ignore:

• Anyone promising AI citations. No source we found shows that anyone can guarantee them, and we don't.
• Special files and markup "for AI." Google says it doesn't use them.
• A single percentage for "how much AI cuts clicks." The studies differ for good reasons.
• Rank-style scores for "AI position."

Sources

• Google Search Central: AI features and your website (updated December 10, 2025)
• Google Search Central: Optimizing your website for generative AI features on Google Search (updated July 10, 2026)
• Google Search Central: Using generative AI content on your website (Google guidance on scaled content abuse)
• Google Search Central: Link best practices for Google (updated December 10, 2025)
• Google Search Console Help: Generative AI performance report (Search) (what the report shows and its limits)
• Google, Liz Reid: AI in Search is driving more queries and higher quality clicks (August 6, 2025)
• Pew Research Center: Google users are less likely to click on links when an AI summary appears (July 22, 2025; 900 U.S. adults, March 2025)
• Ahrefs: Update: AI Overviews reduce clicks by 58% (February 2026; 300,000 keywords)
• Seer Interactive: AIO impact on Google CTR, September 2025 update (3,119 search terms, 42 organizations)
• SparkToro: In 2026, less than one third of Google searches still send a click (Similarweb panel, January to April 2026)
• Semrush: AI Overviews study (10 million+ keywords, refreshed December 2025)
• SparkToro: 20% of Americans use AI tools 10X+/month (Datos panel data, August 2025)
• Ahrefs: AI traffic has increased 9.7x in the past year (June 2025; 81,947 websites)
• Alphabet: Q2 2026 earnings release (July 22, 2026)
• Ahrefs: Brand visibility and AI correlations (December 12, 2025; 75,000 domains)
• Ahrefs: Domain-level link metrics may not be good predictors of AI search mentions (June 25, 2025)
• Bing Webmaster Blog: Introducing AI Performance in Bing Webmaster Tools (February 10, 2026)
• SparkToro and Gumshoe: AIs are highly inconsistent when recommending brands (January 28, 2026)

Checked on 6 October 2026.