How Google AI Overviews Affect SEO in 2026

Google AI Overviews affect SEO in 2026 by taking clicks from the pages ranked beneath them and returning a smaller share to the pages they cite. Ahrefs found that an AI Overview now correlates with a 58% lower click-through rate for the top-ranking page. Seer Interactive found that a page cited inside the Overview earns 120% more organic clicks per impression than a page that is not, though still 38% fewer than on queries with no Overview at all.

Rankings still matter: they are how a page enters the pool Google's AI draws from. But a ranking no longer guarantees the click, and Google visibility does not carry over to other engines. In the Machine Relations Index, fewer than three in ten of the domains Google AI Overviews cited were also cited by ChatGPT.

This guide sets out what the 2026 evidence measures, which content is most exposed, and what to change.

The 2026 evidence at a glance

Study	Sample	Finding
Ahrefs, February 2026	300,000 keywords, December 2023 vs December 2025	AI Overview presence correlates with 58% lower CTR for position one
Seer Interactive, April 2026	53 brands, 5.47 million queries, January 2025 to February 2026	Cited pages earn 120% more organic clicks per impression than uncited pages on the same kind of results page
Pew Research Center, July 2025	900 U.S. adults, 68,879 Google searches, March 2025	Users clicked a traditional result on 8% of visits with an AI summary vs 15% without
Khosravi and Yoganarasimhan, University of Washington, September 2026	English Wikipedia articles matched to German and French editions	AI Overviews reduced Google referrals to English Wikipedia by 5.45% (vs German) and 4.82% (vs French)
Grossman et al., ACM SIGIR 2026	11,500 real-user queries	An AI Overview appeared on 51.5% of queries; Google Search, AI Overviews and Gemini retrieved substantially different sources
Machine Relations Index, November 9, 2026	Answer runs across six engines, May 10 to November 9, 2026	Fewer than three in ten of the domains AI Overviews cited were also cited by ChatGPT

How many clicks AI Overviews take, by position

Ahrefs compared 150,000 keywords that showed an AI Overview with 150,000 informational keywords that did not, using aggregated Search Console data. In December 2023, before AI Overviews, position one on the AI Overview keywords earned an average CTR of 0.073. By December 2025 it earned 0.016. After adjusting for the decline that hit all informational keywords, Ahrefs puts the AI Overview effect at about 58% for position one (Ahrefs, 2026).

Middle positions lose too. This is the answer for anyone asking whether ranking fifth or sixth still pays:

Position	CTR change when an AI Overview is present
1	-58.0%
2	-50.8%
3	-46.4%
4	-38.8%
5	-32.6%
6	-30.5%
7	-29.7%
8	-28.8%
9	-29.7%
10	-19.4%

Source: Ahrefs, February 2026, December 2025 data.

Behavioral data points the same way. Pew Research Center tracked the browsing of 900 U.S. adults in March 2025. On searches with an AI summary, users clicked a traditional result on 8% of visits, against 15% without one. They clicked a link inside the summary on 1% of visits, and ended their browsing session on 26% of those pages against 16% of pages without a summary (Pew Research Center, 2025).

The best causal estimate is smaller than the correlational ones, because it measures total search referrals rather than one position's CTR. Mehrzad Khosravi and Hema Yoganarasimhan of the University of Washington used Google's staggered rollout and compared English Wikipedia articles with the same articles in German and French. In the September 2026 version of their paper, AI Overviews reduced English search referrals by 5.45% relative to German and 4.82% relative to French, roughly 100 million fewer search visits to English Wikipedia a month (Khosravi and Yoganarasimhan, 2026).

The decline is also not one-way. Seer found organic CTR on AI Overview queries fell to 1.3% in December 2025, then recovered to 2.4% by February 2026, while queries without an Overview reached 3.8% (Seer Interactive, 2026). Seer treats the recovery as two months of directional data, not a forecast.

Why blogs take the largest hit

AI Overviews fall unevenly, and they fall hardest on the questions blogs are written to answer. In Seer's data, informational queries showed an AI Overview 36% of the time, commercial queries 8% and transactional queries 5%. Comparison queries triggered one 95.4% of the time (267 of 280) and question-format queries 85.9% (1,214 of 1,413) (Seer Interactive, 2026).

Pew found the same shape in everyday searching: 60% of searches beginning with who, what, when or why produced an AI summary, as did 53% of searches of ten words or more, against 8% of one- or two-word searches (Pew Research Center, 2025).

How often Overviews appear depends on the sample. Pew counted 18% of all searches in March 2025. The SIGIR 2026 benchmark found an Overview on 51.5% of 11,500 representative real-user queries (Grossman et al., 2026). A blog built on how-to, what-is and comparison posts sits at the top of that range.

The practical reading: a blog that earns traffic from informational questions should expect an Overview on most of them, and judge each post by whether it is cited in the Overview, not only by where it ranks.

Being cited is now the click that matters

Seer's most useful finding for content teams is the gap between cited and uncited pages on the same kind of results page. Being cited in the AI Overview delivered 120% more organic clicks per impression than not being cited, and still 38% fewer than a query with no Overview (Seer Interactive, 2026). A citation is an advantage over every other page on that results page, not a return to the old baseline.

Seer also warns against reading CTR alone. Its cited brands' CTR fell through 2025 because impressions more than doubled while clicks held flat: more queries earned citations. Track clicks and impressions separately before calling a drop a problem.

What earns the citation is not the same as what earns the ranking. The SIGIR 2026 study compared Google Search, its AI Overview and Gemini for each of 11,500 queries and found the sources each retrieved were substantially different, with an average Jaccard similarity below 0.2. Traditional Google Search was more likely to retrieve popular and institutional sites in government and education; the generative surfaces were more likely to retrieve Google-owned content. Overviews were also less consistent across two runs of the same query (Grossman et al., 2026).

Blocking Google's AI crawler costs visibility

Some publishers responded by blocking Google-Extended, the token Google uses for AI training and grounding. The SIGIR 2026 study found that sites blocking Google's AI crawler were significantly less likely to be retrieved by AI Overviews, even though Google still had access to their content (Grossman et al., 2026). The authors call for revenue frameworks between publishers and generative search providers. For a brand that wants to be cited rather than paid for its content, blocking removes the page from the answer without returning the click. Google documents its crawlers and tokens in its Search Central documentation.

Google visibility does not carry over to other engines

The Machine Relations Index records which domains six AI engines cite when people ask real buying and research questions across 24 subject categories and a legacy news bucket. In the November 9, 2026 release of the Machine Relations Index (methodology v2.1; 37,851 answer runs from May 10 to November 9, 2026), Google AI Overviews cited 3,498 distinct domains. Of those, 2,146 (61.3%) were also cited by Perplexity, 1,998 (57.1%) by Google AI Mode, 1,678 (48.0%) by Claude, 1,561 (44.6%) by Gemini and 995 (28.4%) by ChatGPT; 1,058 (30.2%) were cited only by Google's own AI surfaces (AI Overviews, AI Mode and Gemini). ChatGPT cited 10,709 domains across 8,956 observed runs, nearly seven times as many runs as the 1,337 observed for AI Overviews, and still overlapped with under three in ten of the Overview's sources.

In the October 18 release ChatGPT had 5,424 observed runs and the overlap was 26.7%. It has risen by less than two points while ChatGPT's observed runs grew by about two thirds, so the gap is not an artifact of too few ChatGPT answers. Google's two AI surfaces share a little over half of their sources.

Two collection changes sit inside this window. Google AI Mode's collection changed between August 10 and August 30, 2026, and answers collected since cite far fewer sources; Perplexity's API has returned 10 sources per answer since October 7. The AI Mode and Perplexity shares therefore pool two collection regimes: read them as whole-window overlaps, not as how either engine behaves today. The release records no collection change for ChatGPT, Claude or Gemini.

Of the 36,722 domains in the November 9, 2026 release of the Machine Relations Index, 512 (1.4%) were cited by all six engines, and those 512 account for 22.5% of all domain citations (51,108 of 227,162). Among the 100 most-cited domains by citation count, 70 are cited by all six.

The sources every engine cites form a short list. Among them are forbes.com, techradar.com, gartner.com, nih.gov, youtube.com and linkedin.com.

That is the strategic point for SEO teams. Ranking work earns a place in Google's pool. A brand that also wants to be named by ChatGPT, Perplexity and Claude needs to appear in the sources those engines cite as well, and the publications on the all-engine list are where one placement is read everywhere. Which publications AI engines cite breaks that list down.

Method note. Every figure in this section can be recomputed from the public release artifact at machinerelations.ai/data/machine-relations-index.json, the November 9, 2026 Machine Relations Index release (file SHA-256 6cd5e904f35fbaaa1a87fb22f381ede7bc5b6236e0c9e01bc5301269c11c6293): each domain's overall engine_breadth lists the engines that cited it, and runs_cited gives its citation count. Engine run counts are in the release manifest at machinerelations.ai/data/mri-release-manifest.json. The Index queries ChatGPT, Claude, Gemini and Perplexity through their APIs and records Google AI Overviews and AI Mode answers alongside them; it measures which sources were cited, not why (methodology).

Google is not losing search; publishers are losing clicks

Search demand has not fallen. Alphabet reported Google Search revenue of $60.4 billion in Q1 2026, up 19% year over year, and Sundar Pichai said queries reached an "all time high". In Forrester's February 2026 Consumer Pulse Survey, 71% of consumers had used Google in the past month to search for products they wanted to buy, against 26% who had used ChatGPT.

The same queries now send fewer visits to the pages that answer them. Google answers the question above the links, keeps the user and monetizes the page. The SEO question is no longer whether people search, but whether your page is the source the answer names.

AI-referred visitors are worth more per visit

The traffic that does arrive from AI is more valuable, at least in retail. Adobe's analysis of more than 1 trillion visits to U.S. retail sites found AI traffic up 393% in Q1 2026 from a year earlier. In March 2026, AI visitors converted 42% better than other visitors, a reversal from March 2025, when they converted 38% worse. They spent 48% longer on site, viewed 13% more pages, and generated 37% more revenue per visit (TechCrunch, 2026). Those figures describe U.S. retail; B2B buyers have no equivalent public dataset yet. Why AI search traffic is worth more covers what the gap means for a pipeline.

SEO tactics vs citation tactics: what the research tested

The Verge reported in April 2026 that AI search has put the SEO industry "through the wringer". The most rigorous test of what works is C-SEO Bench, accepted at NeurIPS 2025, which measured conversational SEO methods (rewriting documents to be more visible in AI answers) across question-answering and product-recommendation tasks. Most of those methods were largely ineffective and often lowered a document's position. Traditional SEO strategies, those that improve the source's ranking in what the model retrieves, were significantly more effective, and the gains shrank as more competitors adopted the same methods (C-SEO Bench, 2025).

That is the case for keeping SEO and dropping the page-rewriting tricks. Retrieval still starts from what ranks; the citation goes to the source the engine trusts.

Discipline	Optimizes for	Success condition	Scope
SEO	Ranking algorithms (Google, Bing)	Top 10 position on SERP	Technical + content
GEO	Generative AI engines (ChatGPT, Perplexity, Gemini)	Cited in AI-generated answers	Content formatting + distribution
AEO	Answer boxes / featured snippets	Selected as the direct answer	Structured content
Digital PR	Human journalists/editors	Media placement in trusted publications	Outreach + storytelling
Machine Relations	AI-mediated discovery systems	Resolved and cited across AI engines	Full system: authority → entity → citation → distribution → measurement

5 things to change in an SEO program in 2026

1. Map which of your queries carry an AI Overview

Search Console does not separate AI Overview impressions, so use a rank tracker's SERP-feature filter. Expect most informational, question and comparison queries to carry one. Those pages are judged by citation; your transactional and navigational pages are still judged mostly by rank.

2. Measure citation status alongside rank

For each high-value query, record whether your domain is cited in the Overview. Seer's 120% gap between cited and uncited pages is the size of what that status is worth. Track clicks and impressions separately, then extend the same check to ChatGPT, Perplexity and Claude, because Google's sources overlap only partly with theirs. How to measure AI search visibility sets out a method.

3. Keep ranking work, and stop rewriting pages for machines

C-SEO Bench found that improving a source's ranking beats rewriting its text to court an AI answer. Keep technical SEO, links and topical depth. Write each section so one sentence answers its question with a named source, because that is the passage an Overview can lift, and do not expect formatting alone to win the citation. Keep Google's AI crawler unblocked.

4. Earn coverage in the publications every engine reads

Only about one domain in seventy in the Machine Relations Index is cited by all six engines, and those domains carry roughly a fifth of all citations. A placement there is read by Google's AI surfaces and by ChatGPT, Perplexity and Claude, which a ranking on your own domain is not. That is the work Machine Relations describes and AuthorityTech practices: earning coverage in the publications engines already cite, then measuring the answers. AuthorityTech's results record where that coverage has run.

5. Put informational content where an Overview cannot replace it

An Overview substitutes for a page when a short synthesis answers the question. Original data, methodology, worked comparisons with sources and decision criteria are harder to compress, and a page that holds a number others quote gives an Overview a reason to cite it. Use the blog for that, and use earned coverage for the questions where engines prefer a source that is not the brand describing itself.

What this means for brands that depend on search traffic

The evidence tells a consistent story. AI Overviews take a large share of clicks from the results beneath them, most of all on the informational questions blogs answer. A cited page recovers part of that. Google's own revenue grows because it answers before the click. And the sources Google's AI cites are not the sources ChatGPT cites, so Google visibility is one engine's answer, not the market's.

Every discipline that touches online visibility is being restructured by the same shift: the reader changed. Search used to reward the page that ranked; synthesis rewards the source the answer trusts.

Earned media in respected publications has always been the strongest trust signal a brand can hold. It was true when your buyers were human, and it is true now that AI systems do the first cut of research on their behalf. Machine Relations is the discipline of making that credibility legible to AI engines: being resolved and cited across them through the same third-party authority that made earned media valuable in the first place.

If you want to see how your brand currently appears in AI answers across ChatGPT, Perplexity, Gemini and Google AI Overviews, start with a visibility audit. The gap between what you think AI says about your brand and what it actually says is usually the most useful data point.

FAQ

How do Google AI Overviews affect organic click-through rates?

Ahrefs found that the presence of an AI Overview correlates with a 58% lower click-through rate for the top-ranking page, and 30% to 51% lower for positions two to six, using December 2025 data on 300,000 keywords. Pew Research Center found users clicked a traditional result on 8% of visits with an AI summary against 15% without one. The University of Washington's causal study of Wikipedia found a smaller total effect, a 5.45% drop in search referrals.

Is SEO still worth it in 2026 if AI Overviews take the clicks?

Yes, with a different goal. Ranking is still how a page enters the pool Google's AI retrieves from, and C-SEO Bench found traditional SEO strategies more effective than rewriting content for AI answers. But informational posts should be judged by whether they are cited in the Overview: Seer Interactive found cited pages earn 120% more organic clicks per impression than uncited ones.

How is Google's AI Overview affecting SEO for blogs?

Blogs are the most exposed content type. Seer Interactive found AI Overviews on 36% of informational queries, 85.9% of question-format queries and 95.4% of comparison queries, against 5% of transactional queries. A blog built on how-to and comparison posts should expect an Overview on most of its traffic queries and measure citation status, not rank alone.

How often do AI Overviews appear in Google search?

It depends on the sample. Pew Research Center found an AI summary on 18% of all Google searches by 900 U.S. adults in March 2025. A 2026 SIGIR study found an AI Overview on 51.5% of 11,500 representative real-user queries. Longer searches and questions trigger them far more often than one- or two-word searches.

Does ranking in Google AI Overviews make a brand visible in ChatGPT?

Not reliably. In the Machine Relations Index release of November 9, 2026, fewer than three in ten of the domains Google AI Overviews cited were also cited by ChatGPT, although ChatGPT was observed on nearly seven times as many answer runs. Being cited across engines takes presence in the sources each one reads.

What is the difference between SEO and GEO?

SEO optimizes content for ranking algorithms; the goal is a top-10 position on the search results page. GEO (Generative Engine Optimization) optimizes for AI answer systems; the goal is to be cited in AI-generated responses. SEO is about position. GEO is about source selection. Both sit within the broader Machine Relations framework, which I coined in 2024 to describe the full system of earning AI citation through authority, entity clarity and distribution.

Should publishers block Google's AI crawler?

The evidence says blocking costs visibility. A SIGIR 2026 study found that sites blocking Google's AI crawler were significantly less likely to be retrieved by AI Overviews, even though Google still had access to their content. Blocking removes the page from the answer without recovering the click.

How do AI search visitors behave differently from traditional search visitors?

Adobe's analysis of more than 1 trillion U.S. retail site visits found that in March 2026 AI-referred visitors converted 42% better, spent 48% longer on site, viewed 13% more pages and generated 37% more revenue per visit than other traffic. The data covers U.S. retail; there is no comparable public dataset for B2B yet.

Updated November 9, 2026: the Machine Relations Index figures now come from the November 9 release, and the cross-engine comparison adds Claude and Gemini (methodology). The October 18 update had refreshed the traffic and citation figures from the latest version of each study.