Ahrefs' AI Search Report 2026: what you need to know about AI search

Over the course of 2025 and 2026, Ahrefs went from speculating about what AI does to search, to measuring it at scale. The result is a series of studies that together make up the company's AI Search Report for 2026 — a benchmark built on 146 million search results pages, 730,000 AI answers and analysis of 75,000 brands. This is not a hunch from a webinar. It is the largest publicly available dataset we have on how AI search actually behaves.

For businesses, the conclusion is uncomfortably clear: the rules for getting found online have changed, and the old SEO habits no longer hold on their own. Below we go through the most important findings — with the numbers, the methodology, and what you should actually do about them.

What the report consists of

The "AI Search Report" is not one single document, but a collection of around a dozen studies Ahrefs published through 2025 and into 2026, summarised in a Q1 2026 benchmark. The main building blocks are:

• A CTR study on 300,000 keywords measuring how AI Overviews affect clicks.
• A citation study on 863,000 SERPs and 4 million AI Overview URLs.
• A brand study on 75,000 brands looking at what correlates with AI visibility.
• A freshness study on nearly 17 million cited URLs.
• A traffic study on 3,000 websites measuring how much traffic AI actually sends.

Together they give a rare full-picture view. Let us take the findings one by one.

Finding 1: AI Overviews eat 58% of clicks

The most striking number is this: when Google shows an AI Overview, the click-through rate of the top-ranking page falls by 58%. Ahrefs compared 150,000 keywords with an AI Overview against 150,000 informational searches without, based on aggregated Google Search Console data for desktop.

The numbers are merciless. In December 2023, a keyword in first position had a click-through rate of 0.076 without an AI Overview and 0.073 with one. Two years later, in December 2025, the same figures had fallen to 0.039 and 0.016. In other words: AI Overview searches lost more than half of their clicks compared with what they would otherwise have earned.

Ahrefs sums it up themselves like this: "For every 100 clicks you could historically earn on a top-ranking page, Google now keeps 58." By comparison, their previous study from April 2025 found "only" a 34.5% reduction. So in eight months the effect nearly doubled. Other independent studies — from Seer Interactive, Authoritas and Kevin Indig — point in the same direction, with drops of between 47% and 65%.

Finding 2: AI Overviews appear more often — especially on long queries

How widespread is this really? Across 146 million search results pages, Google showed an AI Overview for 21% of all keywords. But the average hides an important pattern: the longer and more specific the search, the more often AI kicks in.

• 9.5% for single-word searches.
• 46.4% for searches of seven words or more.

This is crucial to understand: the long, conversational searches — precisely the kind of questions people ask when they are close to a decision — are where AI dominates most. They are also the kind of searches that drive leads and sales. In other words, the zero-click effect hits hardest exactly where it hurts most commercially.

Finding 3: Citations move out of the top 10

A common assumption is that if you just rank in Google's top 10, you get cited in the AI Overview. That is no longer true. In an analysis of 863,000 SERPs and 4 million citation URLs, Ahrefs found that only 38% of AI Overview citations come from the top 10. In July 2025 the figure was 76%. So in half a year the share was cut in two.

The explanation is what Google calls "query fan-out". When a search triggers AI, Google splits it into several related sub-searches, runs each separately, and cites the pages that appear most often across them. So you are no longer competing for one keyword, but for a whole network of related questions.

In practice this means you have to cover a topic broadly and deeply — building content clusters that answer the many sub-variants of a question — instead of optimising one page for one head keyword.

Finding 4: YouTube is the strongest signal for AI visibility

This is perhaps the report's most surprising finding. When Ahrefs analysed 75,000 brands to work out what correlates with visibility in AI Overviews, the answer was not links, not page count, not domain authority — but mentions on YouTube, in video titles, descriptions and transcripts.

Glen Allsopp, Head of Marketing Strategy & Research at Ahrefs, sums up the shift: "As brand discovery spreads across AI platforms that each behave differently, tracking and building visibility online has changed fundamentally."

The numbers back up YouTube's role: in the citation study YouTube accounted for 18.2% of citations that did not themselves rank in search, made up 5.6% of all AI Overview citations, and grew 34% as a citation source in half a year — enough to be the single most-cited source of all.

By comparison, link count and total page count showed only weak correlation. That does not mean links are worthless — but that brand mentions broadly across the web, and video in particular, now weigh more heavily for AI than the classic SEO signals. Earlier studies from Ahrefs show that brand mentions correlate around 0.66 with AI Overview visibility, and higher still when the mentions sit on strongly linked pages.

Finding 5: Freshness counts — AI cites newer content

Ahrefs analysed nearly 17 million cited URLs and found that AI-cited content is 25.7% fresher than ordinary organic results. The average age of an AI-cited URL was 1,064 days (2.9 years), against 1,432 days (3.9 years) for organic results.

ChatGPT showed the strongest preference for fresh content, while Google's AI Overviews sat almost in line with ordinary organic rankings. A telling example: a single update to a HubSpot article triggered 1,135 new AI Overview mentions. Even so, it is worth noting that the average age is still close to three years — established, authoritative content dominates, but keeping it updated pays off.

Finding 6: AI traffic is small — but converts unusually well

Here you have to hold two thoughts at once. On the one hand, AI traffic is still marginal in volume: Ahrefs found that AI makes up around 0.1% of all referral traffic, and that Google sends roughly 190 times more traffic than ChatGPT. Of AI traffic itself, 98% comes from three sources:

• ChatGPT — around 50%
• Perplexity — around 30.7%
• Gemini — around 17.6%

On the other hand, a study of 3,000 websites showed that a full 63% of sites already get at least one visitor from an AI source. And the quality is high: Ahrefs reports that around 3% of their own conversions over the past year came from AI, and that more than 14,000 new users named ChatGPT as their source themselves — despite AI making up a fraction of the visits.

A critical warning from the report: much AI traffic is invisible. Several AI sources (Copilot, Mistral and Jasper among them) send no referral data, so the visit lands as "direct" in Google Analytics. The real AI traffic is therefore higher than the tools show. To measure this properly, you have to supplement with qualitative questions such as "Where did you hear about us?".

Finding 7: Every AI assistant is its own channel

One last, strategically important finding: the AI assistants largely cite different sources. Among the 50 most-cited domains across Google AI Overviews, ChatGPT and Perplexity, only seven appeared on all three lists. In other words, 86% of the sources were unique to a single assistant.

That means "one strategy for everything" does not work. Getting cited in ChatGPT, in Perplexity and in Google's AI Overview takes partly different moves. On top of that, around 5.9% of websites still block OpenAI's GPTBot — which in practice makes them invisible to ChatGPT. So the first rule is banal, but overlooked: let the legitimate AI bots read your pages.

What does this mean for your business?

The report paints a clear picture: visibility in AI has become its own discipline, not a side effect of good SEO. Here are the practical consequences:

• Stop measuring success in clicks alone. When AI answers the question without the user clicking, being mentioned and recommended becomes the new goal. You have to track AI visibility, not just rankings.
• Build brand mentions broadly. AI picks up who is talked about online. Digital PR, guest content, industry coverage and — according to the data — YouTube in particular now weigh more than pure link building.
• Take video seriously. When YouTube is the strongest single signal, video is no longer "nice to have". Even simple explainer videos with good text and transcripts can lift AI visibility.
• Cover topics broadly, not just keywords. "Query fan-out" means you have to answer the whole network of related questions around a topic — build clusters, not single pages.
• Keep content fresh. Regular updates can produce measurable movement in AI citations, as the HubSpot example shows.
• Optimise per platform. ChatGPT, Perplexity and Google AI Overviews cite differently. Check where you are already mentioned, and prioritise accordingly.
• Let the bots in. Check that you are not blocking GPTBot and other legitimate AI crawlers in robots.txt.

Conclusion: from ranking to recommendation

Ahrefs' AI Search Report for 2026 documents a shift that has been building for a while, but is now measurable at scale. Clicks from traditional search are shrinking where AI takes over. At the same time, a new form of visibility is emerging — where being cited and recommended by an AI assistant is what counts.

Google is still by far the largest traffic source, so traditional SEO is far from dead. But businesses that want to be visible in 2026 and beyond have to work on two fronts at once: rank in search and become the one AI chooses to recommend. That takes strong content, a broad brand presence, video and a structure that lets the machines understand who you are. That is exactly the work CitationLab helps businesses with — going from ranking to recommendation.