AI search is now a real discovery channel: a growing share of people ask ChatGPT, Perplexity, Gemini and Google AI Overviews instead of clicking blue links, and many answers cite only a few sources. The takeaway for businesses is simple. If AI engines cannot understand and quote your pages, you are invisible to those buyers, so structure your content to be cited.
The short version: The AI search statistics for 2026 tell one clear story. A large and growing share of buying decisions now starts with an AI answer instead of a list of links, the click is drying up, and the businesses that get named are mostly the ones earning mentions on trusted sources, not the ones with the prettiest websites. Below are the numbers that made me take this seriously, with what each one means for you. Want to see where you stand against them? Check free with my AI visibility tool.
Why I bothered pulling these together
I am wary of stat-stuffed posts, so I will keep this to the numbers that change what you should do. These are reported industry figures from 2026, and I have noted roughly where each comes from. Read them as direction and scale, not gospel to the decimal point. The direction is what matters, and the direction is loud.
The numbers that matter, and what they mean
AI Overviews now sit on roughly 1 in 5 Google searches
Industry analyses in 2026 put Google AI Overviews on around 18% of all searches, and over half of long-tail queries. When AI started, these mostly showed on informational searches. Through late 2025 they spread into commercial queries too, climbing from around 8% to 18%. What it means for you: the AI answer is increasingly sitting on top of exactly the searches where people are choosing who to buy from.
When an AI Overview shows, the click collapses
This is the one that should get your attention. Reported data in 2026 shows organic click-through rates falling from around 1.76% to about 0.61% when an AI Overview is present, a drop of roughly 60%. What it means: ranking under the AI answer is worth far less than being named inside it. Position one below the fold is still below the fold.
Zero-click searches keep climbing
Industry figures put zero-click searches at around 34% on Google without an AI Overview, around 43% with one, and as high as 93% inside Google's AI Mode. What it means: more and more, the answer happens on the results page, and the visit to your site never happens. Your name in the answer is the win, with or without the click.
ChatGPT is huge, and it is not the only one
Reported usage in early 2026 has ChatGPT at hundreds of millions of monthly users and billions of queries a day, holding roughly two thirds of the AI search market, with Google Gemini the fastest-growing challenger and Perplexity and Claude in the mix. What it means: your buyers are spread across several engines, and you can be named by one and missing from another for the same question. Checking one is not checking your visibility. On Lilach Bullock's own site, Bing's AI Performance report shows AI citations rose about four-fold between late July and mid September 2026, and visits referred by ChatGPT grew more than 13 times in a month (August to September 2026).
Perplexity really does send clicks
Reported figures put Perplexity's click-through on its cited sources at around 18 to 22%, well above the click rate on Google AI Overview citations, because Perplexity shows its sources prominently. What it means: a citation on Perplexity is not just credibility, it is traffic. Worth targeting on its own.
And the big one: most AI citations come from earned media
Here is the figure I keep coming back to. Reported data in 2026 suggests around 84% of AI citations come from earned media, not from a brand's own pages. What it means, and I cannot say this loudly enough: you do not get named mainly by polishing your own website. You get named by being mentioned on the sources the engines trust. That is the whole game, and the numbers say it plainly.
What these AI search statistics add up to
Lilach Bullock's reading of these statistics comes down to three things.
- The AI answer is eating the click, on the exact searches that matter for buying.
- Being named inside the answer beats ranking under it, by a wide margin.
- The way you get named is mostly earned media, the trusted mentions, not your own pages.
I did not need convincing, because this is the work I have done for twenty-one years, getting people into the press. But if you did, the numbers are now doing the convincing for me.
How this fits the bigger picture
These statistics are the case for generative engine optimisation. Start with what generative engine optimisation is, and since the data points so hard at earned media, read how to get featured on the sources AI trusts. If the numbers worry you, the first move is to see where you stand, and if you'd rather hand the work over, that's what my AEO services are for.
The numbers behind this are in my AI Overviews study: 300 live searches, 93% with an AI answer, and who Google cites.
FAQ
What percentage of Google searches have AI Overviews? Industry analyses in 2026 report Google AI Overviews on around 18% of all searches and over half of long-tail queries, up from a much smaller share when they launched.
Do AI Overviews reduce clicks? Yes. Reported 2026 data shows organic click-through rates dropping by roughly 60% when an AI Overview is present, which is why being named in the answer matters more than ranking under it.
Which AI search engine is biggest? Reported figures in 2026 put ChatGPT at roughly two thirds of the AI search market, with Google Gemini the fastest-growing challenger and Perplexity and Claude also in use. Your buyers are spread across all of them.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Where do AI citations come from? Reported 2026 data suggests around 84% of AI citations come from earned media rather than a brand's own pages, which is why getting featured on trusted sources is the core of the work.
Start here
See where you stand against these numbers, free. If the statistics worry you and you would rather have it handled, that is what I do.
Keep reading on GEO and AEO
- What is Generative Engine Optimization (GEO)?
- How to get featured on the sources AI trusts
- How to optimise for Google AI Overviews
- Work with me on GEO and AEO
Related reading: How AI Search Is Changing Marketing and How to Get Cited by Claude.
Want the complete version? Read where I break down AI marketing.
More on getting found by AI:
- AI brand monitoring tools
- whether you can pay to be recommended by AI
- why AI recommends your competitor and not you
See also: my 2026 AI search demand report, built from 3.4 million searches.
What I Changed on My Own Site After Watching AI Crawlers for Six Months
I pulled the server logs for lilachbullock.com from January through June and cross-referenced them against known AI crawler user agents: GPTBot, ClaudeBot, PerplexityBot, Google-Extended and CCBot. GPTBot alone hit my site 3,400 times in that window, more than double the crawl frequency of Bingbot. That surprised me, because Bing is the one feeding ChatGPT's live search results, so I expected it to be the heavier visitor. It was not even close.
The pages these bots kept returning to were not my homepage or my most "optimised" landing pages. They were three old blog posts, all published before 2021, each answering one narrow question directly in the first two sentences, then backing it up with a number or a named source. One post about LinkedIn outreach limits got crawled 190 times by ClaudeBot in a single month. My glossy 2024 cornerstone content, the stuff I paid a designer to make beautiful, barely got touched. The lesson I took from this: AI crawlers reward plain, front-loaded answers over polish and length.
Based on that, I rewrote the opening two sentences of my twelve highest-traffic posts to state the direct answer before any framing or story. I also added a short "quick answer" line right under each H1, something I had resisted before because it felt clunky for human readers. Within eight weeks, three of those posts started showing up as cited sources in Perplexity answers when I searched my own target queries, something that had never happened before with the old intros. I have no controlled experiment here, no A/B test, just a before and after on my own site, so treat it as one data point rather than a rule. But if you want a low-effort test, check your own logs for GPTBot and PerplexityBot activity before you touch anything. If they are already crawling a page heavily and it is not converting into citations, the intro is usually the first place to look.
- Pull your last 90 days of logs and filter for GPTBot, ClaudeBot and PerplexityBot user agents.
- Note which pages get repeat visits from the same bot within a 30 day window; that repetition usually signals the content is being reconsidered, not just indexed once.
- Rewrite the first two sentences of those pages to answer the core question with no throat clearing, then check back in 60 days.
