AI Search Is Real. The 740% Stat Was Not.

ChatGPT reached around 800 million weekly users by late 2025, roughly double February's 400 million, and Google searches now return an AI summary about one in five times. The headline isn't a single growth percentage. It's that AI answers are absorbing the click your funnel used to depend on. This piece sorts the verified numbers from the hype, and explains what changes for marketers. Current as of June 2026.

A note on a number you may have seen, including in an earlier version of this post. The claim that ChatGPT search "grew 740%" to "2.1% of global search" traces to a single agency study of 1,200 cybersecurity and managed-service firms, self-published, never independently checked. We've cut it. Precise-looking figures from one unverified source are worse than honest ranges. Everything below cites a primary or reputable source you can open yourself.

How big is AI search, really?

Two things are true at once. AI tools are growing faster than almost any consumer product before them. And they're still a small slice of total search.

On the growth side, OpenAI's own figures tell the story. Sam Altman said ChatGPT hit 800 million weekly active users in October 2025, up from 700 million in July and 400 million in February. OpenAI's February 2026 report put the figure at roughly 900 million. That's a doubling inside a year, measured by the company that runs the product.

On the share side, AI is still small in raw traffic terms. Semrush's analysis of billions of web visits found AI platforms drive about 0.14% of total website traffic, against 48.5% from organic search. The same study clocked AI traffic growing 66% across 2025, from 462 million to 767 million monthly visits.

So the honest framing is this: tiny base, steep curve. If you're waiting for AI search to "matter," you're reading the share figure and ignoring the slope.

Metric	Verified figure	Source
ChatGPT weekly active users (late 2025)	~800 million	OpenAI / TechCrunch
ChatGPT weekly users a year earlier (Feb 2025)	400 million	TechCrunch
AI share of total website traffic (2025)	~0.14%	Semrush
AI traffic growth (2025)	66%	Semrush
Google searches returning an AI summary (Mar 2025)	~18%	Pew Research

What does this change for marketers?

The number that should worry a marketing leader isn't ChatGPT's user count. It's what AI answers do to clicks.

Pew Research tracked the browsing of 900 US adults across 68,879 Google searches in March 2025. When an AI Overview appeared, people clicked a normal search result 8% of the time. Without one, 15%. The click rate roughly halved. And clicks on the links inside the AI summary itself? 1% of visits.

Read that again. The buyer gets your answer, reads it, and never arrives. Your content did the work. Someone else's interface kept the visit.

This is the mechanism behind every "our rankings held but traffic fell" conversation happening in marketing teams right now. You can hold position one and still lose the click, because the AI summary sits above you and answers the question before the user scrolls. Our take: most dashboards are measuring the wrong thing. Sessions and bounce rate were built for a world where the answer lived on your page. That world is leaking.

Where do people actually search now?

Not one place. That's the real shift. Search used to mean Google. Now a buyer's research question might land in ChatGPT, in Perplexity, in Google's AI Overview, or in a normal blue-link result, often all four across one decision.

ChatGPT leads the assistant category but its dominance is thinning as rivals grow. Within AI chatbot traffic, ChatGPT's share fell from 86.7% in January 2025 to 64.5% a year later, as Google Gemini and Perplexity took ground. For a marketer, that fragmentation matters more than any single platform's size. You can't optimise for "the AI." You optimise to be the source good enough that several of them cite you.

One bright spot in the data. Visitors who do arrive from AI platforms tend to be further along. Semrush found AI-referred visitors spend markedly more time on site than organic visitors, and other analyses report sharply higher conversion. Fewer clicks, but warmer ones. The click that survives the AI filter is a buyer who already read your answer and chose to come anyway.

How should marketing teams respond?

Stop optimising only for the ranking. Start optimising to be the cited source.

That's the practical pivot from SEO to AEO, answer engine optimisation. We've written the full breakdown in our GEO versus SEO guide, but the short version for a CMO:

• Write content an AI can quote cleanly. Direct answers near the top, clear headings, defined terms. If a model can lift a sentence and attribute it to you, you've earned a citation.
• Build entity authority, not just backlinks. AI systems trust sources they can identify and corroborate. Who you are, provably, now feeds whether you get named.
• Track citations, not just rankings. Are ChatGPT and Perplexity naming you for your category questions? That's the new position one.
• Keep content current. Models favour fresh, maintained sources over stale ones. A dated page is a page that quietly stops getting cited.

There's a governance edge here too, and it's the part most agencies skip. When an AI system quotes your content to a customer, it's also interpreting it. For regulated sectors, finance, healthcare, anything where a wrong answer carries liability, that interpretation is a risk you now own without controlling. Getting cited isn't only a visibility win. It's a place where marketing meets responsible-AI practice. That's the work we do: engineering AI visibility without the dark patterns the rest of the industry is already getting penalised for. Our guide to AI SEO services goes deeper on the how.

Frequently asked questions

Is the "740% ChatGPT search growth" stat real?

No. It comes from a single self-published agency study of 1,200 cybersecurity and managed-service firms, with no independent verification. We removed it. The verified picture: ChatGPT roughly doubled its weekly users to around 800 million across 2025, per OpenAI, while AI's share of total web traffic stayed near 0.14%, per Semrush.

How much of search is AI now?

It depends what you count. By raw website traffic, AI platforms sit around 0.14% (Semrush). But that undercounts impact, because AI Overviews now appear in roughly 18% of Google searches (Pew Research) and answer the query without a referral. The influence is far bigger than the referral number suggests.

Why is my organic traffic falling even though rankings held?

Most likely the AI Overview. Pew found clicks on traditional results roughly halve when an AI summary appears, from 15% to 8% (Pew Research). You can hold your ranking and still lose the click, because the answer now sits above you.

Should I optimise for ChatGPT or Google?

Both, and Perplexity, and Gemini. ChatGPT's lead is narrowing as rivals grow (market data). The work that gets you cited by one tends to help across all of them: clear answers, provable authority, fresh content. Optimise to be the source, not the platform.

The bottom line

The AI search shift is real, but it's been sold with the wrong number. The story isn't a tidy growth percentage. It's that AI answers are quietly taking the click your content earned, and they're doing it across several platforms at once.

Here's the call I'd make for any marketing team this year. Stop treating AEO as a future project. The base is small and the curve is steep, which is exactly when a head start compounds. Build content an AI wants to cite, prove who you are, track whether you're being named, and do it without the manipulative tactics that are starting to draw penalties. The brands that own their answer now will still own it when the share figure stops being small.

One last thing, and it's the part that's easy to miss. Being cited means being interpreted. For anyone in a regulated market, that's not just a marketing question. It's a responsible-AI one. Which is the whole point of doing this properly.

If you want support with this, VerityAI offers answer engine optimisation.