58% of consumers have already replaced traditional search with AI tools for product and service discovery. Capgemini's 2025 research, 12,000 respondents across 11 countries. Not 'planning to.' Past tense. The shift you've been hearing about for two years has finished happening for the majority of buyers.
Most marketing reporting hasn't caught up. Most teams are still tracking 'organic search' as one bucket while half their audience asks the question somewhere else. This piece is the data, what it actually means, and three specific things to do about it before next quarter.
Volume isn't the story. Behaviour is.
The volume numbers are easy to quote. ChatGPT has 800-900 million weekly active users and processes 2.5 billion prompts per day. It's the fourth most-visited website on the internet, behind only Google, YouTube, and Facebook. Gemini hit 400 million monthly users on the back of being baked into Android, Chrome, and Google Search itself. Perplexity has 45 million monthly users running 780 million queries.
But the volume number that should change how you plan is buried in OpenAI's own usage analysis: 49% of ChatGPT usage is asking questions. Not coding help, not creative writing, not summarising articles. Asking questions. That's the exact behaviour that used to happen on Google, and it's now happening somewhere your analytics doesn't see.
Then there's the buyer-intent signal: 64% of customers say they're ready to purchase products recommended by AI. AI isn't a top-of-funnel research tool sitting alongside the real buying journey. It's increasingly the buying journey.
If you sell to professionals, the Perplexity demographic is worth re-reading: 80% hold a degree, 30% are senior leaders, 65% are high-income. The audience answering 'where should we buy this' for an entire team has already moved.
The click is becoming optional
Standard Google Search has a 34% zero-click rate. With AI Overviews showing, that jumps to 43%. In Google AI Mode, 93% of queries end without anyone clicking through to a website at all.
Read that number once more. Nine out of ten searches on the surface Google is steering toward end with the user reading the synthesised answer and closing the tab.
If your strategy depends on someone clicking through to your page, that strategy is competing for an ever-smaller slice of attention. The mention inside the answer is the touchpoint now. There is no consolation prize for being on the second page of citations.
AI is sending real traffic. The split is messy.
Want to see this in action?
Check how AI models talk about your brand — free, instant, no signup required.
The other side of zero-click is that AI engines do send some traffic, and the volume is no longer rounding-error. Similarweb's January 2026 data shows over 1.1 billion monthly referral visits from AI platforms. Gemini referrals grew 388% year over year. Semrush found ChatGPT sent 4 billion visits to websites in the second half of 2025 alone, accounting for 82% of all AI-platform traffic.
63% of websites now receive at least some AI-search traffic according to Ahrefs. The same research projects AI-search visitors will overtake traditional search visitors by 2028.
What this looks like in your analytics depends on how you've configured it. Many setups bucket AI referral traffic into 'Direct' or 'Other,' which is why teams keep saying 'we don't see any AI traffic.' You probably do. It's just labelled wrong.
Three things to do this quarter
The data is settled. The behaviour shift is real, the click is optional, and traffic attribution lags. Here's what changes if you act on this rather than just bookmarking it.
1. Audit how you currently appear in the answers your buyers actually ask. Pick 20 buyer-stage queries (your category, your competitors, your problem framings). Run each one in ChatGPT, Gemini, Claude, and Perplexity. Note who gets named, who gets cited, what's missing. This is the smallest possible version of the work and it usually surprises people. Our free AI visibility checker does this in 30 seconds.
2. Stop treating AI as one channel. The same prompt run on ChatGPT and Perplexity returns brands from completely different domains 89% of the time. Optimising for one engine doesn't carry to the others. Strategy that doesn't differentiate by engine is strategy that's invisible on at least one of them.
3. Fix your traffic attribution. Add the major AI engines as referral sources in your analytics. Look for utm-less referrals from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com. Your finance team doesn't need to see the channel grow by 388%. They need it counted at all.
The honest bottom line
AI search isn't replacing Google. Google Search usage actually increased after ChatGPT launched, from 10.5 to 12.6 sessions per week per user. The pie is bigger, not just sliced differently.
What's replacing isn't search itself. It's the assumption that your brand only needs to win one surface to be discoverable. Brands that treat AI visibility as a real channel get recommended. Brands that treat it as a 2027 problem get to find out, in 2027, that they should have started in 2026.
For the strategic framework, see our pillar guide on generative engine optimization.
