CHVEKLY

A year ago, “AI search” basically meant ChatGPT. Our own 2025 breakdown of which platforms lead reflects that moment well, and it’s no longer true. The shift happened faster than most marketing teams have adjusted to.

We pulled together the latest 2026 tracking data from Similarweb, StatCounter, and a couple of dedicated AI-referral studies to see what’s actually changed. Short version: ChatGPT is still the biggest single source of AI-driven traffic, but its share has been sliding every quarter, and the runners-up look different depending on who’s measuring and what kind of website you run.

ChatGPT’s lead is real, but it’s shrinking

Every tracker agrees on the direction, even if they don’t agree on the exact number.

StatCounter’s April 2026 data puts ChatGPT at roughly 77% of global AI chatbot referral traffic, down from about 84% a year earlier. Similarweb’s January 2026 numbers show a similar pattern: ChatGPT at around 61%, down from close to 68% the year before. One B2B-focused study (Goodie, covering brand sites between May 2025 and April 2026) found an even steeper drop, from 89% down to about 63% in under a year.

The exact figure depends on whether you’re looking at global consumer traffic or B2B referral sessions, and whether “AI search” includes in-app usage or just referral clicks. But nobody’s data shows ChatGPT holding flat. It’s losing ground every quarter, and the question worth asking isn’t “is ChatGPT still #1” (it is), it’s “who’s picking up what ChatGPT is dropping.”

Claude’s growth doesn’t show up where you’d expect

Gemini and Copilot are the obvious beneficiaries on raw usage numbers, mostly because they’re bundled into products people already use (Google Search, Workspace, Microsoft 365). But the more interesting story is Claude.

On global usage share, Claude is still small, somewhere around 2-4%. On B2B referral traffic specifically, though, Claude’s share jumps to around 18-19%, according to Goodie’s brand-panel research. That’s roughly seven times its global weight. One analysis (Axis Intelligence) called this the largest overperformance ratio of any major AI platform, and it lines up with something a lot of marketers have noticed anecdotally: Claude gets used more for long-form research and professional analysis, and that kind of query is more likely to end in a click-through than a quick factual lookup.

If your audience is B2B, this matters more than the headline market-share number. Optimizing purely for ChatGPT because it’s the biggest player overall can mean missing where your specific audience actually shows up.

Google’s AI Overviews are the bigger disruption, not ChatGPT

Here’s the thing most “ChatGPT vs. Google” framing misses: Google’s own AI Overviews and AI Mode are doing more damage to organic click-through than any external AI platform.

AI Overviews now show up on somewhere between 25% and 60% of US searches, depending on the tracker and the query type (the wide range reflects real methodology differences, not sloppy research). When an overview appears, click-through rates on the organic results below it drop by roughly 40-60%. Some trackers put clicks-to-website as low as 8%, down from 15% without an AI summary present. When Google’s full AI Mode kicks in rather than just an inline overview, one estimate puts the no-click rate as high as 93%.

Google still handles around 80-90% of total search volume by most counts. That’s the paradox: Google isn’t losing the search war to ChatGPT. It’s quietly taking clicks away from its own results pages through its own AI features, and that’s arguably the bigger traffic story for anyone who relies on organic search.

The traffic that does arrive converts better

The flip side of all this zero-click behavior: the visitors AI platforms do send you are worth more than average organic traffic.

Multiple sources point the same direction here. One industry report put AI referral conversion at around 7%, second only to paid search and well ahead of standard organic. Another found AI-referred visitors bounce less and browse more pages per session than typical organic traffic. The logic makes sense: someone who got a direct answer from an AI tool and still clicked through was probably looking for something specific, not just browsing.

So fewer visits, but better ones. If you’re only tracking session count, AI referral traffic looks like a rounding error. If you’re tracking conversion or pipeline value per visit, it starts looking a lot more important than the raw numbers suggest.

What actually gets cited

If you want your content showing up inside AI answers (not just clicked on, but actually cited as a source), the data points to a few consistent patterns across ChatGPT, Perplexity, and Gemini. For the tactical side of this, our GEO strategies guide covers the how-to. Here’s the why, based on the data:

• Third-party coverage beats owned content. Multiple studies put the split at somewhere between 84% and 95% of citations coming from earned media and third-party publishers rather than a brand’s own site. A University of Toronto study found this held up across different AI platforms and different ways of phrasing the same query.
• Freshness matters. One analysis found that the large majority of ChatGPT citations came from pages updated within the last 10 months, and that visible “last updated” dates roughly doubled citation odds. Related: our post on AI-generated content do’s and don’ts covers how to keep pages both fresh and trustworthy
• Structured, specific content wins. Content with clear stats, direct quotes, and defined sections gets cited more than vague long-form writing. Section lengths around 120-180 words per heading were linked to a noticeably higher citation rate in one study.
• Depth still counts, up to a point. Longer articles (2,000+ words) tend to earn more citations per piece, but only when they’re broken into clearly answerable chunks rather than one continuous block of text.

The pattern across all of it: AI platforms cite sources the same way a careful researcher would, favoring independent confirmation over self-promotion, and favoring content that’s easy to extract a clean answer from.

What this means if you’re planning content or link strategy for 2026

A few practical takeaways, based on everything above:

1. Don’t optimize for ChatGPT alone. It’s still the largest single platform, but its lead is shrinking every quarter, and the platform mix that matters depends heavily on your audience (B2B skews more toward Claude than the global average suggests).
2. Third-party mentions matter more than they used to. Earned coverage and links from other sites are now doing double duty: they still help traditional SEO, and they’re increasingly what gets you cited inside AI answers. This is exactly the gap our link building services are built to close
3. Track citations alongside clicks. Raw AI referral traffic is small compared to organic search, but it’s growing fast and converts well. A strategy that only measures session volume will underrate how much AI platforms are already influencing buying decisions before a visit ever happens. Our SEO consulting work now builds citation tracking into standard reporting for exactly this reason.
4. Keep content current. Freshness signals appear to carry real weight in what gets cited. A once-a-year refresh cycle probably isn’t enough anymore for competitive topics.
   
   The AI search landscape looked completely different a year ago and will probably look different again by the time 2027 rolls around. The one constant across every report we pulled: the days of treating “AI search” as a single platform with a single set of rules are over.