Report
State of AI Search 2026
The data behind why your category's buyers stopped finding you, and what the brands they do find are doing differently.
Get the reportFree · No sales call required
$750B
By 2028, an estimated $750 billion in US revenue will flow through AI-powered search rather than the ten blue links.
McKinsey, 2025
“Why doesn't ChatGPT recommend us when buyers search our category?”
That is the question this was written to answer, with data rather than opinion.
- Founders
- CMOs
- VPs of Growth
- Heads of SEO
- Product leaders
What’s inside
Where AI traffic actually comes from
Volume and intent across ChatGPT, Gemini, Claude and Perplexity, and why the four behave differently enough that one strategy cannot serve them.
How often brands get cited, and which ones
Citation frequency by category, and the gap between being mentioned and being cited, which are different problems with different fixes.
What the assistants select on
The page and corpus signals that correlate with being named, separated from the ones marketers assume matter and that do not.
The first-mover position, and how long it lasts
What early citation is worth, whether it compounds, and how quickly a competitor can take it back.
Google rank against AI visibility
The correlation is far weaker than most teams plan around. This chapter is the one that changes budgets.
How smaller brands beat bigger ones
Where category leaders are structurally weak in AI answers, and the positions a challenger can take without their budget.
What the data shows
Three of the findings, in full and sourced, so you can judge the rest before handing over an email.
39.8%
AI Overviews reduce outbound organic clicks by 39.8%, while clicks on sponsored results and Google's own properties stay flat.
61% vs 25%
61% of pages cited by ChatGPT carry structured data markup, against 25% of pages ranking in Google's top organic results.
36
Across 126 million prompts, only 36 brands hold consistent visibility across every major AI platform. Winning one tells you almost nothing about the others.
Get the State of AI Search 2026
How assistants pick the brands they name.
Method
- Real category questions, not keyword lists.
- Every engine queried on its own. No blended score.
- Citations counted separately from mentions.
- Model versions annotated, so a retirement never reads as a drop.
Rather talk it through?Book a demo (opens in a new tab)