AI search statistics: what the evidence actually measures
AI search statistics measure different things: clicks after a search, crawler requests and visits referred to a website. Use a figure only when the population, the period and the denominator behind it match the decision in front of you.
A dated click study, with its limits
Pew compared Google visits with and without AI summaries. These are observations from one study and do not forecast your SaaS funnel.
Click observations from the same study
| Measure | Reported result | Source and period |
|---|---|---|
| Traditional result link clicked when an AI summary appeared | 8% of visits | Pew, 22 July 2025; March 2025 data |
| Traditional result link clicked without an AI summary | 15% of visits | Pew, 22 July 2025; March 2025 data |
| A source link inside the AI summary clicked | 1% of visits with a summary | Pew, 22 July 2025; March 2025 data |
Check what the comparison can establish
Pew studied the browsing of 900 US adults and reconstructed the search results afterwards. It reports an association. It did not test what happens when the summary is removed from the same search. Source: Pew Research Center, 22 July 2025; browsing during March 2025.
So do not write a blanket traffic cut into your revenue plan on the strength of it. Your queries, your audience and your landing pages need their own evidence. Use the study to sharpen your questions about click behaviour, then look at what your own data shows.
Do not call every AI measure market share
A share of recorded visits is not a share of searches, users or buying decisions, and a crawler request is something else again. Before you compare two sources, write down the denominator each one uses and whether they cover the same products and regions.
A tool may show your brand's visibility across a chosen set of questions. That number holds inside the sample. It is not your share of every AI-assisted buying conversation. Keep the question set and the collection method attached to the chart.
Read crawler statistics as infrastructure evidence
Cloudflare Radar defines AI bot and crawler traffic through associated user-agent activity; documentation checked on 16 November 2026. That counts automated activity. It says nothing about how many people are evaluating a product.
On your own site, ask the website owner to separate verified crawler requests from ordinary visits, and to label bots they cannot identify. More requests may justify an access review. They do not mean more qualified demand.
Build a source note before repeating a figure
Keep the original link, the publication date and the observation window next to every figure, along with the population and what was counted. If a later article repeats the number without the method, go back to the original.
If the data does not cover your market, say so. A narrow statement with a clear boundary is something your team can use. A global claim stitched together from unrelated studies can send a content programme the wrong way.
Connect research to a decision you can inspect
Use outside research to pick a question for your own review: which pages are losing search visits, which answers cite you, or which enquiries mention an assistant. Keep those findings apart until you have evidence linking them.
The AI visibility audit gives you a way to record observations. The B2B SEO guide keeps the work tied to buyer decisions, and the SEO, GEO and AEO comparison helps you name the activity precisely.
People also ask
Do AI search statistics prove that our traffic will fall? No. A study describes its own sample and period. Look at your own queries and recorded visits before you change a forecast.
Is crawler activity a measure of human demand? No. Automated requests and buyer visits are different events. Report them separately and say how you identified each.
What should accompany an external number? The source, the publication date and the observation period. Say what was counted and the population or denominator, so the reader can judge the claim.
Hiring a marketing leader to own this work? Brief Scalerr on the role or use the enquiry form below.
Hiring someone to own search?
Scalerr recruits marketing leaders for tech companies, from a first Head of Marketing to a CMO. Brief us on a marketing leadership hire, or use the form below.
Request a call