Evidence
AI traffic converts better — and you are undercounting it
The published first-party numbers agree on the shape: small volume, disproportionate value. Then the measurement problem makes it look smaller than it is.
The published first-party numbers agree on the shape: small volume, disproportionate value. Then the measurement problem makes it look smaller than it is.
Across every company that has published its own figures, AI search sends a small fraction of traffic and a disproportionate share of conversions. Ahrefs reported AI search accounting for 0.5% of visitors but 12.1% of sign-ups. Adobe Analytics, across more than a trillion visits, measured AI-referred traffic converting 31% better than other channels. And because referrers are frequently stripped, the volume you can see is a floor, not a count.
| Source | Finding | Basis |
|---|---|---|
| Ahrefs | AI search = 0.5% of visitors, 12.1% of sign-ups — roughly 23× the conversion rate of organic | Own analytics, self-reported |
| Vercel | ChatGPT referrals rose from under 1% to 10% of new sign-ups over six months | Stated publicly by its CEO |
| Adobe Analytics | AI referrals convert 31% better, spend 45% longer on site, bounce 33% less | Over one trillion visits |
| Semrush | An AI search visitor is worth roughly 4.4× a traditional organic visitor | 500+ topics analysed |
| Broworks | After 90 days of GEO work: 10% of organic traffic from AI engines, 27% of it converting to sales-qualified leads | Self-reported agency result |
Four of those five are first-party figures published by companies with an interest in the result looking good, and that should be weighed. What makes the pattern credible is not any single number but that independent parties measuring different things — a SaaS product, a dev platform, an enterprise analytics suite over a trillion visits, a research team across 500 topics — landed on the same shape.
A visitor arriving from an AI answer has had part of their evaluation done for them. They did not scan ten blue links and pick one; they asked a question, received a recommendation, and clicked through already knowing roughly what you do and why you were named. That is closer to a referral from a trusted colleague than to a search click.
It also arrives later in the buying process. The comparison, the shortlisting and the objections were handled inside the conversation with the assistant — which is precisely why being absent from those conversations is expensive. You lose the deal during a discussion you never saw.
Every figure above is an undercount, for reasons that have nothing to do with the honesty of whoever published it.
This last one is the largest and the least measurable. It is also the reason citation share — how often you are named in the answers themselves — is a better leading indicator than referral traffic. Referral traffic measures the minority who clicked.
What this does not prove. The volumes are still small in absolute terms. Most of these figures are self-reported by parties who benefit from them. None of them establishes that AI traffic will convert better for you, in your category, at your price point. They establish that the channel is real, converts unusually well where measured, and is growing — not that it is a replacement for anything.
Every figure above is linked to the page that published it. Where a number is self-reported by the company that benefited from it, this post says so.
Where it has been measured, yes. Ahrefs reported AI search delivering 0.5% of visitors but 12.1% of sign-ups. Adobe Analytics, across more than a trillion visits, measured AI referrals converting 31% better with 45% longer sessions and 33% lower bounce. Most such figures are self-reported, and none guarantees the same result in a given category.
Because it is systematically undercounted. Referrers are frequently stripped, assistants running in apps and desktop clients often pass nothing usable, and the largest case leaves no trace at all: a buyer who sees you named in an answer and searches your brand later arrives as branded organic.
Both, but citation share is the better leading indicator. Referral traffic only counts the minority who clicked; citation share counts how often you are named in the answers, which is what you can influence and what precedes the traffic.
Small in absolute terms today. The published cases put AI search at roughly 0.5% to 10% of traffic or sign-ups depending on the company and the period, on a steep growth curve. The argument for acting is the conversion quality and the trajectory, not current volume.
