Do AI search visitors convert better?

What does the 23 times claim actually rest on?

The figure circulating is that traffic arriving from AI search converts far better than traditional search traffic. The source is a single post published by Ahrefs on 16 June 2025. What it reports: on Ahrefs’ own website, over a trailing thirty days, AI search accounted for 0.5% of traffic and 12.1% of signups, which works out at roughly a 23 times higher conversion rate.

The author hedges it himself, writing that he is not sure it will scale that way, and notes that AI users click links about 75% less.

Why won’t we present the Ahrefs figure as a benchmark?

There are two reasons, and neither of them is that the figure is wrong. The first is what the dataset covers. It is one site. Not a sample of sites, not a panel, not an industry study. A single business-to-business software company measuring its own traffic.

The second is who published it. That company sells SEO software and has a commercial interest in AI search being important. That does not make the number false, and it does mean it should carry the label.

And the audience is wrong for our clients. A B2B SaaS signup from a researcher who found a tool through an assistant has almost nothing in common with a homeowner choosing who to call about a failed heat pump.

A separate figure of 4.4 times circulates widely and we could not trace it to a primary publication at all.

So the honest position is that the effect is plausible, the evidence is thin, and anyone quoting a multiple at you for your market is quoting something that has not been measured there.

How do the two circulating figures compare?

Two multiples get quoted in this conversation and they do not have the same standing behind them. One has a named publisher, a stated sample and a date. The other does not. Source, scope and what is actually reported for each are below.

What we read from that, and this part is our judgement rather than anything the sources state: only one of these two multiples has a publisher, a date and a stated sample standing behind it, and even that one was measured on a business-to-business software company rather than on a local service business.

Sources: Ahrefs, published 16 June 2025.

What is defensible underneath the claim?

A visitor who arrives having already been given an answer that named you has been pre-qualified in a way a visitor scanning ten blue links has not. That mechanism is reasonable and does not need a number attached to be useful.

It changes what the landing page has to do. Someone who arrived on a recommendation is checking rather than exploring, which means proof, specifics and a way to make contact matter more than persuasion.

It also changes what to measure. If assistants send small volumes of high-intent traffic, a report that only counts sessions will make the channel look irrelevant while it quietly produces work.

What should you measure instead of a conversion multiple?

We measure whether you appear in the answer at all, per surface, using a published protocol, rather than inferring it from traffic. Traffic from assistants is under-attributed by every analytics tool we have tested, so a visibility measurement and a traffic measurement answer different questions.

And we would rather tell you the conversion multiple is unmeasured in your market than repeat a number from someone else’s website.

“We would rather tell you the conversion multiple is unmeasured in your market than repeat a number from someone else's website.”