AI Traffic Converts Better Than Search. It's Also 0.2% of Your Visits.

Two numbers landed within weeks of each other, and read alone each one supports a different strategy. Adobe measured AI-referred visitors to US retail sites converting 54% better than everyone else. Contentsquare measured AI-referred visits at 0.2% of all visits. The first says drop everything and optimize for AI search. The second says ignore it for another year. Both figures are sound, and the reason they coexist is the most useful thing in either dataset.

Key takeaways

Adobe's May 2026 retail data, reported in June, is the strongest single reading on the quality of AI-referred traffic, mostly because of its size: over a trillion visits to US retail sites, with AI referrals tracked since October 2024.

The direction matters more than any single figure, because it flipped. Digital Commerce 360's report on the same data notes that a year earlier AI-referred visitors were converting at close to half the rate of non-AI visitors. Revenue per visit alone moved from a 37% premium in March 2026 to 53% in May. A channel that was worse than average in early 2025 became materially better than average by mid-2026.

Traffic grew alongside it: 138% year over year in May 2026, the highest share of total retail visits since Adobe began tracking. Vivek Pandya, Adobe's Director of Digital Insights, frames the opportunity as personalisation once the shopper continues on the retailer's own site.

Read that carefully, though. A conversion premium describes the visitors a channel delivered. It does not describe how many it delivered, and it is not a growth forecast.

The strongest explanation for the premium came from a company with the full funnel in view. On 5 August 2026, Shopify reported that AI-driven traffic and AI-driven orders each tripled year over year in Q2, on revenue of $3.6 billion.

The mechanism is in the landing behaviour. President Harley Finkelstein said half of all AI-referred sessions on Shopify go straight to a product description page — 2.5x the rate for sessions arriving through traditional search — and that new buyer orders from AI channels arrived at nearly twice the rate of other channels.

That is the whole story of the conversion premium in one statistic. Traditional search delivers a visitor at the beginning of an evaluation. An AI assistant does the evaluation in the chat, then delivers a visitor to the specific product it decided on. The comparison "AI traffic versus search traffic" is not comparing two sources of the same visitor. It is comparing a visitor who is still deciding with a visitor who has decided.

Two more findings from the same call are worth keeping. Shopify said 75% of AI-attributed purchases fell outside its top 100 categories, consistent with retrieval systems surfacing niche inventory that keyword ranking buried. And it reported that AI search running against its structured catalog converted at twice the rate of AI search relying on scraped product data — a claim that happens to describe a product Shopify sells, and which no independent party has replicated, but which points the same way as Adobe's machine-readability finding below.

Finkelstein's own framing is the part most coverage skipped: AI is "a complement to search, rather than a substitute for it." Traditional search sessions were up 1.3x over two years and still account for roughly a third of storefront sessions. The largest disclosed dataset on AI commerce does not describe search being replaced.

Contentsquare's benchmark, built from 99 billion sessions across 6,500 sites, is the counterweight. AI-referred traffic grew 632% year over year — and reached 0.2% of total visits.

Its conversion figure also looks nothing like Adobe's: 1.3% for AI-referred traffic in 2025, up 55% from 0.8% the year before, against 1.9% for email.

These are not rival measurements of one quantity, and treating them as a contradiction is the easiest mistake to make with this data.

The period is doing most of the work. Adobe's own series shows the flip happening between early 2025 and 2026; a full-year 2025 average sits mostly on the wrong side of it. The comparison group does the rest: beating a blended site-wide average is a much lower bar than beating email, the channel that sends messages to people who already gave you their address.

Put plainly: AI-referred traffic converts well relative to the average visitor, is still catching your best-performing owned channel, and arrives in volumes small enough that a 632% growth rate barely moves its share of sessions.

Every number above rests on the same fragile signal: a referrer header naming an AI platform. When that header is absent, the visit is not "AI-referred with a missing label," it is direct traffic, and it disappears from the analysis entirely.

Two mechanisms remove it, and neither involves the user doing anything. Links opened from an assistant's mobile app render in an in-app browser that does not pass a referrer. Links marked rel="noreferrer" block it by design. Cloudflare is explicit about this in the methodology for its own crawl-to-refer ratio, noting that native app traffic lacks referral headers and may therefore overstate how one-sided the crawl-versus-referral picture looks.

The scale of the loss is genuinely unknown — the vendor estimates in circulation range from roughly a third of AI sessions to well over two thirds, measured on samples too small and too self-selected to settle it. What is not in dispute is the direction. Your measured AI referral traffic is a lower bound.

There is a subtler consequence for the conversion numbers themselves. If a meaningful share of AI-referred visits never carries a referrer, then every conversion premium published so far was measured on the subset of AI visitors whose platform, app, and link happened to pass one. Whether that subset behaves like the whole is untested. The premium is well evidenced; its precision is not.

Not the case for AEO, which never rested on referral clicks. What it changes is the reporting.

The temptation with a 54% conversion premium is to run it forward: multiply by projected AI traffic growth and produce a revenue case. That calculation compounds two numbers that are moving for different reasons — a premium that reflects who AI sends today, and a volume that reflects how often engines hand over a click at all. The second is the engine's product decision, and it has been trending toward answering in place rather than passing traffic on. AI Overviews now appear on around 43% of US searches, and Pew measured clicks on 8% of results pages carrying an AI summary against 15% without one.

So the honest position is the unglamorous one. AI referral traffic is a small, high-quality channel whose size you cannot control and whose measurement you cannot fully trust. The mention in the answer, which reaches everyone who reads it whether or not they click, is the thing worth managing — and most of the citations behind those mentions point at domains you do not own anyway.

One tactical item does come straight out of this data. Adobe scored retail product pages at 66% machine-readable, against 75% for homepages and 74% for category pages. The pages where AI-referred visitors actually land are the pages agents parse worst. Fixing that — real HTML, visible prices and specifications, structured data, no content that only exists after JavaScript runs — is cheap, and it is the same work that makes those pages retrievable in the first place.

Elmo is an open-source, self-hosted AI visibility platform that runs your prompt sets across ChatGPT, Claude, Gemini, Perplexity, and Google's AI surfaces, recording every mention, cited URL, and competitor named alongside you. Because the data is yours to query directly, joining it to whatever AI referral sessions your analytics does capture is a query rather than a vendor integration.

For the fundamentals, start with answer engine optimization and how to track your brand in AI search. For what to report to a board, see the IAB's AI visibility metrics; for the referral data platforms do expose, Microsoft Clarity's AI citation reporting. For the vocabulary, see the AI search glossary.