AI Referral Value by Engine

A vendor benchmark reports higher revenue per visitor from Perplexity and Claude than from ChatGPT. A separate client study reports a higher conversion rate from ChatGPT. The metrics differ, and the benchmark's own arithmetic also needs checking.

A blended AI-traffic total hides differences among engines and among business models. The useful comparison keeps the conversion event, revenue definition, cohort, and period visible.

Revenue per visitor and conversion rate answer different questions. A recorded conversion can be a lead or another configured event; it is not automatically a purchase. Neither metric supplies a universal engine ranking.

What the revenue benchmark reports

Attrifast's May 26, 2026 vendor benchmark reports 200 Stripe-connected SMB sites through May 2026. Its blended revenue-per-visitor values are Perplexity $1.42, Claude $1.18, ChatGPT $0.87, Gemini $0.41, and AI Overviews $0.29.

The same report lists ChatGPT at 71 percent of sessions and 53 percent of revenue, Perplexity at 8 and 19 percent, and Claude at 6 and 14 percent. Those shares do not reconcile with the reported revenue-per-visitor figures under one pooled denominator. Engine RPV multiplied by session share and divided by revenue share implies a pooled RPV of roughly $1.17, $0.60, and $0.51 respectively. Rounding cannot explain that spread.

The source labels both comparisons cohort-blended, without resolving the discrepancy. Different aggregation methods could matter, but I cannot invent one to make the figures agree. Retain these as reported vendor figures, not a verified allocation model or a basis for shifting budget. The Stripe-only SMB sample also limits where the results apply.

What the conversion-rate study reports

Seer Interactive's single-client case study, covering October 2024 to April 2025, reports conversion rates of 15.9 percent for ChatGPT, 10.5 percent for Perplexity, 5.0 percent for Claude, 3.0 percent for Gemini, and 1.76 percent for Google organic. Those are the client's recorded conversions, not a universal purchase rate.

The case reports 1,370 AI-referred conversions and nearly 14 million organic sessions. Its industry is undisclosed. One funnel and buying cycle cannot establish the ranking for another business.

Rate and revenue are not the same question

Conversion rate divides recorded conversion events by the eligible visits or users under the study's definition. Revenue per visitor divides the defined revenue by visitors. Within one comparable dataset, larger purchases can offset a lower purchase rate.

That mechanism is a hypothesis here. These studies use different sites, windows, and measurement methods, and one contains unresolved arithmetic. Product mix, conversion definitions, attribution windows, and customer mix could also change the rankings. They do not establish that ChatGPT buyers spend less.

Measure both metrics in your own funnel before deciding which assistant deserves more work. If the conversion event is a lead, follow it to accepted pipeline and the appropriate financial outcome rather than treating it as a sale.

What a later referral report shows

Goodie Research Lab's May 21, 2026 report compares May to August 2025 measurements, published that October, with March to April 2026 averages from a fresh anonymized brand panel. It reports Claude at 1.4 and 18.5 percent, and ChatGPT at roughly 89 and 63 percent. It does not establish that the same brands were retested. Its five reported engine percentages total 103 percent, an unresolved allocation problem in the published figures.

Use a dated study as a comparison point. Refresh your own engine-level results on a schedule that fits the volume and buying cycle.

What actually changes if you take this seriously

Stop reporting a single AI-traffic number. Split revenue and conversion rate by engine, the same way you already split them by campaign. GA4's native AI Assistant channel gives you the medium, but not a clean per-engine revenue view out of the box; you have to build that cut yourself, using source dimension against the domains each assistant actually refers from.

Allocate effort using your own accepted outcomes and measured revenue contribution. More sessions do not automatically justify more budget, and the unresolved vendor-share table cannot settle the decision.

Keep the reporting window, source definitions, and conversion event stable enough to compare periods. Dark AI traffic measurement covers unknown origins; pipeline reporting covers the commercial handoff.

Frequently asked questions

Do Claude and Perplexity really convert better than ChatGPT?

Attrifast reports higher revenue per visitor for Perplexity and Claude, but its revenue shares do not reconcile with those values under one pooled denominator. Seer reports a higher ChatGPT conversion rate in a separate single-client study. Neither establishes a universal ranking.

What is revenue per visitor and why does it differ from conversion rate?

Conversion rate measures a defined conversion event among eligible visits or users. Revenue per visitor divides defined revenue by visitors. Purchase size can change their ranking within one dataset, but unrelated studies do not prove that mechanism.

How fast is Claude's referral traffic growing?

Goodie reports Claude at 18.5 percent for March to April 2026 versus 1.4 percent for May to August 2025. Wave 2 used a fresh brand panel. The report does not establish an unchanged cohort or a universal growth rate.

Should I trust a single conversion-rate study?

Use it as a scoped comparison. Seer's study covers one undisclosed client from October 2024 to April 2025, with 1,370 AI-referred conversions. The conversion definition, buying cycle, and site mix can differ from yours.

How do I track AI referral value by engine instead of lumping it together?

Report visits, the defined conversion rate, accepted pipeline, and defined revenue by observed assistant source. Keep periods and attribution rules consistent, retain unknown sources, and investigate small samples or conflicting figures before moving budget.

Sources

Sources: Attrifast AI Traffic Revenue Benchmark 2026 (May 26, 2026; its published RPV/share inconsistency remains unresolved); Seer's single-client conversion study (October 2024 to April 2025); Goodie's AI search traffic report (May 21, 2026). These are separate datasets.

About the author

Jeff Brokaw is a CMO and Certified Chief AI Officer who builds AI workflows, measurement systems, and the commercial operations around them.

AI Referral Value by Engine: Revenue, Conversion, and Conflicting Figures