Does AI-Referred Traffic Convert? What the Evidence Actually Says

Short answer: on the most recent large dataset, yes. Adobe Digital Insights reports that AI-referred visitors to U.S. retail sites converted at a 48% higher rate than non-AI traffic in August 2026, and generated 43% more revenue per visit. But that is a recent reversal, it covers large retailers on Adobe Analytics, and a peer-reviewed study of 973 ecommerce sites found the opposite for August 2024 to July 2025. Whether AI traffic converts for your store is a question your own GA4 and Shopify data can answer, and this article shows how.

It is written for ecommerce merchants and the agencies reporting to them who are deciding how much to invest in AI channels and keep running into contradictory headlines. We opened every study cited here and state who ran it, when, and on what sample. Studies we couldn't open, or that don't disclose a sample, are named as such.

AI-referred traffic

Website sessions that arrive by a click from an AI assistant such as ChatGPT, Perplexity, Gemini or Copilot, identified by the referrer the assistant passes. It excludes Google's AI Overviews and AI Mode, which GA4 counts as Organic Search, and it misses AI visits whose referrer was stripped, which land in Direct.

Conversion rate

The share of sessions that end in a purchase or another key event. GA4 calls the session version Session key event rate: "the percentage of sessions that converted". Revenue per session (or per visit) divides purchase revenue by sessions and captures order value as well as conversion.

What do the published studies say?

Five sources get quoted most often. They are listed here by sample size and how much of their method they disclose, not by how flattering their result is.

Adobe Digital Insights: the largest dataset, and a reversal

Adobe's figures come from Adobe Analytics on U.S. retail sites, which Adobe describes as covering more than 1 trillion visits and 100 million SKUs, aggregated and anonymized. Its September 2026 holiday forecast says AI-referred retail visitors converted at a 48% higher rate than non-AI traffic in August 2026, generated 43% more revenue per visit (down from a record 53% in July) and added to cart at a 32% higher rate.

The trend matters as much as the level. Adobe's Q1 2026 analysis says AI traffic converted 42% better in March 2026, against 38% worse in March 2025. Its June 2026 report puts May 2026 at 54% higher, and says that twelve months earlier non-AI visits were worth 128% more than AI visits. So the sign flipped within about a year.

Limits: Adobe's sample is retailers that use Adobe Analytics, which skews toward large brands. The non-AI baseline is everything else, described in Adobe's Q1 post as covering channels such as paid search and email marketing, and it isn't broken out channel by channel in the material we opened. The reports we read don't list which referrers Adobe counts as AI.

Kaiser and Schulze: the peer-reviewed counterpoint

Maximilian Kaiser (University of Hamburg) and Christian Schulze (Frankfurt School of Finance & Management) studied 973 ecommerce websites with combined revenue above $20 billion, from August 2024 to July 2025. The University of Hamburg's summary (November 26, 2025) says the data was first-party data from the participating sites, comparing over 50,000 transactions attributed to ChatGPT with 164 million from traditional channels.

The finding: ChatGPT traffic converted and earned revenue per session above paid social but below all other established channels, including Google's paid and organic search. ChatGPT referrals bounced less, conversion improved during the study period, and falling order values partly offset that gain. The Frankfurt School release (June 22, 2026) confirms publication in Marketing Science, puts LLM traffic below 0.2% of total traffic in the sample, and says ChatGPT is most useful for complex, information-heavy purchases. We could not open the journal page itself, so the details here come from the two universities' summaries.

Similarweb: an estimate, not a measurement

Similarweb's September 15, 2025 release says it "estimates" the conversion rate for ChatGPT-referred visits at 11.4%, against 5.3% for organic search, for July 2024 to June 2025. Similarweb describes its data as estimated metrics generated by proprietary algorithms from multiple sources. It's useful as a market-level signal, but it is modelled from panel and partner data rather than read from stores' analytics.

Ahrefs: one site, one kind of conversion

Ahrefs' June 16, 2025 post reports that AI search was 0.5% of its own site's traffic over 30 days but produced 12.1% of sign-ups, a 23x higher conversion rate than traditional organic search. It is one software company's website and one kind of conversion, a sign-up, not an ecommerce purchase. The author called the engagement picture a close call: bounce rate and pages per visit were higher for AI visitors, visit duration lower.

The numbers to set aside

Semrush's July 2025 AI search study says the average AI search visitor is 4.4 times as valuable as a traditional organic visitor, "based on conversion rate". The article doesn't state which sites, how many or what period the figure comes from, so it can't be weighed against the studies above.

Shopify's Spring '26 Edition (June 17, 2026) says AI searches powered by Shopify Catalog convert at twice the rate of those using scraped data. That compares two ways of supplying product data to AI channels. It is not a comparison of AI traffic with search or email traffic, so don't use it as one.

Google says that clicks from results pages with AI Overviews are "higher quality", with users more likely to spend more time on the site, in its AI features documentation. It publishes no figures with that statement, and those clicks sit in Organic Search, not in any AI channel.

Why do the studies disagree?

Mostly they aren't measuring the same thing. Six differences explain most of the spread.

Two more effects apply to every study. First, intent selection: someone clicking through from an AI answer has often already compared options in the chat, so the visit is the end of a journey that other channels see from the start. That inflates the conversion rate of the last click without proving AI created the demand. Second, small denominators: AI is a small share of traffic (under 0.2% in Kaiser and Schulze's sample, 0.5% on Ahrefs' site), so a modest number of orders moves the rate a lot.

How do you measure your own AI conversion rate?

Set up a channel that catches AI referrers, compare it with specific channels rather than the site average, and wait until there are enough orders to trust the comparison. The AI traffic and attribution guide explains where every kind of AI visit lands; the steps below are the conversion part.

1. 01

   Catch every AI referrer in GA4

   GA4's default AI Assistant channel covers referrers on Google's list, which isn't published. To catch the rest, create a custom channel group under Admin > Data display > Channel groups > Create new channel group (Editor role or higher). Add an AI channel whose condition is Source matches regex .*(chatgpt|openai|perplexity|gemini\.google|bard\.google|copilot|claude\.ai|deepseek|grok|meta\.ai|chat\.mistral|you\.com|phind|poe\.com).*, then use Reorder to place it above Referral. Google says custom channel groups apply to reports retroactively.

2. 02

   Know the regex's limits

   It over-matches any source containing those words, and AI tools change domains, so review the list every quarter. It can't see AI visits that arrive with no referrer, and it doesn't include Google's AI Mode, which stays in Organic Search. See how to estimate AI Mode traffic.

3. 03

   Pull the four metrics per channel

   For each channel in your custom group: Sessions, Session key event rate, Ecommerce purchases and Purchase revenue. Divide revenue by sessions for revenue per session. Use complete calendar months.

4. 04

   Compare like with like

   Put the AI channel next to Organic Search, Paid Search, Email and Referral, not the site average. If your volume allows, split new and returning users, because a returning customer asking ChatGPT for your store converts very differently from a stranger.

5. 05

   Cross-check in Shopify

   In the Shopify admin, Sales channels > Agentic shows sales, orders, online store sessions and online store conversion for each AI channel. Shopify notes that sales there combine referral-based sales with direct checkout, so they won't match GA4 exactly. Shopify's Sessions by referrer report is a third view of the same visits.

6. 06

   Check the paths, not just the last click

   GA4's attribution paths report, under Advertising > Attribution, shows whether AI visits tend to be early touches followed by a later branded or direct visit. Google says all GA4 attribution models exclude direct visits from credit, so if a shopper clicks from ChatGPT and later returns by typing your URL, the sale goes to the last non-direct touch, which may be the AI visit.

How many orders before the rate means anything?

A conversion rate from a small number of sessions has a wide margin of error. The standard error of a rate p over n sessions is √(p(1−p)/n), and a rough 95% interval is two standard errors either side. At a 2% conversion rate over 1,000 AI sessions, that's 20 orders and an interval of about 1.1% to 2.9%, wide enough to overlap most channels on a typical store. At 10,000 sessions it narrows to roughly 1.7% to 2.3%.

What should you do with the answer?

Whatever your rate turns out to be, AI referrals are probably a small share of sessions for now. The decision isn't "AI or not" but how much effort AI visibility deserves relative to its revenue and its growth.

• If AI converts at or above your organic search rate: the traffic is worth more of it. Prioritize the product data and eligibility work that gets you recommended; the AI shopping channels guide maps where products get into each assistant.
• If AI converts below organic but is growing: check landing pages first. AI tools often send visitors to deep product pages; if those pages contradict what the assistant said about price, stock or sizing, the visit fails at the landing.
• If you have too few orders to tell: track revenue and sessions monthly and revisit the rate each quarter. The AI revenue calculator gives a rough sense of what's at stake in the meantime.
• In every case: look beyond the first order. Whether AI-acquired customers come back is a separate question; see LTV math for AI-acquired customers.

For the setup itself, the GA4 AI Assistant channel guide covers the default channel and the regex in detail, tracking ChatGPT traffic on Shopify covers the Shopify side, and the monthly AI revenue report shows how to present the conversion comparison every month without overclaiming.