AI traffic converts worse than the hype, except for complex products
An analysis of 164 million purchases across 973 ecommerce sites found AI-referral traffic drove less revenue per session than every channel except paid social, and was only about 0.2% of visits. It performs far better for complex, research-heavy products.
Quick Summary
There is heavy pressure to pour budget into generative engine optimization so AI chatbots recommend your products. Before you do, weigh the evidence. An analysis of 164 million purchases across 973 ecommerce sites in 49 countries found that AI-referral traffic drove less revenue per session than every digital channel except paid social, and made up only about 0.2% of visits in the study period.
The nuance that matters: AI traffic performs much better for complex, research-heavy products (electronics, health, vehicles, many B2B purchases), where it had 4.6 times the traffic share of simple-product stores. Treat AI as a mid-funnel channel, prioritize it if your products are researched heavily, and keep it lower on the list if you sell simple goods.
Key Finding
0.2%
Share of ecommerce visits that came from AI referrals across 973 sites between August 2024 and July 2025.
Kaiser & Schulze, 2026
Every store is being told to optimize for ChatGPT and other AI assistants right now. The instinct is understandable: as people ask chatbots for recommendations instead of searching, you want to be the answer. But how good is that traffic actually? Researchers analyzed 164 million online purchases across 973 ecommerce sites to find out, and the picture is more measured than the hype.
What the data shows
- Low revenue per session. AI-referral visitors generated less revenue per visit than every channel measured except paid social.
- Weaker conversion. AI referrals converted about 11.5% below organic search and 46.2% below affiliate traffic.
- Still a small slice. Only around 0.2% of visits came from AI referrals between August 2024 and July 2025, mostly ChatGPT.
- But rising and uneven. Conversion from AI traffic is improving over time, and it skews toward younger (5.5x) and more tech-savvy (3.8x) audiences.
Where AI traffic does pay off
The standout finding is about product type. Stores selling complex, research-heavy products had 4.6 times the AI traffic share of stores selling simple ones. The logic is human: shopping through an assistant adds steps, so people only bother when the research payoff is worth it. Choosing a mattress, a laptop, or a supplement rewards that effort. Buying a plain t-shirt does not.
What to do on your Shopify store
- Treat AI as a mid-funnel channel, useful for shoppers actively comparing options, not a top-of-funnel firehose.
- Prioritize it if your products are researched heavily. Publish clear, text-based comparisons against alternatives, since that is what people ask assistants for.
- Add substantive Q&A content to product pages. It helps assistants answer questions about your product and tends to lift ratings too.
- Lead with benefits, not just specifications, so both shoppers and assistants can match your product to a need.
- Keep it in proportion if you sell simple products. Do not divert budget from proven channels to chase a 0.2% slice yet. Get the fundamentals right first.
A caveat worth keeping
This is a working paper, not yet peer-reviewed, and it used last-click attribution, so some journeys that started with AI but finished elsewhere may be undercounted. Showing up in AI answers also builds brand awareness that this kind of analysis does not capture. The direction is clear enough to plan around, but expect the numbers to move as adoption grows.
The practical takeaway lines up with our companion piece on how to bias AI agents toward your products: make your titles, ratings, and comparisons strong, but keep investment proportional to how much your customers actually research before buying. If you want help deciding where AI readiness sits against higher-impact conversion work, that is a good question for a focused UX audit.
Research: Kaiser, M. & Schulze, C. (2026). Working paper, University of Hamburg and Frankfurt School of Finance & Management. https://dx.doi.org/10.2139/ssrn.5585812
Is this broken on your store?
Get a free UX review.
We review your store against 50+ conversion principles - including this one - and send you a detailed breakdown of what to fix and why.
Get a UX AuditFind your conversion leaks.
A focused expert review of your store with Figma redesigns and a Loom walkthrough. Pick one page or get the full picture.