Key takeaways
- About 22% of US checkouts had AI use in the 24 hours before by mid September, up from 16% in June.
- AI chat apps alone went from about 10% to 12%. Most of the late-August jump came from Google AI Mode.
- Desktop checkouts lean in more, about 3 in 10 compared with about 2 in 10 on phones.
- Analytics mostly miss this because the AI step happens before the visit to the store.
I wanted to know how close to an actual purchase people are using AI. A lot of what I talk about with customers is AI showing up early in research, so we looked at the other end, the hours right before someone checks out.
We used an opt-in US consumer panel that records web browsing and AI app use. Every time a panelist loaded a checkout or cart page on any site, we checked whether that same person had used an AI assistant in the 24 hours before. In the first week of June that was 15.9% of checkouts. By the week of September 14 it was 22.5%, so about 1 in 5 and still climbing (we can see that they used AI, not what they asked it, so some of that is people asking ChatGPT about something totally unrelated to the purchase).

What counted as AI
The list is ChatGPT, Gemini, Claude, Perplexity, Copilot, Grok, Meta AI, DeepSeek and Google AI Mode. Google AI Overviews are left out, since pretty much anyone who googles anything sees one and that would blow the number up. A checkout here is a day where someone loaded a checkout or cart page, on any site.
Google drove the late-August jump
If you only count the chat apps (everything above minus AI Mode), the line moves a lot slower. It went from 9.8% in early June to 11.9% by mid September, which is real growth but pretty steady. The gap between the two lines opened up in late August, and most of that came from Google AI Mode.
The weekly numbers bounce around a bit too. Any AI hit almost 23% the last week of August, dropped under 19% the week after, then came right back to 22.5%, so I would watch the trend more than any single week. Either way, I honestly think a lot of brands track ChatGPT and skip AI Mode, and AI Mode is where most of the recent growth showed up.
Desktop leans in more
About 3 in 10 desktop checkouts had AI use in the 24 hours before. On phones it was about 2 in 10. My guess is bigger purchases still happen on a laptop, where people have a few tabs open and are comparing stuff, but that part is me guessing (we did not split it by product type).
Why your analytics don’t show it
The AI step happens before the visit to the store. Someone asks ChatGPT which running shoes are good for flat feet, gets a few names, then opens a new tab and types in the brand or googles it. By the time they check out, analytics credits direct or branded search and the answer that sent them never shows up anywhere.
What I’d do with this
- Ask customers how they found you. Add AI as an option on your post-purchase survey or “how did you hear about us” field (ChatGPT, Google AI Mode and such), because nothing else will show it to you.
- Watch branded search and direct traffic. If those are climbing and you can’t explain why, some of it is probably AI.
- Test what AI says close to the purchase. Ask the questions people ask right before buying, like which one is best for a specific need or whether a brand is worth it. I put together 50 prompts to test your brand in ChatGPT if you want a starting point.
- Don’t skip Google AI Mode. It drove most of the recent jump, so it should be on your list when you decide which AI models to track.
Sources
- Searchable analysis of an opt-in US consumer panel that records web browsing and AI app use, weekly, June 1 to September 20, 2026.




