ChatGPT picked three names, and you either were one of them or you weren't.
AI-referred retail traffic grew 393% year over year in Q1 2026, and it converts about 42% better than traditional organic traffic, according to Adobe Analytics (via Triangle Direct Media's 2026 agentic commerce writeup). Here's a free 5-minute test to find out which side you're on.
The sale that happens without you
Picture it: 9pm, a customer types "best sustainable activewear under $100, ships here" into ChatGPT. The response comes back with three brands, each with a Buy button. She taps one. The order is confirmed thirty seconds later, straight from Shopify checkout. No store visit, no comparison shopping, no chance for anyone else to enter the conversation.
That scenario comes from a guest post by Kristina Tyumeneva on the Debutify blog, and it isn't hypothetical. OpenAI shipped Instant Checkout with Shopify in 2025, and the integration lives in millions of stores. The plumbing for buying inside a chat is done. The open question is a different one: does the AI recommend your store at all?
Here's the uncomfortable part. In classic SEO, you could at least watch impressions climb in Search Console. You knew Google was showing you, even when nobody clicked. AI search gives you nothing like that. Either your brand is in the recommendation or it isn't, and unless you actively look, there's no notification that a customer searched for exactly what you sell and bought from someone else thirty seconds later.
The 5-minute test (free)
The method below is Tyumeneva's, from that same Debutify post, and it's the best free baseline I've seen. Five minutes, zero cost:
- Open a private browser window. ChatGPT personalizes answers from your history, which ruins the test, so go incognito.
- Log out of your ChatGPT account for the same reason.
- Go to chatgpt.com and ask the way a customer would: "Best sustainable activewear under $100 that ships to the US." Replace the example with your category and your buyers' typical filters.
- Run eight to ten variations. Swap the price point. Change the region. Use different adjectives your customers actually use ("ethical" vs. "sustainable", "under $50" vs. "affordable").
- Write down every store that gets named across all your variations.
That's it. You now have a baseline: the set of stores ChatGPT actually recommends for your category, and whether yours is in it.
What to do with what you find
Zero mentions. ChatGPT doesn't know your store exists. This is more common than people think, especially for stores under a year old or with thin content beyond product pages. A tracking tool won't fix this. You need to give the internet enough to say about you that AI has something to pull from: reviews on independent sites, product content that answers real customer questions, mentions from niche blogs in your category. Then come back and re-test in three months.
Some queries, not others. This is the interesting case. The pattern of where you appear and where you don't is a roadmap. Usually the gap runs between branded queries ("best activewear brands 2026") and discovery queries ("good workout top for cold weather running"). If you're in the first and missing from the second, you need content that matches how people describe their problem instead of how you describe your product.
Consistent appearances. Screenshot your baseline before your next campaign or content push, run it for a month, and look at what moved. Same discipline as A/B testing, different signals. And because almost nobody else in your category is measuring this yet, you'll know what works before your competitors figure out there's anything to measure.
Why the manual version stops working
The five-minute test is a sanity check, and it's honest about its limits. It can't catch drift week over week. It can't compare your share of voice against competitors over time. And as Tyumeneva puts it, running it manually once a month is the kind of thing you mean to do and then don't.
A systematic version of the same idea has three upgrades. First, a fixed prompt panel: the same 20 to 40 prompts, in the same order, every run, mirroring how buyers describe their problem. That's the only way week-over-week comparisons mean anything. Second, cross-model coverage. ChatGPT holds roughly 80% of the AI search market according to the Debutify piece, so it's the right place to start, but Perplexity, Gemini, and Google's AI Overviews resolve the same query differently. Third, evidence instead of vibes: screenshots of every response, which stores were named, what facts the AI stated about each. Six weeks later, when a competitor launches a content push or OpenAI tweaks retrieval, you have a before-and-after instead of a feeling.
What should the prompts actually be? Write them the way buyers talk, not the way your product page talks. Mix three query types: discovery ("best running shoes for flat feet under $120"), comparison ("Brand X vs Brand Y for marathon training"), and problem-led ("knee pain when running, what shoes help"). Ten of each is a solid starter panel of thirty. Avoid your own brand name in the prompts; branded queries tell you about existing awareness, not about being discovered.
Run the manual test today. It costs nothing, and the answer might genuinely surprise you.
Top comments (0)