By Carson Cobb. Updated November 2026.
AI search attracts more confident claims than evidence. Some are true, some are reasonable guesses, and some are sold as facts with nothing behind them. This is my running scorecard of the claims I hear most from Shopify brands and the agencies pitching them, graded by what can actually be shown.
Grades: Proven means a primary source or a measurement shows it. Likely means good evidence points that way, but nothing settles it. Unproven means there’s no real evidence either way yet. False means a primary source or a measurement contradicts it.
The short version
- Proven: AI referred shoppers convert better than search shoppers, and most of them land on product pages.
- False: AI is replacing Google for most stores, blocking GPTBot keeps you out of ChatGPT, and AI crawlers see everything on your page.
- Unproven: llms.txt and “AI friendly” content chunking change what AI says about you.
- Likely: structured data helps AI get product facts right, and outside sources decide who gets recommended.
- Unproven: that duplicate product data on its own confuses AI. Splitting the rating away from the product is what caused errors in my test.
The scorecard
| # | Claim | Grade |
|---|---|---|
| 1 | AI referred shoppers convert better than search shoppers | Proven |
| 2 | Most AI referred shoppers land on product pages | Proven |
| 3 | AI search is replacing Google for online stores | False |
| 4 | Blocking GPTBot keeps you out of ChatGPT | False |
| 5 | AI crawlers see everything on your product page | False |
| 6 | You can buy directly inside ChatGPT from most Shopify stores | False |
| 7 | You need an llms.txt file to show up in AI answers | Unproven |
| 8 | Rewriting content into small “chunks” helps AI rank you | Unproven |
| 9 | Structured data helps AI understand and describe products | Likely |
| 10 | Any duplicate product data confuses AI | Unproven |
| 11 | Outside sources decide which brands AI recommends | Likely |
| 12 | Your AI traffic numbers show the full picture | False |
The evidence, claim by claim
1. AI referred shoppers convert better. Proven.
Shopify reported in May 2026, from its own storefront data for the first quarter, that AI referred visitors who land on a product page convert at nearly 50% higher rates than organic search visitors, and that AI attributed orders run 14% higher in value. AI referred orders grew nearly 13 times year over year. One caveat: the conversion comparison only covers sessions that start on a product page.
2. Most land on product pages. Proven.
Same Shopify data: about 55% of AI referred sessions start on a product page, compared with about 20% for organic search. That’s why product page clarity matters more for AI traffic than for search traffic. The visitor arrives already told what the product is, and the page has to confirm it.
3. AI is replacing Google. False, for now.
Shopify’s same report says organic search still sends more sessions to its merchants than all tracked AI platforms combined, and same store organic sessions grew about 5%. AI is a fast growing, high intent channel next to search, not a replacement for it.
4. Blocking GPTBot keeps you out of ChatGPT. False.
OpenAI’s crawler documentation is explicit: GPTBot is for model training, and only OAI-SearchBot’s robots.txt setting decides whether a site can appear in ChatGPT search answers. In my crawler access report, none of 94 Shopify stores blocked OAI-SearchBot.
5. AI crawlers see everything on your page. False.
A 2024 analysis of crawler traffic on Vercel’s network found the crawlers from OpenAI, Anthropic and Perplexity don’t run JavaScript. In my scan, 11% of Shopify stores had product data that only appeared after scripts ran, and 3% had no product data at all until then.
6. You can buy inside ChatGPT from most Shopify stores. False.
OpenAI launched Instant Checkout in September 2025 and pulled back from it in March 2026, after only a handful of Shopify merchants went live. ChatGPT now works mainly as a discovery and referral channel that sends shoppers to stores and retailer apps.
7. You need llms.txt. Unproven.
Google’s AI optimization guide lists llms.txt as something you can skip. No major AI search product says it reads the file, and the public crawler log data shows AI bots rarely request it. Shopify now generates one for 95% of stores anyway. Details in Does llms.txt do anything?
8. Chunking content for AI helps. Unproven.
Google’s same guide groups content chunking and AI specific rewriting with the tactics you don’t need. What does help is a page that answers real questions clearly, which was good practice before AI. See product data depth.
9. Structured data helps AI describe products. Likely.
Microsoft’s Fabrice Canel said publicly that schema markup helps Microsoft’s models understand content. Google says no special markup is required for its AI features, while its standard product structured data still feeds its shopping results. In my own test of 24 products, ChatGPT gave the right price for all 20 that had product structured data, and couldn’t find a price for 2 of the 4 that had none. Ratings were less reliable either way. See Does schema change what ChatGPT says?
10. Any duplicate product data confuses AI. Unproven.
29% of Shopify stores publish two or three separate Product blocks on the same page, usually because a review app adds its own. In my test, ChatGPT described 6 of 8 duplicate products correctly when the blocks agreed. The errors came from a rating split into its own fragment, or added by a script. Fix that specific problem first. See the duplicate schema report.
11. Outside sources decide who gets recommended. Likely.
In my test of 48 buyer questions across six categories, 45 answers showed sources, led by independent review and testing sites and magazines, with medical and veterinary sources in health categories. Stores that ChatGPT named scored only slightly higher on site readiness than stores it didn’t. See the sources ChatGPT cites.
12. Your AI traffic numbers show the full picture. False.
Shopify notes that clicks from Google’s AI Overviews are counted as organic search, so AI influence is undercounted. Many shoppers also read an AI answer, then search the brand name or type the URL. A post purchase survey question (“Where did you first hear about us?” with an AI option) catches what analytics misses.
What to do with this
- Ignore anyone leading with llms.txt or content chunking. They’re selling the least important work.
- Check the one robots.txt line that matters, OAI-SearchBot, and leave the rest to a business decision.
- Make your product pages machine readable and consistent. One product entity, real ratings, the right brand name, and facts in text.
- Earn outside mentions in the sources AI reads in your category. That’s what gets you named.
- Measure with a survey question, not just referral traffic.
How this was done
- Grades are mine, based on the sources listed under each claim: Shopify’s AI search insights (May 2026), OpenAI’s crawler documentation, Google Search’s AI optimization guidance and staff statements as reported by Search Engine Land and Search Engine Journal, Vercel’s AI crawler analysis (December 2024), reporting on OpenAI’s Instant Checkout changes (March 2026), and Microsoft’s public comments on schema.
- My own data comes from the other reports in this series: a structured data scan of 94 US Shopify stores and live ChatGPT tests run in November 2026.
- I’ll update grades as new evidence comes out, and note each change here.
No store is named in this report. Run the free AI Visibility Check to see what AI says about your brand today.