Start with the thing nobody checks
Before anything else, confirm the AI crawlers can fetch the pages that matter. GPTBot, ClaudeBot, PerplexityBot and Google Extended all have to be allowed at the server and the CDN. A store can be silently blocking every one of them.
It is still search underneath
The same things that decide whether Google ranks you decide whether an AI engine cites you: pages it can reach, real authority, and something only you can say. What AI search adds is a measurement problem, not a new set of tricks.
A catalog is not an article
Advice written for software companies assumes your value is explained in prose. A store's value is in a catalog: cuts, weights, breeds, grades, shipping windows. If that detail only exists inside a variant picker, an engine has nothing to quote. For a food brand that means writing down the breed, whether it is grass finished, how long it hangs and who processes it.
Get the catalog onto the free surfaces first
On one store the highest-value products had never been published to Google's free listings. Sales from Google went from $285 in the slowest 30 days before the work to $8,668 in the best 30 days after, by Shopify's own reports. Before optimizing for answer engines, get the catalog onto the surfaces that are already free.
Measure mentions and citations separately
Build a fixed list of the questions your buyers ask. Run it across ChatGPT, Perplexity, Gemini and Claude. Record whether you are mentioned, whether you are cited, and who is named instead. Most tools blend those into one score, and the score is quietly wrong.
What not to buy
A text file no major engine has committed to using for ranking. A special schema block, because there is no separate AI version of markup. And the idea that ranking first wins the answer, because AI answers pull from a wider set of sources than the top blue links.
The proof, from one store
Straight from the case study, with the source and what it does not prove. See all five results.
Result 3 · GoogleSales from Google: $285 to $8,668 in 30 days
Mar 20 to Apr 18, 2026$285
→Aug 17 to Sep 15, 2026$8,668
30x- Baseline: the worst case before the work
- $285 from 4 orders that came in from Google, March 20 to April 18, 2026. The highest-value products in the catalog had never been published to Google.
- What we changed
- Rewrote product names, titles, descriptions and search snippets around the terms buyers search, with real weights (August 7 to 30). Built six city landing pages (August 1 and 2). Published the beef share products to the Google channel so they could show in free Shopping listings (August 24). Cleaned up crawl errors and expanded thin product pages to more than 400 words (August 27 to 30).
- Result: the best case after
- $8,668 from 16 orders that came in from Google, August 17 to September 15, 2026, 30 times the baseline.
- Evidence source
- Shopify Analytics, Total sales by referrer, Google row, pulled September 16, 2026, beef shares at full price.
- What this does not prove
- Shopify does not split Google search from free Shopping listings, so this cannot say which sold more. Four of the 16 orders were beef shares, $6,120 at full price. Search Console only began collecting on July 28, 2026, so there is no clicks comparison from before the work.
- Test it on your store
- In Shopify, open the Google and YouTube channel and check that your best sellers are published to it. In Search Console, open Performance, sort queries by impressions, and note the ones in positions 5 to 15. Then compare the Google row in Total sales by referrer for your last 30 days with your slowest 30.
If this sounds like your store
This is what AI search for online stores fixes. See it on one store, with the before and after, in
the case study.