- AI search engines retrieve pages and product data first, then write an answer from what they retrieved. If your store can't be crawled, isn't in the index those engines draw from, or has no clear product data, it can't be cited.
- Being cited depends on four things you can influence: access (crawlers can reach and read your pages), answerability (pages directly answer the questions shoppers ask), product data (accurate, structured prices, availability and reviews) and corroboration (other sites describe you the same way).
- Start with an audit of the prompts that matter to your category, fix access and product data, then build the content and third-party mentions that make you the obvious source.
How do AI search engines build a shopping answer?
When a shopper asks ChatGPT “what’s the best running shoe for flat feet under $150?” or sees an AI Overview on Google, the system isn’t recalling a fixed list. In most cases it does two things:
- Retrieve. It runs one or more searches against an index, and for shopping queries it also pulls product data such as price, availability and reviews.
- Generate. A language model reads what came back and writes an answer, citing or linking some of the sources it used.
That means the old SEO question, “can search engines find and understand my pages?”, still sits underneath everything. Generative engine optimization (GEO) adds a second question: when a model reads your page next to ten others, does yours give it the clearest, most trustworthy answer to quote?
Where does each AI engine get its information?
The engines don’t share one index. What you control differs for each.
| Engine | Where it retrieves from | What you control |
|---|---|---|
| Google AI Overviews and AI Mode | Google’s search index, crawled by Googlebot; Google’s shopping data, fed by Google Merchant Center among other sources | Normal indexing and snippet eligibility, page content, structured data, your Merchant Center product feed |
| ChatGPT search and shopping | Web results crawled by OpenAI’s search crawler (OAI-SearchBot) and third-party search providers; product metadata from first- and third-party providers | Crawler access in robots.txt, page content, product data (Shopify Catalog integration or a direct product feed) |
| Microsoft Copilot | Bing’s index | Bing indexing (Bing Webmaster Tools), page content, structured data |
| Perplexity | Its own crawling (PerplexityBot) plus search results | Crawler access in robots.txt, page content |
A few platform facts worth knowing:
- Google says there are no special requirements for appearing in AI Overviews beyond normal search eligibility: the page must be indexed and eligible to show a snippet. Blocking Google-Extended in robots.txt controls some uses of your content for Gemini, but it doesn’t remove your pages from Google Search or its AI features.
- OpenAI states that ChatGPT’s product results are selected independently and are not ads. When choosing which products to show, it considers structured metadata from first-party and third-party providers (such as price and product description) and other third-party content. Merchants are ranked by factors such as availability, price, quality and whether they are the maker or primary seller. For Shopify merchants, OpenAI says product data is already integrated through Shopify Catalog; other merchants can apply to provide a direct product feed.
- Robots.txt matters. OpenAI separates its search crawler (OAI-SearchBot) from its training crawler (GPTBot). You can block training while allowing search. Blocking OAI-SearchBot or PerplexityBot by accident, often through a security tool or a copied robots.txt, removes you from those engines’ search answers.
What makes an AI engine cite one store over another?
No engine publishes a ranking formula, so treat the following as principles drawn from how retrieval and generation work, not as a guaranteed checklist.
1. Access: can the engine read the page at all?
- The crawler is allowed in robots.txt and not blocked by a firewall or bot-protection rule.
- Key content is in the HTML the server returns, not only rendered later by JavaScript. Many AI crawlers don’t execute JavaScript the way Googlebot does.
- Important pages are indexed in Google and Bing, linked internally and listed in your XML sitemap.
- Faceted navigation and parameter URLs don’t swallow crawl activity on large catalogs, a crawl budget issue.
2. Answerability: does the page answer the question directly?
Models pull passages, not whole pages. A product page that says “elevate your everyday with our premium essentials” gives a model nothing to quote. One that says “Cushioned midsole with a medial post for overpronation; 10 mm drop; true to size; 30-day trial” does.
Pages that tend to answer shopping prompts well:
- Product pages with specific attributes: materials, dimensions, compatibility, who it’s for and who it isn’t for.
- Category buying guides that answer “which one should I choose?” for your category.
- Comparison content: your product lines against each other, and honest comparisons with common alternatives.
- Policy pages for shipping times, returns and warranties, written in plain language.
- FAQ sections built from real customer questions in support tickets and reviews.
3. Product data: is it structured, complete and consistent?
For shopping answers, structured product data often matters as much as the page text. That means:
- Structured data on product pages using schema.org Product and Offer markup, with price, currency, availability, brand, GTIN where you have one, and review ratings where they’re genuine. Our guide to product schema for eCommerce covers the implementation.
- A clean product feed in Google Merchant Center, with titles and attributes that match the page.
- Consistency. If the page says $89, the schema says $79 and the feed says out of stock, every system that compares them trusts you less.
4. Corroboration: do other sources say the same thing about you?
AI answers often lean on third-party content: reviews, editorial roundups, forum discussions, comparison sites and video reviews. A brand that’s only described on its own site is harder for a model to recommend with confidence than one that independent sources describe consistently.
This is where entity clarity helps. Use the same brand name, product names and core claims everywhere: your site, marketplaces, social profiles, retailer listings and press. Earn coverage in the publications and communities your customers already read.
What does an AI visibility audit look like? A worked example
Here’s an illustrative process for a mid-size outdoor gear store. All numbers are for illustration.
- List 30 prompts that matter. Mix category (“best lightweight tent for backpacking”), problem (“tent that packs small for bike touring”), comparison (“Brand X vs Brand Y tents”) and brand (“is [your brand] good?”) prompts.
- Run them in ChatGPT, Google (AI Overviews and AI Mode where available), Copilot and Perplexity. Record whether your brand is mentioned, whether your site is cited or linked, and which sources are cited instead.
- Group the results.
| Finding (illustrative) | Count | Likely cause | First fix |
|---|---|---|---|
| Brand mentioned and site cited | 4 of 30 | Strong product pages on niche prompts | Protect and expand these pages |
| Brand mentioned, not cited | 6 of 30 | Engine relies on review sites and forums | Improve product pages; pursue reviews and roundups |
| Brand absent, competitors cited | 14 of 30 | Competitors have buying guides; store has none | Build category buying guides |
| Brand absent from one engine entirely | 1 engine | OAI-SearchBot blocked by a bot-protection rule | Allow the crawler and recheck |
| Wrong price shown | 3 of 30 | Schema price out of date after a sale | Fix the schema source |
- Prioritize by revenue. Fix access and data errors first, since they block everything else. Then build content for the prompts closest to purchase.
- Re-run the same prompts monthly. Answers vary between runs and users, so track trends across many prompts rather than reacting to a single answer.
How do you measure AI search traffic?
Set expectations: you won’t get a clean “AI rankings” report from the engines themselves.
- Referrals in GA4. Build a channel group or exploration for referral sources such as chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com. Track sessions, revenue and conversion rate, not just visits.
- Search Console. Clicks from Google’s AI features are included in your overall web search data rather than shown as a separate line, so watch query and page trends instead of looking for an AI-only report.
- Prompt tracking. Manual prompt audits or a dedicated AI visibility tool show mentions and citations over time, including when people read the answer and never click.
- Branded search. If AI answers mention you without linking, some shoppers search your name afterward. Rising branded search can be a sign.
What should you fix first?
- [ ] Robots.txt allows Googlebot, Bingbot, OAI-SearchBot and PerplexityBot; firewall and bot-protection rules don’t block them.
- [ ] Core product and category content is present in the server-rendered HTML.
- [ ] Top products and collections are indexed in Google and Bing.
- [ ] Product schema includes price, currency, availability, brand and genuine review data, and matches the page.
- [ ] Merchant Center feed is approved, accurate and consistent with the site.
- [ ] Product pages state specific attributes, use cases and fit, not just brand copy.
- [ ] Buying guides exist for your top categories and answer “which one should I choose?”
- [ ] Shipping, returns and warranty policies are clear and easy to find.
- [ ] Brand and product names are consistent across your site, marketplaces and profiles.
- [ ] A monthly prompt audit and a GA4 AI-referral report are in place.
Most of this is good technical SEO and good merchandising. The difference with AI search is that weak spots are less forgiving: an engine picks a handful of sources per answer, and if yours isn’t among them, there’s no page two.
FAQ
Is generative engine optimization different from SEO?
It builds on SEO rather than replacing it. Crawling, indexing and structured data still decide whether you can be retrieved. GEO adds a focus on pages that answer specific questions clearly, on consistent product data, and on third-party sources that corroborate what you say about yourself.
Can I pay to appear in ChatGPT’s product results?
OpenAI states that ChatGPT’s product results are selected independently and are not ads, and that ads are shown separately. To improve your chances of appearing, focus on accurate product data, crawler access and strong, specific product content.
Should I block AI crawlers from my store?
Blocking training crawlers is a business choice, but blocking search crawlers such as OAI-SearchBot or PerplexityBot removes you from those engines’ search answers. Most stores that want to sell through AI search allow the search crawlers and decide separately about training.