Seven things, in the order that matters: a page it can read, exact price and stock in the HTML, product schema with offers, category pages that answer "best X for Y," OAI-SearchBot and PerplexityBot allowed in, reviews on the sites engines cite, and one consistent brand entity. Fix them out of order and the later ones do less work.
This is the catalog-wide version of the question how to get recommended by ChatGPT answers for a single page: the same order a fetcher meets a site in, applied to every product and category page in a store instead of one. Reflexa for online stores walks through the same seven as a product feature list; this is the checklist version.
Only if the title, price and description exist in the HTML your server sends, not only after JavaScript renders the page in a browser. A Shopify theme with a custom app that injects price and stock, or a headless React or Vue storefront, can look complete to a shopper and be an empty shell to a fetcher that never runs scripts.
| Checked by | Free tool |
|---|---|
| Text is present without JavaScript | What AI actually sees |
Do this: view the raw page source (not the browser inspector) and look for your product's price. If it is not there, it is drawn in by JavaScript after the fact.
Yes, in both the visible copy and the Product schema's Offer, because one without the other reads as a guess, not a fact an engine can quote. A price that exists only in JSON-LD and not on the page a shopper actually sees is unconfirmed; a price only in the copy gives an engine nothing structured to extract. The mechanics of a price mismatch, including why a stale JSON-LD Offer is a common cause, are in why does AI get my pricing wrong.
Why it matters: an out-of-stock item quoted as available, or a sale price that expired last month, is the fastest way to lose a buyer's trust the moment they click through.
An Offer with price, currency and availability, and a rating or review count if you have one, because a Product block without an Offer tells an engine what the item is, not whether or for how much it can be bought. The Organization behind the store needs its own stable identity too, linked to every product so an engine can tell one listing belongs to one seller, not a dozen unrelated pages.
| Checked by | Free tool |
|---|---|
| Structured data matches the visible text | Schema & entity check |
| Structured data entities are linked (@id) | Schema & entity check |
Do this: generate the schema from the same source as the visible price and stock figure, in the same edit, so the two cannot drift apart silently.
Yes, because when a buyer asks an engine "best running shoes for flat feet" or "best standing desk for a small room," the answer is assembled from the comparison pages and roundups the engine can retrieve, not from a single brand's product page alone. A category page written to answer that exact comparison, rather than just listing items, has a chance to be one of the pages named.
| Checked by | How to see it |
|---|---|
| Comparison pages in your category mention the brand | Ask ChatGPT, Claude and Perplexity the buyer's exact question and open the sources they cite |
Why it matters: this is a human judgment call, not something a structural check can score alone — it needs the buyer question actually asked and the sources actually read.
They need to be allowed separately, because GPTBot trains future models while OAI-SearchBot and PerplexityBot fetch a page right now to answer a live shopping question. Allowing GPTBot and forgetting the other two, or leaving a blanket Disallow from an old anti-scraping rule in place, is the single most common mistake a store's robots.txt makes.
Proof: OpenAI — crawlers overview · Perplexity — bots documentation.
| Checked by | Free tool |
|---|---|
| robots.txt allows AI crawlers | AI crawler access check |
Yes, because engines are cautious about taking a brand's word for its own value, so they look for reviews on the sites they already trust and cite, not just the stars on your own product page. G2, Capterra and Trustpilot for software, and category-specific review sites for physical goods, are the material an engine can point to when it backs up a recommendation with something other than your own copy. The full mechanics are in do reviews change whether AI recommends you?
| Checked by | How to see it |
|---|---|
| Press and reviews confirm the brand | Search your brand name with "review", "vs" and "alternative" and read what surfaces |
Yes, the same Organization name, URL and sameAs links on every product and category page, so an engine assembling an answer about your store has no doubt it is reading pages from one seller, not several. A storefront that renders its Organization block differently on different templates, or that changed its legal name without updating schema everywhere, gives an engine a reason to merge it with the wrong company, or none.
| Checked by | Free tool |
|---|---|
| JSON-LD identifies the organization | Schema & entity check |
Each of these takes minutes and costs nothing:
Seven things, roughly in order: a page an engine can read without running JavaScript, exact price and stock in the HTML, Product schema with offers, category pages that answer a "best X for Y" question, OAI-SearchBot and PerplexityBot allowed in robots.txt, reviews on sites engines cite, and one consistent brand entity across the catalog.
Not enough on its own. GPTBot trains future models, while OAI-SearchBot and PerplexityBot fetch a page right now to answer a live shopping question. Allowing GPTBot and forgetting the other two is the single most common mistake in a store's robots.txt.
No. Schema tells an engine what an item is and what it costs, but it has to match the visible page, and it says nothing about whether other sources back you up. It is one of seven conditions, not the whole answer.
Because a "best running shoes for flat feet" answer is usually built from the roundups and comparison articles an engine can retrieve, not from a single brand's product page. A strong product page still has to exist, but being named in the roundup is what gets you into the answer.
The free check at dashboard.reflexaagent.com/audit runs the structural checks (readability, schema, crawler access) on your own pages in about three minutes and flags which of the seven you are missing.
The free check reads your top pages the way ChatGPT, Claude and Perplexity read them, and checks your schema against what's actually on the page. 3 minutes, evidence included.
Let the engines in, with one click: the robots.txt switch for AI crawlers · install Reflexa Agent · på svenska