Can AI actually read your store? Five checks you can run in fifteen minutes

SEO
, , , ,
1

More buyers are asking ChatGPT, Perplexity, and Google’s AI for product recommendations before they ever land on a store. Those tools do not browse the way people do. They read the raw page and the structured data underneath it, then decide if your store is clear enough to name.

Plenty of beautiful Shopify stores are nearly invisible to them, and the owner has no idea, because nothing in your analytics shows it.

Here are five checks you can run on your own store in about 15 minutes. No apps, no cost. Count how many you pass. Most stores pass one or two.

  1. The Front Door. Can a machine read the page at all.
    Do this: open one product page, right click, choose View Page Source, not Inspect. Press Ctrl F and search for your product title and price.
    Pass: the title, price, and description are right there in the source.
    Fail: they are missing because they only load after JavaScript. Many AI crawlers do not run JavaScript, so they see a blank page.
    Why it matters: this is the most basic test, and it quietly knocks out the most stores. Nothing else counts if the page cannot be read.
  2. Entity Integrity. Are your products clean, machine readable facts.
    Do this: paste a product URL into Google’s Rich Results Test (free, just search that name).
    Pass: it detects Product data with price, availability, and a rating, and Organization data with real social links.
    Fail: missing price or availability, empty social links, or no Product data at all.
    Why it matters: AI cites facts, not prose. With no clean facts, it quotes a competitor it can read instead. This is where most stores fail hardest.
  3. AI Readability. Is the answer stated, or buried.
    Do this: read your top product or collection copy and try to pull one fact in a single pass, what it is, who it is for, why it is different.
    Pass: the key answer is stated plainly near the top.
    Fail: the answer is buried three paragraphs into marketing copy.
    Why it matters: AI extracts the plain statement. If it has to dig, it moves on to a store that made it easy.
  4. The Mesh. Do your pages connect into one map.
    Do this: start on a product page and try to reach its collection, then a related product, using only on page links.
    Pass: you can move from product, to collection, to related product easily.
    Fail: your pages are islands with no clear connections.
    Why it matters: those links are the map AI reads to understand that your brand sells X for use case Y.
  5. Authority. Is your identity consistent and current.
    Do this: check your brand name, contact details, and key info across your pages and footer.
    Pass: same name, same details, current content, everywhere.
    Fail: inconsistent or stale identity.
    Why it matters: a model needs a consistent identity before it will confidently name you.

Your score.
5 of 5 is rare. You are genuinely legible to AI.
3 to 4, you are most of the way, fix the gaps.
0 to 2 is where most stores sit today.

If you only fix two things, do these in order. First, open the Front Door so crawlers can read the page. Second, complete your Product and Organization data so AI has clean facts to cite. Those two carry most of the results.

The pattern I keep seeing is simple. Stores are strong on design and weak on all five. The fix is rarely to write more content. It makes your products machine-readable and confirms the front door is open.

Drop your store URL and the check you are stuck on, and I will tell you what AI sees.

2

Solid list. The front door and entity integrity checks really do carry most of it, so I’d put those first too.

One thing worth adding that’s specific to Shopify. A store can pass the Rich Results test and still get ignored, because a lot of themes inject Product schema and then an SEO app injects its own on top. You end up with two conflicting Product entities on the same page, which reads as less trustworthy than having none. Worth checking you only have one clean block, not two fighting each other.

There’s also a step before any of the five even matter: check your robots.txt isn’t blocking the AI crawlers. GPTBot, PerplexityBot and ClaudeBot are separate user agents, and some stores or apps quietly disallow them. If they’re blocked you can be 5 of 5 and still invisible.

And readable only gets you eligible, not cited. AI tends to name stores it can corroborate off your own domain too, so consistent details on third-party pages matter as much as the footer.

Curious how often you see the duplicate-schema one, since on Shopify it feels almost default once someone installs an SEO app.

3

This is a great addition, Lumine, and you are right on all three. Let me take them in order, because each one slots into the five rather than sitting outside them.

The duplicate schema one is the most common entity problem I see on Shopify, and you named exactly why. A theme outputs Product schema, then an SEO app outputs its own, and now there are two Product entities on one page with different prices or availability. To a model that is worse than empty, because it is not missing data, it is contradictory data, and contradiction reads as untrustworthy. It really is almost default once an SEO app goes on. Quick way to catch it, open View Page Source, Ctrl F for Product, and count the JSON-LD blocks. If you see two, that is the fight. It lives inside Entity Integrity, it is just the failure mode nobody looks for, because each block on its own looks fine. Entity Integrity was the single weakest layer when we measured 216 stores, 72 percent failed it, and conflicting or incomplete product data was a big part of why.

The robots.txt point belongs right at the Front Door, even before the JavaScript check. GPTBot, PerplexityBot, and ClaudeBot are separate user agents, and on Shopify they get blocked two ways, a disallow added to robots.txt.liquid, or a bot protection app that stops them at the edge before robots.txt even matters. You can be 5 of 5 and invisible, exactly as you said. It deserves to be step zero. For scale, the Front Door was one of the layers most stores actually passed, and 41 percent still failed it, usually JavaScript only rendering, with the blocked crawler cases hiding inside that.

And the corroboration point is the honest ceiling on all of this. Readable gets you eligible, cited is a separate bar. That is the Authority layer doing its work, a model is more confident naming a store whose identity it can confirm off the domain too, in reviews, directories, and mentions. Consistent details everywhere, not just the footer.

On how often I see the duplicate one, it is one of the most frequent entity failures on stores running an SEO app, and the frustrating part is the owner did everything right, they added schema, they just added it twice. Are you seeing any theme and app combinations that handle it cleanly, or is your fix usually to turn one source off.

4

Hi @margareta.petrovic .

Hope you are doing well.
Really good post. Practical, honest, and easy to follow.

The one thing worth adding is to tell people to simply search their own store name in ChatGPT or Perplexity first. If nothing comes up or the info is wrong, that tells them everything instantly without any technical tests.

The scoring system at the end is a great touch. People love knowing where they stand.

Good work, this is genuinely useful advice.

6

You can also go into Claude or ChatGPT and prompt something like ‘Can you access the content at your store link. I want to know if you can access the page content’

If not, it’ll likely also tell you why and how to fix it.

7

That can definitely be one of the checks. You just have to force AI to perform checks and be consistent in evaluations. It tends to be overly optimistic and skip steps.

8

this is becoming way more important than most merchants realize. the stores that show up in chatgpt shopping recommendations and google AI overviews aren’t necessarily the ones with the best backlinks, they’re the ones with clear, descriptive product information that AI can actually parse.

biggest quick wins: make sure every product has a genuine description (not just specs), use natural language in your titles (how a real person would describe the item), and have clear category/collection structure. AI models pull from structured content, so the more organized and descriptive your product data is, the more likely you show up when someone asks an AI for recommendations.