You can check whether AI can reach, read and recommend your store in about twenty minutes, with a browser and one assistant account. Run the seven checks below in order, since a failure at check 1 or 2 hides everything the later page checks look for. Each check states what a passing result looks like.
Nothing here needs a tool, a login to your store or a developer. Keep a note open and write down each result as pass or fail, with the page you tested.
Check 1: can AI crawlers reach your site?
Type your domain followed by /robots.txt into the browser, for example yourstore.com/robots.txt, and read the file from top to bottom.
- Look for a line that begins with User-agent and names GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Bingbot.
- Under each of those, look for a Disallow line. "Disallow: /" blocks the whole site for that crawler.
- Look at the group under "User-agent: *", which applies to every crawler not named elsewhere.
You pass when no rule blocks those crawlers from your products, collections or blog. You fail when any of them is named with "Disallow: /", or when "User-agent: *" carries "Disallow: /". Google-Extended is a separate token that controls Gemini training and grounding, and Google says it does not affect inclusion in Google Search. Our robots.txt guide explains each line.
Check 2: does your text exist without JavaScript?
Open a product page, right click and choose View page source. That view shows the HTML as the server sends it, before scripts run. Press Ctrl+F, or Cmd+F on a Mac, and paste in a sentence you can see on the page.
You pass when the browser finds the sentence. You fail when it does not, which means scripts draw the text after the page loads and some crawlers will never see it. The rendering guide explains what to change.
Check 3: is the page indexed?
Copy one distinctive full sentence from a product page and search for it on Google inside double quotation marks.
You pass when your page comes back. You fail when nothing comes back, which usually means Google has not indexed the page. Search Console's URL Inspection tool tells you why.
Check 4: can a machine read your product data?
Go back to the page source from check 2 and search for "application/ld+json". Read each block the search finds.
You pass when one block has "@type": "Product" with a name and an offers section. The offers section must carry price, priceCurrency and availability that match the page. You fail when there is no Product block or there are two with different prices. You also fail when the markup price differs from the price a shopper sees.
Check 5: does an assistant know who you are?
Open ChatGPT or Perplexity and ask "What is [your brand name], and what do they sell?" Use the brand name exactly as your site writes it.
You pass when the answer describes your business in a way you recognize. You fail when it describes somebody else with confidence, or says it does not know. Either result usually means your identity is not stated clearly where a machine can check it.
Check 6: would an assistant recommend you?
Ask the buying question a customer would ask, without naming your brand, for example "Where can I buy [your category] online?"
You pass when your store is named. You fail when it is not. Write down the stores that are named, since those stores are the competitors an assistant weighs you against.
Check 7: are your price and stock current in the markup?
Pick a product whose price changed recently. Compare the price and availability in its markup from check 4 with what a shopper sees on the page today.
You pass when they match. You fail when they differ. That is worse than having no markup, since an assistant can then send a buyer to a price that no longer exists.
What does a ceramics studio find when it runs the 20 minute audit on a Monday?
A ceramics studio on Shopify with 35 products runs the checks on a Monday morning. It passes checks 1 to 3, since the robots file is the Shopify default, the product text sits in the page source and Google returns its product sentence. At check 4 it finds two Product blocks on every page, one from the theme and one from a reviews app. At check 5 ChatGPT describes a different studio with a similar name in another country.
That studio has two jobs and a clear order. The owner removes the duplicate block first. Then a plain sentence goes on the homepage that names the studio, what it makes and the city it ships from.
What do I do with the results?
Fix the failures in the order of the checks, starting with any failure at check 1 or 2. A crawler cannot read markup on a page it cannot reach.
- On Shopify, robots.txt is edited in Online Store > Themes, under Edit code, in the robots.txt.liquid template. If the file carries rules nobody on your team remembers writing, look for a robots.txt.liquid template that a developer or an app added.
- Product markup comes from the theme and from apps. Check each app's settings for a structured data option before you edit theme code.
- Product descriptions are edited in Products, in the Description field of each product.
- Repeat all seven checks every month, and always after a theme change, an app install or a migration.
When you want to read your score in layers and pick a first fix, what your AI visibility score means explains the scorecard. The 90 day plan turns the fixes into a schedule. The free scan runs the same logic on five of your key pages and ranks what it finds.
Common questions
Do I need any tools for this audit?
A browser and one assistant account are enough. Every check uses View page source, a Google search or a question typed into ChatGPT or Perplexity.
How often should I repeat the audit?
Repeat it monthly, and always after a theme change, an app install or a migration. Those are the moments when crawler access and product markup most often break.
What if everything passes but I am still not recommended?
If you pass all seven checks, readability is not what holds you back. Readability only gets a store into the comparison. Mentions on other trusted websites, reviews and a clear identity are what an assistant uses to choose between eligible stores.
How is this different from a full scan?
This audit tests a few pages by hand. A full scan reads the structured data and text of many pages, scores five layers and ranks every finding with its fix.
Sources
- OpenAI, Overview of OpenAI crawlers, read November 9, 2026.
- Perplexity, Perplexity crawlers, read November 9, 2026.
- Anthropic, Does Anthropic crawl data from the web, read November 9, 2026.
- Google, Google's common crawlers, on Google-Extended, read November 9, 2026.