Before you sign up for any AI visibility tool, run this audit. It takes about an hour, costs nothing, and gives you a fairly accurate read on whether AI search engines can see your store. Most merchants who do it find at least one issue they didn’t know about — usually something that’s been quietly costing them visibility for months.

This isn’t a substitute for ongoing monitoring. AI search results shift constantly, and one-time checks won’t catch regressions. But it’s the right starting point. If you can’t pass these basic tests, paying for tracking software just gives you a slightly better-instrumented view of your store being invisible.

The five tests

Each test answers a specific question, and each one has a fix when it fails. You can run them in any order, but I’d suggest doing them top to bottom — the earlier tests catch issues that affect everything else.

1. Ask the AI engines directly

This is the bluntest test, and also the most informative. If a real shopper asks an AI engine for recommendations in your category, do you come up?

Open fresh chat sessions (sign out of your account if you can — personalized history can skew results) in:

ChatGPT (chatgpt.com)
Perplexity (perplexity.ai)
Google AI Overviews (run a search on google.com — the AI Overview appears at the top for shopping-intent queries)
Gemini (gemini.google.com)
Claude (claude.ai)

Now run a few prompts a shopper might genuinely use to find a product like yours. The trick is to phrase them like a person, not like a search query. Compare:

Bad test prompt: “running shoes brand”
Good test prompt: “I’m looking for trail running shoes for someone with wide feet, under $150, and I prefer brands that don’t run small. What would you recommend?”

Run five to ten prompts per engine. Write down:

Whether your store or product is mentioned at all.
Position in the response — first product cited, fifth, buried in a footnote.
Whether the AI links to your actual product page or just names the product without a link.
Who else shows up. This is your real competitor list for AI search, which is often different from your competitors in Google search.

What “good” looks like: you appear in answers for at least your three or four most obvious category prompts, and you’re cited by at least two of the major engines.

When this fails: the most common cause is that your products are technically findable but lack enough structured signal for the AI to confidently recommend them. The AI doesn’t trust thin descriptions or missing identifiers. Move on to the next tests to figure out which signal is weak.

2. Look at your raw HTML source

Right-click on a product page and choose “View Page Source” (not “Inspect Element” — that shows the rendered DOM, which is different from what crawlers see). What you’re looking at is the HTML that AI crawlers see when they fetch your page.

Press Ctrl+F (Cmd+F on Mac) and search for these strings inside the source:

application/ld+json — this is the structured data block. You should find at least one. On a product page, it should contain "@type": "Product".
"price" — should appear inside the JSON-LD block with the actual price.
"sku" and "gtin" — these are the product identifiers AI engines use to match your product to the real-world entity. Missing GTINs are the single most common gap in Shopify catalogs.
Your product description — search for the first sentence of your description. Is it actually in the HTML, or is it being injected later by JavaScript?

What “good” looks like: you find a complete application/ld+json block with name, description, price, availability, brand, image, sku, gtin (or mpn), and ideally aggregateRating. Your description text is in the raw HTML, not loaded after the fact.

When this fails:

Missing JSON-LD entirely. Your theme isn’t outputting Product schema. This is the highest-priority fix.
JSON-LD only has name and price. Common on Shopify themes — they include minimal schema. You need to add brand, GTIN, MPN, condition, AggregateRating, and images. A theme app embed (delivered through a Shopify app) is the cleanest way to fix this without editing theme code.
Description not in raw HTML. Some apps inject descriptions or specs client-side via JavaScript. Crawlers that don’t render JS won’t see this content. Fix this by rendering server-side (in Liquid for Shopify) or using a theme app embed that runs at the right point in the load order.
Missing GTIN/MPN. Add these in your product metafields. Perplexity’s merchant program specifically requires GTINs. Without them, AI engines have a harder time matching your product to real-world product entities and reviews.

3. Read your robots.txt

In your browser, navigate to yourstore.com/robots.txt. This file tells crawlers what they’re allowed to access.

Search the file for these user agents:

OAI-SearchBot (ChatGPT Shopping)
PerplexityBot (Perplexity)
Google-Extended (Google AI Overviews and Gemini)
ClaudeBot (Claude)
Applebot-Extended (Apple Intelligence)

If any of these have a Disallow: / rule, that AI engine cannot index your store at all. This sometimes happens unintentionally — some SEO apps add aggressive bot blocks, or developers add them once and forget about them.

Also look for:

A blanket User-agent: * followed by Disallow: /. This blocks every crawler. Sometimes it’s leftover from when the store was in development and was never removed for launch.
Custom rules in robots.txt.liquid (your theme’s robots template) that override Shopify’s defaults.

What “good” looks like: AI search bots are not blocked. The only Disallow lines should be Shopify’s defaults blocking /admin, /cart, /checkout, /search, /policies/, and filtered collection URLs.

When this fails: find the source of the block. If it’s in robots.txt.liquid, edit the file. If you can’t find it in your theme but it’s appearing in robots.txt anyway, check if a Shopify app is overriding it. If you’re using Cloudflare or another CDN in front of your store, also check whether AI bots are being blocked at the network level — Cloudflare has been blocking AI bots by default for new sites since mid-2025.

4. Validate your structured data

This is the most reliable way to confirm that the structured data in your HTML is actually parseable by machines. Two free tools, both worth running:

Google’s Rich Results Test — paste your product URL. It tells you what Google sees and whether it’s eligible for rich results. Errors here often correspond to issues AI engines have parsing your data, even if AI engines aren’t Google.
Schema Markup Validator — same idea, tests against the schema.org spec rather than Google’s specific eligibility rules. More forgiving but better at catching schema-level issues.

Run a few different product types — your top-selling product, a product with variants, a product that’s out of stock, a product on sale. They might surface different issues.

What “good” looks like: Both tools report no errors. Warnings are acceptable for fields that don’t apply to your products, but errors mean the structured data is malformed in a way that machines can’t reliably read.

When this fails: the validator tells you exactly which field is invalid. Common ones:

Price formatted as a string instead of a number (“$29.99” instead of 29.99).
Missing image URL or image URL pointing to a 404.
Brand object malformed — should be {"@type": "Brand", "name": "..."} not just a string.
AggregateRating present but no review count — AggregateRating without a reviewCount is invalid.

These tend to come from theme defaults or app conflicts (two apps both injecting JSON-LD that don’t agree on the schema).

5. Check your AI referral traffic

Open Google Analytics 4 (or your preferred analytics tool) and look at acquisition data for the past 90 days. Filter referrals or sources for these domains:

chat.openai.com
chatgpt.com
perplexity.ai
gemini.google.com
copilot.microsoft.com
bing.com/chat

Note the session count from each, the engagement rate, and the conversion rate.

What “good” looks like: you see at least some traffic from these sources, and the engagement metrics are stronger than your site average — AI-referred sessions tend to have lower bounce rates and longer time on site than the typical visitor.

When this fails:

Zero AI referral traffic across all engines. Either AI engines genuinely aren’t recommending you (which the previous tests should help diagnose), or they’re recommending you in a way that doesn’t generate clicks (some engines name products without linking, or users copy product names and search separately). The first scenario is fixable; the second is something to live with for now.
Traffic from one engine but not others. Different engines crawl differently. ChatGPT relies more heavily on its own search index; Perplexity has its own crawler; Google AI Overviews piggybacks on Google’s index. If only one engine sends traffic, you may have a crawl-access issue specific to the others.
Low engagement on AI traffic. If AI-referred visitors bounce quickly, the AI may be sending them for a query your product doesn’t actually fit. This is usually a description or categorization mismatch — your product is being recommended for the wrong queries because your data implies a different fit than reality.

A worked example

Here’s how this looks in practice. I’ll use a hypothetical: a Shopify store selling natural deodorants, mid-size catalog, decent organic traffic, no AI strategy yet.

The merchant runs the audit and finds:

Asking AI directly: They appear once across 30 prompts, and only when the prompt is extremely specific to their brand name. For category prompts (“aluminum-free deodorant for sensitive skin”), they’re missing entirely from all five engines. ChatGPT and Perplexity both repeatedly recommend three competitors.
Raw HTML: Product schema is present but minimal — name, price, image. No brand, no GTIN, no aggregateRating. Description loads via JavaScript and isn’t in raw source.
Robots.txt: Default Shopify, no AI bot blocks. Clean.
Schema validation: No errors, but several warnings about missing recommended fields (brand, GTIN, AggregateRating).
Analytics: Approximately 12 sessions per month from AI referrers, almost entirely from Perplexity. Zero from ChatGPT or Google AI Overviews despite weak coverage in both.

The audit gives them a clear priority list:

First: fix the JavaScript-rendered description (server-render it). This is invisible to roughly half the AI crawlers right now.
Second: add GTINs to the metafields and switch to a theme app embed that injects complete Product schema with brand, identifiers, and aggregate ratings.
Third: rewrite the top 20 product descriptions with the specific factual detail (active ingredients, fragrance-free vs not, ounces, application method) that AI engines extract for category queries.

That’s a plan they can actually execute, prioritized by which fix unblocks the others. They didn’t need to buy any software to find it — they needed an hour and a willingness to look at their own store the way a crawler would.

When the manual audit isn’t enough

The DIY approach hits a ceiling fast:

It scales linearly. Auditing 20 products by hand is fine. Auditing 800 isn’t. Once you have a real catalog, you need automation just to know which products have issues.
It’s a snapshot. AI search ranks shift constantly. The query that you appear for this week might drop you next week. Re-running the audit manually every Monday is not realistic.
The structured data rules evolve. Schema.org adds fields, AI engines change which signals they weight. Rules that pass validation today might be insufficient in six months.
Cross-engine breadth is hard manually. Five engines, ten prompts each, run weekly, with results stored and compared — that’s a job for software.

This is the point where the merchants who care about AI search start using a tool. Not before — there’s no point paying for tracking when you can’t pass test #2 — but once the foundation is in place and the catalog is too big to audit by hand.

Where StoreBeam fits

StoreBeam is built for Shopify specifically. It runs the structured data, content, taxonomy, and crawlability checks from this audit automatically across your entire catalog, scores every product, and gives you a prioritized fix list. On Pro and Business plans, it adds visibility tracking against AI engines so you can see how recommendations shift as you fix things.

The free plan covers up to 25 products and surfaces the same issue list this manual audit would, just done in seconds. If you want to start there before committing to anything, install StoreBeam from the Shopify App Store.

But honestly: if you’ve made it this far in the post, the most useful thing you can do today is run the manual audit on your store. You’ll either find issues you can fix in a weekend, or you’ll confirm your foundation is solid — and either outcome is worth knowing before AI search becomes an even bigger share of how shoppers find products.