What this audit is designed to answer
The first audit should answer two separate questions: can an independent system retrieve and interpret the public facts of the store, and does the store appear in a small, controlled set of relevant buyer questions? The first is storefront readiness. The second is recorded visibility.
Do not begin by adding generic “AI optimized” copy. Establish the public baseline first.
1. Select one representative product
Choose a product that matters commercially and has a stable canonical URL. Record its product name, selected variant, price, availability, SKU or other identifier, brand, image, shipping constraints, and the primary use case a buyer would mention.
Open the page in a private browser window. Confirm that the product information is visible without logging in and without relying on a hidden interaction.
2. Check the public response and canonical destination
Request the product URL and follow its redirects:
curl -sSIL https://example.com/products/example-product
The final response should normally be 200, use the expected public hostname, and resolve to the URL shoppers should share. Inspect the page source and verify that its canonical URL points to the same destination.
3. Compare visible product facts with JSON-LD
Locate every script[type=\"application/ld+json\"] block. The existence of Product JSON-LD is not enough. Compare the product name, price, currency, availability, brand, identifiers, image, URL, and selected variant with the visible page.
Continue with the Product JSON-LD inspection guide.
4. Check crawler and agent-facing routes
Request the following public paths:
/robots.txt
/sitemap.xml
/agents.md
/llms.txt
Confirm their status, content type, final URL, and referenced pages. A generated file is not useful if it links to redirected, private, missing, or contradictory content.
Continue with the AI crawler access guide.
5. Define five buyer questions
Write questions that include a product category, a use case, and a meaningful constraint. Avoid prompts that only ask whether the model “knows” the brand.
Useful groups include discovery, comparison, best-for, compatibility, and purchase intent. Keep the wording stable so future runs are comparable.
6. Save the baseline
For every check, save the URL, timestamp, command or prompt, output, platform context, and visible sources. If the store appears, distinguish a verified storefront domain from a brand-only mention.
7. Make one repair and verify it
Fix the clearest inconsistency first. Confirm the public output after publishing. Then rerun the same check and buyer-question set. Treat any change as an observation, not proof that one field controls an independent recommendation.