How to Audit a Product Page for AI Search

What an audit covers

An audit looks at the product facts a page states, not at the wording. It records what is present, what a buyer in this category would expect and cannot find, what contradicts itself, and what is claimed without support.

Step 1 — Fix the category first

The expected field list depends on the category. Footwear needs sizing systems, width and upper material; a cosmetic needs volume, ingredients and shelf life. A gender segment such as "women" is not a category on its own. If the category is wrong, every later finding is wrong with it — so pick the most specific category the page actually supports, and a broader family when the evidence is thin.

Step 2 — Check identifiers

• Is there a GTIN, EAN or barcode, and is it well-formed?
• Is the SKU stable and unique within your own store?
• Is the brand stated explicitly rather than implied by the logo?
• Is the model or MPN present where the category expects one?

Record an invalid identifier as invalid. Do not repair it and do not substitute a similar one — a silently corrected identifier merges products that are not the same.

Step 3 — Required and recommended fields

Split expected fields into two lists: the ones without which the product cannot be understood at all (title, brand, category, price, currency, availability, primary material or key spec), and the ones that decide whether it gets recommended (sizing, compatibility, care, dimensions, certifications, what is in the box).

Step 4 — Look for contradictions

Compare the title, the description, the spec table, the images and any structured data. A capacity in the title that disagrees with the spec table is a harder failure than a missing capacity, because it makes every other statement suspect.

Step 5 — Separate claims from facts

"Waterproof", "hypoallergenic" and "lasts all day" are claims. If nothing on the page supports them, mark them unverifiable rather than deleting or defending them, and decide whether you can supply the evidence.

Step 6 — Check machine legibility

• Product structured data present and matching the visible page.
• Units, currency and country stated explicitly, not inferred from language.
• Images with meaningful alt text.
• Specs in a readable structure rather than buried in a paragraph.

What an automated audit adds

Running this by hand takes an hour per page. LinqCheck performs the same reading automatically, applies the category's expected field set, and reports observed, missing, conflicting and unverifiable data — then lets you supply the answers it could not find. It does not change your store; you decide what to apply.