How LinqCheck Assesses Product Data
This page explains what LinqCheck reads, how it classifies what it finds, and where it deliberately stops. It is written so you can judge a result rather than take it on trust.
What is read
LinqCheck reads the public product page at the URL you provide: its visible text, its structured product data where present, its specification blocks and its images. It works from what the page states. It does not have access to your admin, your inventory system or your supplier data.
Category first
Before anything is judged, the product is placed in a category from a language-independent taxonomy. The category determines which fields are expected. Where the evidence is weak, a broader family is used, and where it is absent the product stays unclassified — a broad category is better than a confident wrong one. You can correct the category, and the expected field set follows your correction.
The four states
Observed
The page states the value and it was read as written. The raw wording is preserved.
Missing
The field is expected for this category and the page does not state it. Missing means "not found on this page", not "does not exist".
Conflicting
The same field is stated more than once with different values. Both values are reported. LinqCheck does not decide which is correct.
Unverifiable
A claim is made but nothing on the page supports it. This is a description of the evidence, not an accusation that the claim is false.
Where "verified" applies
A fact counts as confirmed only when you confirm it — for example by answering a missing field or accepting a value during review. Content produced by AI is a suggestion, never a verified fact, and reading a value from a page makes it observed, not verified.
What LinqCheck will not do
- Invent a product fact, measurement, material or compatibility statement.
- Invent or repair an identifier — an invalid GTIN is reported as invalid.
- Merge two products because their identifiers look similar.
- Resolve a conflict on your behalf.
- Change anything in your store, or publish anywhere.
How generated content is constrained
Content is generated only from facts the page supports or you supplied. Where a fact is unknown, the content leaves it out rather than filling it in, and unsupported claims are rejected. Content language is a separate choice from your interface language.
What a result does not mean
A readiness result describes the data on one page at one moment. It is not a prediction of traffic, ranking or placement in any AI assistant, and a clean result is not a guarantee that a third-party system will use your product.
Try it on one product page
See the observed, missing, conflicting and unverifiable breakdown for yourself.