Ten myths
Mentioned means recommended
No. In our evidence, one site was mentioned 26 times across 30 answers while its official site was cited only 12 times. Mention, recommendation and citation are three different things; mixing them inflates your self-image.
A refusal means I am excluded
No. Refusal, non-mention and non-recommendation are three different observations. Some categories draw batch refusals (adult products, for instance), which only describes sample behaviour — not proof that users cannot see you.
A high static score means AI will recommend me
It will not. Static rule checks and real AI answers are two classes of evidence: the score says whether signals are complete, not what a model answers or why.
Schema or llms.txt guarantees inclusion
No. They help machines parse structure and sources, but guarantee nothing about crawling, inclusion, citation or recommendation. llms.txt is a community proposal with no verified evidence of mainstream AI reading it.
A service promising No.1 in AI is worth buying
Nobody controls model answers. For any service promising fixed ranks or guaranteed uplift, ask for traceable raw answer evidence first; without it, the promise is empty.
One test is the final verdict
No. Model, time, region and phrasing all affect answers; a conclusion holds only for that sample. Re-test under identical conditions after fixes; improvements you did not measure cannot be claimed.
Famous in the industry means AI recalls me
Not necessarily. In our evidence, a well-known site was mentioned only 4 times in 30 category answers. Industry fame does not carry into category questions where buyers never name a brand.
If the page opens, machines understand it
Different things. Certification numbers trapped in images, key facts crammed into one wall of text, or markup that contradicts visible content can all be unreadable or incoherent to machines.
Whoever AI cites must be right
Cited sources may be retail platforms, competitors or media — not the official voice. If your site is not cited, the discourse power sits elsewhere. That is exactly why entity verification matters.
A 0% result means it is over
0% only means: not mentioned in this sample — this model, these questions, this time. It does not imply all users cannot see you, nor the future. Widen the sample and break it into layers before judging.
Further reading: the same material is also published in our Zhihu knowledge base (self-published, not independent evidence). Open the knowledge base
Free tools, no sign-in needed