LLM visibility myths: schema, llms.txt and mentions
Someone adds every schema tag, waits, and sees no change. That is the typical story, and it fits what Google actually says. Here is each popular idea against the record.
AlgoLink research team ·
Seven ideas, one line each
Myth
Reality as far as we can tell
Adding full schema makes models cite you
Markup helps parsers read a page. No test shows it changes citations
llms.txt is the new robots.txt
A September 2024 proposal; major vendors have not confirmed they read it
AI search has killed backlinks
Link data still describes a site's web footprint; its role in citations is untested
Reddit, Wikipedia and YouTube are the only sources
They are cited often, but not exclusively, and shares move over time
You need a special agency
The checks are cheap and mostly overlap with ordinary SEO
Blocking GPTBot hides you from every assistant
It affects one crawler. Other products use other crawlers or other indexes
Unlinked mentions are worthless
Plausibly useful, since models read text. Untested at scale
The schema question, slowly
Schema.org markup describes a page in a machine-readable way. Simple. Google has said that structured data is not required for its AI features, though it still helps with rich results, which is its own good reason to add it.
Does it feed models directly? No one outside the vendors can say. We would test it the dull way: mark up half of a set of similar pages, leave half bare, and log citations for eight weeks.
Practitioners report mixed experiences, including plenty of silence after full markup. Silence is data. Do not bury it under another round of tags.
Author bios and 'expert' credentials get the same treatment. They help readers decide whether to trust you, which is reason enough, and no study we have seen shows a model preferring a page because a bio was present.
What about freshness? Pages with a recent update date do appear often in AI answers about news-like topics, and Google has long used freshness for queries that deserve it. Still, bumping a date without changing content is a habit we would not rely on, and Google has said it does not like it.
Mentions: the cheaper bet
If a model reads text, a clear description of your brand on other sites may matter even without a link. Finding such pages is ordinary outreach work.
Start simple. Search your brand name, drop sites that already link, and list the rest. Then ask for a link or a plainer description. Tools like Ahrefs and Semrush have mention alerts, and our link-profile pull shows who links to the competitors you envy.
A final caution about agencies. Selling LLM visibility audits is a new line of business, and some reports are thin. Ask what was sampled, when, and what would have changed their conclusion.
None of this makes markup pointless. It says the claimed effect is unmeasured. Spend an hour, not a quarter, and keep your main effort on pages people actually want to read.
What we would do first
Fix indexing, then write passages that stand alone
Check that Google and Bing can crawl and index the page. Then open sections with a direct answer, name the entities, and earn description from third parties. Everything else is an experiment, so run it as one.
Confirm indexing in Search Console and Bing Webmaster Tools