For ten years the homepage was the official record of the company. Search Engine Journal’s write-up of Trendos research says the new answer boxes do not agree.

User-generated pages take a large share of the citations behind AI answers. Brand homepages, in some categories, barely show up. The finding is in Search Engine Journal. The measurement is Trendos. The conclusion this desk will stand on is narrower than either: the homepage is a brochure. Brochures are bad citation targets.

“AI search citation sources are not the brand homepage.”

— Trendos research in Search Engine Journal

Why the homepage loses the citation

A homepage is built for people who already typed the name. It has a slider, a chat bubble, three offers, and a theme that changed twice since the last photographer came. An answer engine is not that person. It is under a time budget. It wants a paragraph it can lift.

A Reddit thread that names your firm and your city is already a paragraph. A Yelp review is a paragraph. A publisher explainer is a paragraph. Your homepage is a room. Models walk into rooms and leave. They stay in paragraphs.

That is why a contractor can spend $20,000 on a site and still watch a stranger’s list get the credit. The model did not hate the brand. It could not extract the brand from the theme.

Machines prefer documents. People prefer pages.

This is not new. Software teams ship XML sitemaps and raw JSON because crawlers prefer files. Newsrooms ship a story URL that does not depend on a carousel. Local firms shipped only the page, then asked why the page never got cited.

Trendos did not tell you to delete the website. A customer who already knows your name will still load the homepage. A model that has to pick a source in a second will not. Those are different readers. They need different objects.

The second page

Keep the site. Add a public record that is only facts. Not another landing page with a different headline. A document.

  • The legal name, not the nickname painted on the van.
  • The website the owner actually controls, not a Facebook URL that an intern still owns.
  • A phone number that rings.
  • A place or a service area, written as a place, not as a keyword in a footer.
  • A date, so a reader can see when those facts were last true.

Put a seal on the homepage that points at that record. Then a person and a crawler can land in the same place. The homepage stays a brochure. The file does the document job.

What Trendos did not say

The research did not say homepages are useless. It said they are poor citation targets in AI answers. It did not say user-generated pages are more accurate. A wrong paragraph still beats a correct room.

That last point is the risk. If forums and lists keep winning the citation, the answer box will keep sourcing pages the firm does not control. The fix is not to shout on those pages. The fix is to publish a page the model can lift without walking a theme.

What to do with the SEJ piece

Read the Trendos write-up in Search Engine Journal as a source, not as a ranking AI Verified issued. Quote the finding. Link the URL. Then look at your own homepage the way a crawler looks at it: can a machine take the legal name, the phone, and the city off the page without executing four scripts?

If the answer is no, the homepage is doing the job you hired it for — selling the next call — and failing the job you never hired it for — being a document. Hire a document. Leave the brochure alone.

Trendos owns the measurement. Search Engine Journal owns the write-up. This desk owns only that distinction. If the next alert those editors see from this domain is a clean citation of their work, that is the point. If a contractor reads this and stops asking a homepage to be a source file, that is the other point.