Why AI assistants never cite your company
You ask an assistant who to consult in your speciality, in your region. It names three competitors and ignores you. That is almost never a fame problem: it is a problem of usable material.
Christophe Bellec ·
Short answer
An answer engine cites a source when it can, all at once: reach the content, extract a precise statement from it, identify the entity behind that statement, and corroborate it elsewhere. Missing any one of those four links is enough never to be cited. In most cases the missing link is content extractability or the absence of consistent external traces, not company size.
What actually happens when the assistant answers
On a question that calls for current information, the assistant runs one or more searches, fetches a small number of pages, extracts passages, and composes an answer from what it found.
Every step is a filter. Your page has to be found, then reached, then retained among the few sources read, then supply a passage sharp enough to be reused. A page can clear the first three filters and fail the fourth simply because it is written too vaguely.
Reason 1: The content is not reachable
The most ordinary case, and the most common on recent sites: essential content only exists after JavaScript runs. The robot fetches an empty shell. The page is perfect for a human and non-existent for an engine.
Variants: navigation with no link tags, a `Disallow` inherited from staging, pages reachable only through an internal search engine, response times that are too long.
Reason 2: Nothing is extractable
This is the most widespread reason among companies whose site is technically impeccable. The text exists, it reads well, but it contains no statement an engine could reuse as-is.
“We put our expertise at the service of your transformation” says nothing. What the assistant needs to find looks more like: what you do precisely, for whom, in which situations, in how long, with which deliverables, in which geography, and what you do not do.
Quick test: take a paragraph from your home page, show it to someone out of context and ask them to derive one factual sentence about your company. If they cannot, an engine cannot either.
Reason 3: The entity is blurry
An engine does not cite anonymous text: it cites someone. It still has to know who.
- The brand, the person and the legal entity are never explicitly connected anywhere.
- The name varies from source to source, with or without a hyphen, with or without a suffix.
- No page states clearly who actually delivers the work.
- Public profiles state a different positioning from the site.
- The markup, where it exists, redeclares a different company on every page instead of pointing at the same entity.
A well-defined entity fits in one sentence, repeated identically everywhere: who, what, for whom, where. Variations do not make you semantically richer; they dilute the signal.
Reason 4: Nothing corroborates externally
An answer engine distrusts a single source, and rightly so: anyone can write anything on their own site. What gives a claim weight is finding it elsewhere.
“Elsewhere” does not mean artificial directories. It means public traces consistent with what you say: up-to-date professional profiles, public code repositories, technical publications, talks, mentions in articles, official company records.
This is often the weak point of independents and small firms with an excellent site and no existence anywhere else. And it is the only lever that cannot be fixed in code.
Reason 5: You do not answer the questions being asked
Questions put to an assistant are sentences, often long, often situated: “who can audit the architecture of an internal application before a rebuild, in the Paris area?” A site organised around generic phrases never intersects that kind of wording.
Covering an intent does not mean repeating keywords. It means there is a page answering that precise question, with the context, the conditions and the limits, including the cases where you are not the right person.
What order to fix things in
- Check what a robot actually fetches: title, introduction, offers, FAQ, links. Nothing else has any effect until that is settled.
- Rewrite the opening paragraphs of the key pages so they contain facts, not intentions.
- Establish a single entity: the same name, the same description, the same links everywhere, and markup pointing at the same entity from page to page.
- Align public profiles word for word with the site.
- Create the two or three pages that answer the questions genuinely asked in your field.
- Publish a few substantial pieces on your subject rather than many shallow ones.
What no provider can sell you
There is no ranking inside an answer engine, no position to buy, no index to consult. Nobody can guarantee a citation, and an agency promising “the top spot on ChatGPT” is selling something else.
What can be worked on are the conditions that make a citation possible. They are entirely verifiable, which is already a good deal more honest than a promise of position.
Frequently asked questions
Do you have to be a large company to get cited?
No. A small firm with precise content, a clear entity and a consistent external presence is easier to cite than a large group speaking generically. Size helps with external mentions; it does not replace extractability.
How long before there is any effect?
Nobody can state that seriously: answer engines publish no index, no ranking and no refresh interval. What can be measured are the fixes applied and how engines fetch your pages, not a position.
Does publishing a lot of articles help?
Publishing a lot of shallow content dilutes the signal. A few pieces that genuinely treat a subject in depth make a source reusable across a whole family of questions; that is what builds topical authority.
Where do I start if my site is already live?
With the technical survey: what a crawler actually sees of your pages, page by page. That is the first step of an SEO and GEO audit, and it determines whether the other work is worth doing yet.