Why ChatGPT doesn’t mention your company

A buyer asks an assistant which suppliers to consider in your category. It names three or four companies. You are not among them, and at least one of the names is a competitor you routinely beat on merit. This is a common experience, and the causes are more mundane than most people expect.

First, the reason it is probably not

It is almost never a judgement about quality. These systems have no view on whose product is better. They assemble an answer from material they could find, read and verify. A company with an excellent product and unreadable documentation loses to a mediocre competitor with a clear specification table, every time.

The seven causes, in rough order of how often they turn out to be the problem

1. Your specifications are trapped in PDFs

This is the single most common cause in manufacturing. Every meaningful technical detail — dimensions, tolerances, materials, performance data, certifications — sits inside a downloadable datasheet. Some systems extract PDF content unreliably; many do not attempt it at all. To a machine, a page consisting of a product name and a download link contains almost nothing.

2. Your product data is rendered by JavaScript

Catalogues built as applications frequently deliver an empty document to anything that is not a full browser. The page looks complete to you and arrives as a shell to a crawler. This is invisible unless you look, which is why it survives for years.

3. Your best material is behind a form

Gating a comparison guide or specification library behind a lead-capture form was defensible when the objective was harvesting contact details. It now means the most citable asset you own is unavailable to the systems increasingly deciding who gets considered. The trade has changed and most gates have not been revisited.

4. You have no structured data

Without Organization, Product or FAQPage markup, a machine has to infer what your pages are about. Sometimes it infers correctly. Often it does not resolve you to a distinct entity at all, which makes citing you a risk it will simply avoid.

5. Your content answers nothing directly

Most manufacturer pages describe a product. Very few answer a question. Generative systems are assembling answers, so they look for passages that already are one. A page headed “Product Overview” followed by three paragraphs of positioning offers nothing liftable. A page headed with the question a buyer asks, answered in the first fifty words, offers a great deal.

6. Nothing corroborates you

These systems weigh whether a claim appears in more than one place. If everything about your company exists only on your own website, each claim rests on a single self-interested source. Industry directories, association listings, trade publications and specification databases all raise the confidence with which a system will repeat something.

7. Your content is unattributed

Pages with no named author, no credentials and no clear organisational identity are cited noticeably less. Systems favour material they can attribute to somebody, because attribution is a proxy for accountability.

Find out before you fix anything

Do not start remediating on a guess. Establish which of these actually applies to you by testing what the engines currently say, then working backwards from what they cited instead. The cited competitor page will tell you precisely which format the system preferred — it is a better specification than any general advice, including this page.

Related reading: what generative engine optimization involves and how GEO, AEO and SEO differ.