NavlioRun a full report

How to show up in ChatGPT

If you ask ChatGPT about your own company and get a blank, a hedge, or a description of somebody else, that is worth understanding before you try to fix it.

Why it draws a blank

A model answers from two places. There is what it absorbed while being trained, which is frozen until the next release. And there is what it retrieves while you are talking to it, which is live. A young company written about in few places is in neither, so the model has nothing to work from and either says so or guesses.

Why it sometimes gets you wrong

Being described inconsistently is worse than being described rarely. When your homepage, your directory listings and your social profiles each frame the company differently, a model has several conflicting accounts and no reason to prefer any of them. What comes out is often a blend that matches none of the sources.

What actually moves it

The retrieved half responds to work you can do this month. The trained half moves on the timescale of model releases and is mostly out of your hands.

  • Say plainly, on your own site, what the company does and who it is for.
  • Use the same description everywhere the company appears.
  • Get written about somewhere the model already reads.

Check before you start

Measure what the models say now, so you can tell later whether anything you did made a difference.

Questions

How long before ChatGPT knows my business?

Retrieval can pick up a new page within days. Training memory moves when a new model ships, which is months. Expect the retrieved half to change first, and judge your work on that.

Does an llms.txt file make ChatGPT cite me?

On its own, no. It is a convention for telling a model what a site contains, and it is cheap to publish, but nothing about it obliges anybody to read it. Treat it as tidy housekeeping rather than a lever.