ChatGPT uses three different crawlers, not one
It's easy to think of "ChatGPT" as one thing looking at your site. In practice, OpenAI operates three separate, named crawlers, and each does a different job. Knowing which one matters for your situation changes what's actually worth fixing.
GPTBot gathers content for training future models. Allowing or blocking it in robots.txt affects whether your content becomes part of that future training data — it has nothing to do with whether ChatGPT can find you today.
OAI-SearchBot is what powers ChatGPT's search feature — the one that matters most for a customer asking ChatGPT to search the web for a business like yours. If this is blocked, you won't show up in ChatGPT search results at all, regardless of how good your site is.
ChatGPT-User fetches a page live, in real time, whenever someone explicitly asks ChatGPT to look at a specific site or check current information. This is the one that reads your actual page content in the moment — the same reason a stable, machine-readable page matters more than a polished-looking one.