Your site ranks in Google, but ChatGPT doesn't know it exists. What to check

It's a fairly common situation. A business ranks decently in Google, traffic comes in, the contact form works. Then someone asks ChatGPT for a recommendation in that line of work and that area, and the answer lists competitors. Or no one at all.

There's no single switch for this, and no one can promise that an assistant will start recommending you. There are, however, a few things that can cut a site off from assistants entirely, and they're easy to miss because Google doesn't treat them as problems.

1. AI crawlers aren't allowed in#

Start with your-domain.com/robots.txt. Look for names like GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot. If one of them has Disallow: /, that assistant won't read the site.

Sometimes that was added on purpose, to keep content out of model training. Fair enough, but the crawlers do different jobs. GPTBot collects training data, while OAI-SearchBot and ChatGPT-User read your page when someone asks about it in the chat. You can block the first and let the others in.

The second place to look is Cloudflare. It has a setting that blocks AI crawlers, and on some accounts it's on by default. Your robots.txt can look perfectly innocent while the crawlers get turned away before they ever reach the site.

2. The content only appears after JavaScript runs#

Googlebot runs JavaScript. Most AI crawlers, as far as anyone knows, don't. They read whatever the server sends in the first HTML.

If the server sends an empty shell and a script fills in the descriptions, prices and reviews, an AI crawler sees the empty shell. A quick test: open the page, right-click and choose "View page source" (not "Inspect"). Search for a sentence from the content. If it isn't there, the crawler doesn't have it either.

3. The answer is there, just not as text#

Opening hours sit in an image, prices in a PDF, the service area only on a map. A person takes all that in; a model building its answer from text usually doesn't.

Take three questions customers actually ask you by phone or email and check that each answer is on the site as a plain sentence. Text you can select and copy, not a picture or something hidden behind a tab.

4. The site never says plainly who you are#

It sounds obvious, but plenty of home pages have no sentence like "We're an accounting firm in Leeds handling the books for sole traders and small companies." There's a slogan, a photo and a button.

An assistant choosing three businesses to recommend needs that sentence. Structured data (JSON-LD) with your name, address and type of business helps too, as do proper titles and headings. It just leaves the model less to guess.

Where to start#

Points 1 and 2 are the gate: if the crawler can't get in or sees an empty page, nothing else matters. That's where I'd start. Points 3 and 4 are content work you can spread out over time.

The Anole audit reads your pages the way a crawler does, without running scripts, and shows the title, H1, canonical URL and JSON-LD for each one. So you'll see part of points 2 and 4 straight away. Check robots.txt and Cloudflare by hand, though. Those are your settings, and only you know which blocks are deliberate.