Why is ChatGPT ignoring my website?

First, check that ChatGPT actually can't see you

Before you rewrite anything, run the test. Open ChatGPT with browsing on, ask the five queries a customer would actually type, and watch what it cites. If you show up once and you think you should show up five times, that's a different problem than not showing up at all.

People skip this step and end up fixing the wrong thing. A site that gets cited for one query and ignored for four others has a content gap, not a crawler problem. A site that never appears for any query usually has a foundational issue. Diagnose first.

1. The crawler is getting blocked

The OpenAI fetcher is called OAI-SearchBot, and it has its own user agent separate from the training crawler GPTBot. A surprising number of sites block one, the other, or both in robots.txt without realizing it. Cloudflare's default bot fight mode also blocks them at the edge.

Check your robots.txt at yoursite.com/robots.txt. Look for Disallow lines under OAI-SearchBot, GPTBot, or *. If you're on Cloudflare, check the AI Scrapers and Crawlers toggle in the dashboard. If you're on Vercel or Netlify, check your firewall rules.

2. The important content needs JavaScript to appear

OAI-SearchBot does not run JavaScript the way Googlebot does. If your hero copy, your H1, your pricing, or your FAQ only renders after a client-side fetch, the bot sees an empty shell.

Right-click your page, view source, and search for a sentence from the body. If it isn't there in the raw HTML, ChatGPT isn't reading it either. Server-render the parts you want quoted, or pre-render them at build time.

3. You're not answering a question anyone is asking

ChatGPT cites pages that answer the user's actual prompt. Marketing pages that lead with brand story, mission, and a hero video tend to get skipped because there's no answer to extract.

Pull up your top five service pages and read the first 100 words out loud. If a stranger couldn't tell what the page is for, neither can a model. Rewrite the opening to state the problem the page solves and answer the obvious follow-up question in plain language.

• ·Lead with the answer, not the lede
• ·Use the exact phrase a customer would type as a heading
• ·Put the price, the timeline, and the location in the body, not a PDF

4. Nobody else on the internet talks about you

ChatGPT cross-references. If your site claims you're the best dog trainer in Austin but no review site, no forum, and no local news mentions you, the model has nothing to corroborate. It picks the option three other sources agree on.

This is the slowest fix and the most important. Get listed in the directories your customers actually check — Moz Local pushes your NAP out to the big aggregators in one shot. Pitch a local reporter, or build the question-shaped pages journalists and models both pick up with Frase. Answer the same question on the relevant subreddit. Ask happy customers for reviews on the platforms that actually feed ChatGPT (Google, Yelp, industry-specific sites) — NiceJob automates the ask so you stop forgetting. Then watch the mentions roll in with Brand24 so you know when a new source picks you up.

5. Someone else already owns the answer

For a lot of queries the model has a strong default citation, usually a well-known site that wrote the canonical answer in 2019 and has been ranking for it ever since. Outranking the default in classic search took years. In ChatGPT it can take weeks, but only if your version is meaningfully better.

Read the page ChatGPT cites today. Find the parts that are vague, out of date, or missing context. Publish a version that closes those gaps and links out generously. Models reward depth and specificity, not word count.

6. Your schema is wrong, missing, or fighting your copy

Models lean on structured data to confirm what a page is about. A FAQPage schema with five real Q and As often gets a page pulled into an answer when the same page without schema is ignored.

Audit your schema in Google's Rich Results Test. If a page has Article schema but no author, no datePublished, and no organization linked, fix that first. If you have Product schema without a price or availability, expect to be skipped.

7. The page reads like it was written by a robot

Ironic, but true. Pages stuffed with phrases like "in today's fast-paced digital landscape" and "unlock the power of" get filtered out because models recognize their own slop. So do thin pages that repeat the keyword 14 times without saying anything.

Write the page for a specific person with a specific problem. Use real numbers, real examples, real names of products and places. The same instinct that makes a page useful to a reader makes it citable.

8. You're brand new and ChatGPT hasn't met you yet

Browsing tools fetch live pages, but the model's prior beliefs come from training data that's months or years behind. A site three months old has almost no training-data footprint, so it can only show up through live search, which is a smaller surface.

Speed this up by getting your URL into places that get crawled often. A real GitHub repo, a Wikipedia citation, a guest post on an established site, a Reddit AMA. Anything that puts your domain near other established domains shortens the discovery curve.

9. Your site is technically fine but the topic is one ChatGPT avoids

Health claims, legal advice, financial recommendations, and a handful of other regulated topics get conservative treatment. Even a clean, well-structured page in those areas may get summarized in general terms and not cited specifically.

You probably can't fix the policy. You can earn citation by being the source the model paraphrases. Publish primary research, original survey data, or first-person case studies that other sites will quote, and let the citation come through the corroboration loop in problem 4.

Where to start tomorrow

1. 01Run the five-query test in ChatGPT and write down what it cites today.
2. 02Check robots.txt and any edge firewall for OAI-SearchBot and GPTBot rules.
3. 03View source on your three most important pages. Confirm the headline and the first paragraph are in the raw HTML.
4. 04Pick one page and rewrite the opening 100 words to lead with the answer.
5. 05Add or fix FAQPage schema on one high-intent page.
6. 06Find one place where a customer is asking your exact question online, and answer it without selling.