Why Does ChatGPT Still Cite Your Old Page After a Fix?

Q: Does ChatGPT follow 301 redirects?

Yes, but less reliably than Google does. ChatGPT's crawler already burns over a third of its fetches on 404s and redirects combined, so a redirect is more likely to get missed or take longer to resolve than the same redirect would with Google.

Q: How often does ChatGPT or Perplexity recrawl a site?

There is no fixed, published schedule. Check directly by asking the engine the real question and reading what it actually cites, rather than planning around a guessed number.

Q: Does this affect a small local business site, or only big sites?

The crawler behavior is not checking site size before deciding whether to follow a redirect cleanly, so a small site's broken redirect gets missed by the same mechanics as a big one's.

Aidan Blandford

Sep 8

4 min read

You redirected an old page, checked Google Search Console, and it shows the redirect resolved clean. Then you ask ChatGPT the same question a customer would ask, and it answers off the page you just killed. You didn't do anything wrong. AI crawlers run behind Google on this, and the gap is measurable, not a guess.

Why does this even happen?

Google has spent close to two decades tuning exactly how fast it drops a dead URL and picks up a redirect. The crawlers behind ChatGPT, Claude and Perplexity are new by comparison, and they are nowhere near as good at it yet.

Vercel tracked its own network traffic over several months, across nextjs.org and two job board sites built on different setups. ChatGPT's crawler spent 34.82% of its fetches landing on 404 pages and another 14.36% following redirects. Claude's crawler hit 404s at almost the same rate, 34.16%. Googlebot, on the same kind of traffic, spent 8.22% on 404s and 1.49% on redirects. Source: Vercel, "The Rise of the AI Crawler."

That means the crawler reading your site for ChatGPT is burning more than a third of its visits on dead ends, close to four times Google's error rate. A redirect Google resolves the same day can sit half followed by one of these crawlers for a while before it catches up.

Is this the same as ChatGPT just getting your business wrong?

No, and it is worth telling apart. If ChatGPT has a fact about your business wrong, that is usually because the information about you is inconsistent somewhere on the open web and the model picked up the wrong version. This is different. The content was right, you moved it, and the crawler has not caught up to the move yet. Same symptom, a stale-sounding answer, a completely different cause and a completely different fix. If it is the first problem, we covered that one separately.

How long should you actually wait before you worry?

There is no published schedule for this, nothing like Google's index you can check a date against. Give it a few weeks before you decide something is actually broken instead of just slow. Instead of guessing at a timeline, check it directly the way the next section covers. That beats waiting and hoping.

What actually helps it catch up faster?

A few things are in your control:

Use one clean redirect, not a chain. Every extra hop is one more chance for a crawler that already fails on a third of its fetches to give up before it reaches the real page.
Never let the old URL just 404. A dead page reads as content removed. A redirect reads as content moved. Those are different signals to a crawler.
Keep the lastmod date in your sitemap current on the page that changed, so the freshness signal is clean whenever a crawler does show up.
Do not assume Google catching up means every engine caught up. They run on separate crawls, on separate clocks.

How do you actually check what it is showing, instead of guessing?

Ask the exact question a customer would type, in ChatGPT, Perplexity, and Google's AI overview, and read exactly what each one cites and links back to. Search Console telling you the redirect resolved only tells you about Google's crawl. It says nothing about the other three.

This is also why we build our own site's articles to answer the real question in the first two sentences, so whichever crawler eventually gets there has something clean to lift right away. It is the same reason an agent trained straight on your own current content does not have this lag at all. It is not waiting on a third party crawler to notice anything changed, it reads what you actually have right now. You can see what that looks like at demo.ajmarketingresults.com.

Frequently Asked Questions

Does ChatGPT follow 301 redirects?

Yes, but less reliably than Google does. The Vercel data above shows ChatGPT's crawler already burns over a third of its fetches on 404s and redirects combined, so a redirect is more likely to get missed or take longer to resolve than the same redirect would with Google.

How often does ChatGPT or Perplexity recrawl a site?

There is no fixed, published schedule, so do not plan around a guessed number. Check directly instead: ask the engine the real question and read what it actually cites.

Is a page ChatGPT still cites the same as being deindexed by Google?

No. Google removing a dead page from search results and an AI engine still citing an old page in a chat answer are two separate systems running two separate crawls. Fixing one does not automatically fix the other.

Should I redirect the old URL, or just let it 404?

Always redirect, no matter how old the page is. A 404 tells any crawler the content is gone. A redirect tells it the content moved, and that is the signal you actually want it to pick up.

Does this affect a small local business site, or only big sites?

The Vercel numbers come from large-scale traffic, but the crawler behavior is not checking site size before it decides whether to follow a redirect cleanly. A small site's broken redirect gets missed by the same mechanics as a big one's.

AJM