How Do I Recover a Deindexed Website When ChatGPT and Google AI Overviews Stop Recommending It?

Which index dropped you?

Start by working out where you are missing, since each assistant relies on different sources:

If you vanished from ChatGPT but still appear in Google, the problem is probably not Google's. The reverse is also true. Losing everything at once points to your site.

What should you check first?

1. Robots.txt: is there a rule such as "Disallow: /" for all crawlers, or for specific search and AI crawlers?
2. Robots meta tags and headers: do pages carry noindex, in the HTML or in an X-Robots-Tag header? Redesigns and CMS updates often leave these behind.
3. Canonical tags: do they point to a staging site, an old domain or the wrong page?
4. Status codes: do pages return 200, or errors such as 5xx, 403 or 404? Check with URL Inspection and a header checker.
5. Bot protection and firewalls: are search and AI crawlers being challenged or blocked? Some content delivery networks now block AI crawlers by default in certain configurations. Cloudflare, for example, changed its defaults for new domains during 2025 and 2026.
6. DNS, hosting and certificates: has the domain expired, has DNS changed, or has the SSL certificate lapsed?
7. Removals: has anyone used Search Console's Removals tool on the site?
8. Manual actions and security issues in Search Console.

Recovering after a robots.txt block specifically is covered in how to recover AI search visibility after robots.txt blocked ChatGPT, Perplexity and others.

How do you recover from an accidental block?

1. Fix the cause across every template, not just the home page.
2. Confirm with URL Inspection that Google can fetch and index key pages, and request indexing for the most important ones.
3. Resubmit sitemaps in Search Console and Bing Webmaster Tools.
4. Use IndexNow to notify Bing and other participating engines of changed URLs, if your site supports it.
5. Check server logs for returning crawler activity from Google, Bing, OpenAI and Perplexity.
6. Add safeguards: a deployment check that fails if production pages carry noindex or a blanket disallow.

Key pages often return within days. Full recovery for a large site can take weeks.

How do you recover from a manual action?

Manual actions are issued when a reviewer finds spam policy violations, such as thin or scaled content, unnatural links, cloaking or site reputation abuse. Search Console's Manual actions report lists what was found.

1. Read the report carefully and fix every issue on every affected page. Partial fixes lead to rejection.
2. Document what you changed. A clear record helps the review.
3. Request a review from the report. Google says most reconsideration reviews take several days or weeks, and link-related requests can take longer. You will be told by email when the review starts and ends. Do not resubmit before you get a decision.
4. If you bought the site recently, say so in the request. Google's guidance allows for sites that violated policies before their current owner.

How do you recover from a hacked site?

Treat it as a security incident. Clean the injected pages or code, close the vulnerability, change all passwords and keys, and check for spam pages or redirects. Then request a review from the Security issues report. Hacked content is a spam policy violation in its own right, and it also damages trust with users and assistants.

What should you tell customers and staff while you recover?

A deindexed site often means fewer enquiries for a few weeks. Tell staff what happened and roughly how long recovery should take, so they are not surprised by questions. Keep your Business Profile, social profiles and directory listings accurate and active, since those can still appear in search and AI answers while your own pages are missing. If customers ask why they cannot find you, a direct link and phone number solve the immediate problem. Track enquiries by source during recovery, so you can see when search and AI referrals return.

When will AI answers recommend you again?

• Google AI Overviews and AI Mode: they return as Google reindexes your pages, often within days to weeks of the fix.
• Copilot: follows Bing's reindexing.
• ChatGPT search answers: return as the third-party search indexes it draws on and OAI-SearchBot pick up your pages again. Allow OAI-SearchBot in robots.txt and in your firewall.
• Perplexity: returns as its crawler refetches your pages.
• Answers from model memory: may never have stopped mentioning you, since training data does not change when you are deindexed.

Some visibility may take longer to return than rankings. Assistants cite what ranks and what they have recently fetched. Pages that lost their position during the outage need time to regain it. If citations stay low after the site is back in the index, the content changes made during the redesign may be the cause, as explained in why ChatGPT citations disappear after a website redesign or content update.

What might an illustrative case look like?

Consider an illustrative law firm whose site disappeared from Google, and from AI Overviews for its practice-area questions, a week after launching a redesign. ChatGPT still mentioned the firm from memory but stopped citing its pages. The new templates carried a noindex tag from staging. The firm's CDN had also started challenging automated traffic, which blocked Bing and OpenAI's crawler. The developer removed the tag, allowed verified search crawlers, requested indexing for the 20 most important pages and resubmitted sitemaps. Key pages were back in Google within four days. AI Overview citations and ChatGPT search citations returned over the next three weeks.

What do people ask about deindexing and AI visibility?

Can I be deindexed from ChatGPT but not Google?

Yes. ChatGPT's search draws on different indexes and crawlers. Check your rules for OpenAI's crawlers and Bing's index status.

Should I create a new domain to escape a manual action?

No. Fix the issues and request review. Moving the same content to a new domain rarely solves the problem and can repeat it.

How can I stop it happening again?

Add automated checks for noindex tags, robots.txt changes and crawler blocking to every deployment, and monitor index coverage weekly.