Your brand disappearing from ChatGPT answers usually means one of three things changed: ChatGPT can no longer read your site, your pages are no longer eligible for its search results, or another source now answers the question better than yours does. There is no single switch that turns a brand back on, and no vendor publishes a refresh schedule for brand knowledge, so do not look for a fixed waiting time. The work is to check access first, then check eligibility, then check the pages themselves, in that order. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and that placement is not guaranteed even for sites that are eligible, so access is necessary and still not a promise.
Start with the simplest test. Ask ChatGPT the same question that used to cite you, with search turned on, and save the answer and its citations. Then ask it again with slightly different wording. If your brand is gone from every wording, the cause is likely access or eligibility. If it is gone from one wording and present in another, the cause is likely the pages: the winning page answers a different question than the one being asked now. That distinction decides what you fix first.
Where it goes wrong
A disappearance has a cause you can find, and the cause is usually one of these.
- OAI-SearchBot is blocked. A robots.txt rule, a firewall rule, or a bot management rule stops the search crawler, and OpenAI says opted out sites will not be shown in ChatGPT search answers. The block often arrives with a security change nobody connected to the disappearance.
- The search crawler is allowed in robots.txt but blocked at the edge. The file says allow, yet the host, the CDN, or a bot rule returns an error to the crawler's addresses. OpenAI publishes the search crawler IP ranges and recommends allowing requests from them, so a robots.txt check alone is not enough.
- GPTBot was blocked by mistake for search. GPTBot and OAI-SearchBot are separate settings, and OpenAI documents them as independent. Blocking GPTBot signals that content should not be used for training, and it does not by itself remove a site from search. But a broad rule that blocks both tokens, or a wildcard rule that catches both, removes the site from search as well.
- The page that used to win is gone or changed. A product page was removed, a category page was rewritten, or a guide was taken down, and nothing replaced it as the answer to that question. ChatGPT cites pages, not brands, so a missing page is a missing citation.
- The page is live but no longer readable. A redesign moved the key facts into images, a script, or a login wall, and the crawler now sees a shell instead of the facts. Google makes the same point for its own AI features: important content should be available in textual form.
- The question changed. A prompt that used to match your page now matches a competitor page, a review roundup, or a retailer listing more closely. Nothing about your site broke. The answer to that question simply has a better source now.
- The answer came from training, not search. If the old answer had no citations and read like remembered knowledge, it likely came from training data. Training based answers change only when a newer model arrives, and no source fix moves them sooner. No vendor publishes when that happens, so do not plan around a fixed time.
The steps
Step 1. Reproduce the disappearance with saved prompts. Write down the exact prompts that used to cite you, and ask them again with search turned on. Save each answer, each citation, and which assistant gave it. Test three wordings: the exact old wording, a shorter version, and a version that names your product category. If your brand appears in one wording and not the others, the pages are the problem. If it appears in none, access or eligibility is the problem.
Step 2. Check robots.txt for OAI-SearchBot. Open your robots.txt file and look for a group naming OAI-SearchBot. Confirm it allows the paths you want cited. Look for a wildcard group that disallows everything, since a named group overrides the wildcard for that crawler, and a missing named group leaves the wildcard in charge. Confirm GPTBot has its own group, so a training decision does not accidentally become a search block. OpenAI says it can take about 24 hours from a robots.txt update for its search systems to adjust, so note the time you change the file and wait before you retest.
Step 3. Check the edge, not just the file. Ask your host or CDN whether requests from OpenAI's published search crawler addresses are allowed, and check your server logs for OAI-SearchBot visits to your key pages. A visit that returns a success response is the evidence that matters. A run of errors or challenge pages on those addresses is the warning sign. If your firewall blocks by user agent, add an explicit allow for the search crawler by name and by its published addresses.
Step 4. Check that the winning page still exists and still answers the question. Open the page that used to be cited. Confirm it loads, shows the key facts in plain text, and matches the question being asked. If the page was removed, restore it or redirect it to the closest live page that answers the same question. If the page was rewritten for shoppers and the facts moved into images or a script, put the facts back in text. A page that no longer answers the question cannot win the citation back.
Step 5. Check the same access for the other engines. A brand that disappeared from ChatGPT alone is a ChatGPT access problem. A brand that disappeared everywhere at once is a site wide problem, such as a bad deploy, a new login wall, or a platform change. Check PerplexityBot and Googlebot access the same way, and check your sitemap for dead addresses. A sitemap that still lists removed pages, or a sitemap that was never submitted, slows every engine down at once.
Step 6. Rebuild the page that should win. Pick one page per question you want to win back, and make it the clearest answer on your site: the product name, the price, the availability, and who it is for, in plain sentences at the top. Make the structured data match the visible text. Link to the page from the category page and the homepage so crawlers find it without guessing. Do not create five thin pages for one question. One strong page beats five weak ones, and five weak pages split the citation.
Step 7. Request a fresh look from Google and retest ChatGPT. For Google, inspect the rebuilt page with the URL Inspection tool in Search Console and request indexing. Google says a request does not guarantee the page will appear in the index, and that indexing can take up to a week or two. For ChatGPT, there is no equivalent request button for site owners. The route is access plus a clear page, then a retest. Ask the same saved prompts again after the robots.txt change has had time to take effect, and log each answer and citation. If the brand is still missing after access is confirmed and the page is clear, the remaining cause is competition for that question, not access, and the fix is a better page, not another access change.
The machine-readable piece
This robots.txt snippet is the access fix for ChatGPT search. It allows the search crawler, keeps the training crawler out, and leaves the user triggered fetcher open. The settings are independent, as OpenAI documents, so each token needs its own group. Check your full crawler reference in the fundamentals section before you add other engines.
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /
User-agent: ChatGPT-User
Allow: /
Check the file after you publish it. Open the robots.txt address in a browser and confirm the OAI-SearchBot group is there. Check your server logs for OAI-SearchBot visits and confirm they return a success response on your key pages. If the logs show errors or challenge pages, the fix is at the host or the firewall, not in the file. Wait about 24 hours after a robots.txt change before you judge the result, since OpenAI says its search systems take about that long to adjust.
What it looks like when it works
On the page, nothing looks different to a human reader. The change is under the surface: the robots file names OAI-SearchBot and allows it, the firewall allows its published addresses, and the page that should win shows the key facts in plain text. A crawler that visits the page gets a success response and reads the facts. An assistant asked the same question cites your page again, with a link a shopper can open.
Say you run a skincare brand and ChatGPT used to cite your moisturizer page when shoppers asked for a fragrance free moisturizer for sensitive skin. Then the citations stop, and a competitor page wins the question instead. You check robots.txt and find a new security rule added a wildcard block that catches OAI-SearchBot. You add a named group that allows OAI-SearchBot, confirm the firewall allows its published addresses, and check the logs until the crawler returns success responses on the moisturizer page. You also find the page was rewritten and the words fragrance free moved into an image, so you put them back in plain text at the top. You wait about 24 hours after the robots.txt change, then retest the same prompt. The answer cites your page again. If it does not, you compare your page with the competitor page that won, and you improve the page instead of changing the access rules again.
The fundamentals
These mechanics are shared across every engine and covered in full in their own guides:
- •The llms.txt format and a complete sample, see our guide to writing a good llms.txt file
- •The markup AI systems read to verify facts, see our product schema guide
- •The complete crawler rule reference, see our robots.txt guide for AI search
What this guide does not cover
This guide covers diagnosing a disappearance from ChatGPT answers and fixing access, eligibility, and the winning page. It does not cover disappearing from every assistant at once for a site wide outage, which is a hosting and platform problem before it is an AI problem. It does not cover training based answers, which no source fix moves until a newer model arrives, and no vendor publishes a schedule for that. It does not cover the monitoring work of catching the next disappearance early, or the full technical audit that checks every crawler and every page. Those are separate jobs, and this guide is the diagnosis, not the whole program.
Check your own store now
Run the free scan. One click, no signup, and it shows whether AI crawlers can read your store today: aizii.ai/aeo-scan
Run the free scanSources
- OpenAI, Overview of OpenAI Crawlers (specific): https://platform.openai.com/docs/bots
- Perplexity, Perplexity Crawlers (specific): https://docs.perplexity.ai/docs/resources/perplexity-crawlers
- Google Search Central, AI features and your website (specific): https://developers.google.com/search/docs/appearance/ai-features
- Google Search Console Help, URL Inspection tool (specific): https://support.google.com/webmasters/answer/9012289
- Google Search Central, Build and submit a sitemap (specific): https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap
- Google Search Central, Product structured data (specific): https://developers.google.com/search/docs/appearance/structured-data/product
Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
