AI search
Why did a brand vanish from ChatGPT answers?
First confirm it: run a fixed set of prompts several times, because answers change with wording, session, location and model. If the brand is really gone, check the likely causes in order: blocked crawlers or removed pages, inconsistent brand facts, a competitor's new content, lost reviews or mentions, then a model or index refresh.
What you'll see in 10 minutes
- Whether the brand appears in ChatGPT, Perplexity and Google AI results for its category questions
- A site and SEO audit with crawl and indexing findings ranked by severity
- Review and mention monitoring across the sources the brand is talked about in
Did the brand really vanish, or did the answer just change?
It usually arrives as a screenshot: the founder asked ChatGPT for the best brands in their category, and their name wasn't there. Last month it was.
Start by checking whether anything has actually changed. ChatGPT answers vary with the exact wording of the prompt, the session and its history, the user's location, whether search was used, and which model answered. One screenshot is one sample.
Build a fixed prompt set
Write down the questions a real buyer would ask: category ("best handmade wall clocks in India"), comparison ("brand A vs brand B") and problem questions ("how do I choose a rug for a small bedroom"). Freeze the wording so runs can be compared.
Run it more than once
Run every prompt several times, in fresh sessions, logged out where you can, and note the date, location and model for each run. Record whether the brand is named, whether it is linked, and who is named instead. If the brand appears in some runs and not others, that's normal variation, not a disappearance. If it is absent from every run where it used to appear, you have something to explain.
What does it cost to get this wrong?
Treat noise as a crisis and the team spends a fortnight rewriting pages that weren't the problem. Dismiss a real drop and a competitor fills the answers for months. Either way the agency loses the conversation, because it had no evidence to show.
What are the likely causes?
OpenAI doesn't publish how ChatGPT chooses which brands to name. What follows is a working list built from what is documented (crawler access) and what is reasonable inference (everything else). Say which is which when you report it.
The site stopped letting the crawler in
This is the one cause that is documented. According to OpenAI's crawler documentation, OAI-SearchBot is used to surface websites in ChatGPT's search features, and sites that block it won't be shown in ChatGPT search answers, though they can still appear as navigational links. GPTBot is a separate crawler for model training, and the two settings are independent. OpenAI also notes that a robots.txt change can take about 24 hours to take effect.
Check robots.txt, CDN or firewall bot rules, and any recent site migration. A security setting switched on during a sale can block crawlers without anyone noticing.
Pages were removed or changed
A redesign that deleted the buying guides, merged collection pages or moved content behind scripts can remove the pages an answer used to draw on. Compare the current site with the pages that ranked or were cited before the drop.
Brand facts don't agree
If the website, marketplace listings, Google Business Profile and social bios give different founding years, cities, price ranges or product claims, a model has less reason to state any of them with confidence. This is inference, but cleaning it up costs little and helps search too.
A competitor gained ground
New comparison content, press mentions or listicles that name a rival can change which brands an answer picks. Look at who now appears in the runs where the brand is missing, and what has been published about them recently.
Third-party mentions or reviews fell away
Answers that rely on search often draw on reviews, listicles and forums. If a roundup that named the brand was updated without it, or review volume dropped, the brand may have lost a source it relied on. Again, inference: worth checking, not worth asserting.
The model or index changed
Sometimes nothing on the brand's side changed: a new model or a search-index refresh can reorder results. You can't fix this directly, only make sure the brand is strong on the causes above.
What should you fix first?
Fix in order of certainty and cost.
- Access. Allow OAI-SearchBot and confirm that the CDN isn't blocking it. Restore or redirect removed pages that carried category answers.
- Facts. Make the brand's name, location, categories and key claims match across the site, listings and profiles.
- Answer pages. Make sure the site has clear, crawlable pages that answer the questions in the prompt set directly.
- Mentions. Work on reviews and genuine third-party coverage. This is slow, so start it now and judge it over months.
How do you track it from here?
Turn the prompt set into a monthly measure. Run it on a schedule, record how often the brand is named against named competitors, and report the trend rather than single answers. That share is what we call AI share of voice. A trend line ends the screenshot arguments, because the brand can see whether a gap is noise or a slide.
Pair it with a regular site audit, so a blocked crawler shows up before the answers change. The AI search guide for D2C brands covers the content side in more depth.
How does ai-dm handle this?
ai-dm's AI search visibility monitoring runs category questions across ChatGPT, Perplexity and Google AI results and records whether the brand and its competitors appear, so the agency has a trend instead of a screenshot. The site and SEO audit flags crawl and indexing problems by severity. Review and mention monitoring shows where the brand is being talked about. ai-dm recommends what to look at; the team decides what to change.
What should you do this week?
Write the prompt set, run it several times, and check robots.txt and CDN rules first. Then send the brand a short note: what you tested, what you found, which causes are documented and which are inferred, and what you'll measure each month from now on.
Related pages
More posts
Impressions up, clicks down: is it AI Overviews?
Search Console shows more impressions and fewer clicks. How to tell whether AI Overviews explain it, what to tell the brand and what to do next.
Read the post →What did 344k AI-result impressions teach us on Moolwan?
Our D2C home décor brand Moolwan earned 344k impressions in Google's AI results. What we see, what we think helped, what we don't know yet.
Read the post →
Questions agencies ask
How do I know a brand has really disappeared from ChatGPT?
Run the same fixed set of prompts several times, in fresh sessions, and note the location and model each time. One missing answer proves little; a consistent absence across repeated runs is a real signal.
Does blocking GPTBot remove a brand from ChatGPT search?
Not according to OpenAI's crawler documentation. GPTBot is for model training; OAI-SearchBot is the crawler that surfaces sites in ChatGPT's search features, and the two settings are independent.
What should we fix first?
Access first. Check robots.txt, CDN rules and removed pages, because nothing else matters if the crawler can't read the site. Then fix inconsistent brand facts, then work on mentions and content.
Is there a documented ranking formula for ChatGPT answers?
No. OpenAI documents its crawlers and how to allow them, but not how answers choose which brands to name, so treat any cause you find as a likely explanation, not a proven one.
No plan to pick. Billed in USD and INR. A brand's data is deletable on request.