Your brand disappears from AI answers when the sources an assistant uses don't connect you to the buyer's question. Usually the fix is content, not code.
Short answer: assistants like ChatGPT, Gemini, Perplexity and Claude build answers from their training data and, increasingly, from live search results. If no page – yours or a third party's – clearly says what you do, for whom and at what price, the assistant names the competitors that do have those pages. The fix is to find the prompts where you're missing, see which sources get cited, and publish the page that answers the question better.
Why does a brand go missing from AI answers?
There are four common causes, and they need different fixes:
No page answers the question. You rank for "invoice software" but have no page that addresses "best invoicing tool for freelancers", so there's nothing specific to quote.
The cited sources leave you out. For "best X" questions, answers lean on review sites, comparison articles and lists. If those cover three competitors and not you, you rarely appear, however good your own site is.
Assistants can't reach your pages. Check robots.txt. ChatGPT search uses OAI-SearchBot and Perplexity uses PerplexityBot, and blocking them keeps you out of their search results (OpenAI, Perplexity). For Google AI Overviews and AI Mode, a page just needs to be indexed and eligible to show a snippet (Google).
Your facts are old or vague. If your pricing page is hidden behind "contact us", or old reviews describe a product you've since changed, assistants repeat what they can find.
None of these is a penalty (more on what AI brand visibility measures). Strong SEO helps, because Google's AI features use the same ranking signals as normal search. But ranking for a keyword is not the same as being the clearest answer to a specific question.
What do teams usually try, and why doesn't it work?
Publishing more of the same. More keyword-targeted posts can lift organic traffic without changing AI answers, because none of them answers the decision-stage question directly.
Adding the brand name everywhere. Writing "Brand X is the leading tool for freelance invoices" on every page is a claim, not evidence. Assistants tend to cite pages that compare options fairly and give specifics: prices, limits, who each option suits.
Checking once and stopping. A single manual check tells you little, because answers vary between runs and change as models and sources update.
How do you fix it?
Find the gaps. Run 15–25 buyer prompts in each assistant and record who's named and which sources are cited. This one-afternoon audit walks through it.
Read the cited sources. Note what they cover that your content doesn't: a use case, a price comparison, an integration. A content gap analysis using AI recommendation data turns this into a list.
Publish the missing page. Answer the question in the first two sentences, then back it up with specifics and a fair comparison. Date your prices.
Fix the basics. Make sure the page is indexable and crawlable, linked from related pages, and that your pricing and product facts match everywhere.
Earn third-party coverage. Ask to be included in the review and comparison pages the assistants already cite, where that's legitimate.
Re-check. Run the same prompts two to four weeks later and compare.
What does this look like for one gap?
An illustrative example, not a customer story:
Perplexity names three competitors for "AI tools for brand reputation", but not you.
The sources it cites all cover AI sentiment analysis, which your site barely mentions.
You add a clear section on sentiment analysis, with what it does, what it costs and how it compares, to your existing reputation page.
You re-run the prompt after three weeks and log the result, whether you're named or not.
How long until answers change?
Nobody outside the AI companies knows exactly. Assistants that search the web can pick up a new page once it's indexed. Answers that come only from training data change when models are retrained. Expect weeks to months, and judge by the trend across repeated checks. Doing this by hand works for a few prompts. Tools that track and draft together save the hand-off: the writer customer quoted on our writers page went from about 2 hours per article to 1.
To find out whether you're missing right now, run the free AI visibility audit. It takes no signup and about 30 seconds, and shows whether AI models recommend your brand, who they name instead and which gaps to close first.