Why doesn't ChatGPT mention my business?

When someone asks ChatGPT for the best tool in your category, the assistant does not scan every company that exists. With browsing on, it retrieves a handful of pages (roundups, review sites, directories, forum threads) and builds its answer from the businesses those pages agree on. Your own website barely counts. It is a single voice, and the one with the most obvious motive to flatter you. We break down the retrieval step in how ChatGPT decides which businesses to recommend.

How to check: search Google for "best [your category]" and open the top ten results. Count how many mention you. Then ask ChatGPT the same question and compare its answer with those pages. If you are absent from the sources, you will be absent from the answer, and the overlap between the two lists usually makes that obvious. If it names a competitor in your place, why AI recommends your competitors shows how to read the pages behind that choice.

What fixes it: independent coverage. That means reviews on the platforms your category uses (G2, Trustpilot, Capterra or their equivalents), inclusion in the comparison articles that already rank, accurate directory listings, and mentions in the communities where your buyers ask questions. None of this is fast. Building enough consensus to shift answers takes months of steady work, which is why it is worth confirming this is your actual problem before you start.

Assistants match language. If you position yourself as a "revenue orchestration platform" but buyers ask for "sales pipeline software", the model has to bridge that gap on its own, and it often will not. This is a naming problem, not a quality problem: the assistant may know your business exists but file it under a category nobody queries.

How to check: write down the five phrasings a real buyer would actually type, not the ones on your homepage. Ask ChatGPT each one. Then search your business name alongside your category and read how third parties label you. If review sites call you one thing and your site calls you another, the mismatch is costing you retrieval.

What fixes it: pick the category term buyers use and apply it everywhere you control, starting with your homepage and your directory profiles. Then nudge the sources you do not control: most review platforms let you edit your own listing category. This is one of the quicker fixes on this list, though re-crawling and re-indexing still mean weeks before answers move.

Assistants blend what they memorised in training with what they retrieve live, and both layers can be wrong. In a July 2026 test of 30 brands, 27 of 30 (90 per cent) were described with at least one materially false claim when the assistant answered from memory, and 13 of 30 (43 per cent) still contained one with live web search. Old pricing sitting on a 2022 review, or a pivot that never reached your Crunchbase profile: when sources disagree, the model either hedges or quietly leaves you out.

How to check: ask ChatGPT "what does [your business] do and what does it cost?" and read the answer as a fact-checker would. Then search your business name and skim the first two pages of results for anything stale. You are looking for the description a stranger would assemble, not the one you would write.

What fixes it: a cleanup pass. Update every profile you control (LinkedIn, Crunchbase, review site listings) and correct third-party pages wherever the site allows it. Then publish one clear, current description of what you do and what it costs on your own site, so there is something accurate to retrieve.

ChatGPT's browsing relies on crawlers, chiefly GPTBot and OAI-SearchBot. Plenty of sites block them in robots.txt without realising it: a CDN default, or a blanket bot rule left over from 2023. Even with crawlers allowed, a site that renders everything with client-side JavaScript can serve them a nearly empty page, because most answer-engine crawlers read the raw HTML and do not execute your scripts.

How to check: open yoursite.com/robots.txt and look for GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot. Then load your homepage with JavaScript disabled (or fetch it with curl) and see whether your product description survives. If what remains is a loading spinner and a script tag, that is what the crawler sees too.

What fixes it: allow the crawlers you want in robots.txt, and make sure your key pages (your homepage and pricing page in particular) are server-rendered or prerendered so the text exists in the HTML. An llms.txt file is a cheap addition that gives assistants a plain-text map of your site, though treat it as a supplement to readable pages, not a substitute. Unblocking takes minutes; being re-crawled and reflected in answers takes weeks.

Assistant answers are probabilistic. The same question asked twice can return different brand lists, and small changes in wording shift the results further. Your own ChatGPT account adds another distortion: memory and custom instructions mean your session is not a neutral test. It is entirely possible to be recommended regularly for phrasings you have never thought to try, while being absent from the two prompts you did.

How to check: test from a fresh session with memory off, because your account history skews what you see. Run each question five times and vary the phrasing (best X, X for [use case], alternatives to [competitor], is X worth it). Score mentions across the full set rather than reacting to a single answer. One absence is noise. Twenty absences across varied prompts is a signal.

What fixes it: nothing, directly. This one is a measurement error rather than a visibility problem, but it matters because it decides whether you have a real problem at all. Get a proper baseline across a spread of prompts before you spend months on the fixes above.

Be realistic about timescales. Mechanical fixes such as unblocking crawlers or correcting your own profiles take a day, but answers only reflect them after re-crawling, which takes weeks. Positioning changes take weeks to months to propagate through third-party sources. Building genuine consensus, the most common gap, is a months-long effort. And a model's trained memory only updates when a new version ships, so some stale answers persist until then no matter what you do. Anyone promising to get you recommended within days is selling something that does not exist. Measure first, then fix what the measurement shows. How to get your business recommended by ChatGPT puts the whole job in order, from checking where you stand to getting named on the websites ChatGPT reads.

1. Tighten your own entity (days). Write one clear sentence describing what you do and who for, and use it verbatim on your homepage, About page, LinkedIn and every directory listing. Add schema.org Organization markup, keep your name, location and category identical everywhere, and kill any conflicting old descriptions. This is the cheapest fix and the only cure for the ambiguity problem. A model cannot recommend an entity it cannot pin down.

2. Build reviews on the platforms your category uses (weeks). Trustpilot and Google reviews for consumer brands; G2 and Capterra for software. These sites are heavily represented in training corpora and rank well in Bing, so they work in both of ChatGPT's modes. Ask real customers systematically; a steady drip of genuine reviews beats a one-off burst.

3. Get into the listicles that already rank (weeks to months). Search your category query on Bing and note the top ten comparison articles. That shortlist is roughly what a browsing ChatGPT reads. Pitch the publishers: offer updated information, original data, or free access so they can review you properly. One placement in an article that ranks is worth more than ten guest posts nobody finds.

4. Show up in community conversation (months, ongoing). Reddit and niche forums feed both training data and browsing results, but they punish shortcuts. Astroturfed praise gets deleted and can permanently sour how your business is discussed, and models learn from that. The sustainable version is founders and staff genuinely answering questions in your niche, disclosing who they are, and being useful first.

5. Earn independent coverage with original data (highest effort, biggest compounding). Publish something citable, such as a benchmark, a piece of research or an annual report on your industry, and do the digital PR to get it referenced. Every independent article that describes you accurately strengthens the consensus a model sees, and it keeps paying off in future training runs long after the coverage goes quiet.

Whether you can list your business on ChatGPT directly, and what OpenAI offers shops and booking platforms, has a guide of its own.

Your business is not showing up on ChatGPT because the pages it reads about your kind of business do not name you, describe you differently or cannot be read at all. To find out which, or whether it names you for questions you have not tried, check if your business is on ChatGPT: Discoverable's free checker, which needs no login, puts 3 buyer questions about your category to ChatGPT in a logged-out session and shows whether each answer names you and which pages it cites.

Three answers are a sample, so to see whether your fixes move the answers, a plan checks your 10 tracked prompts weekly on every surface, ChatGPT included; see plans and prices.