Why doesn't my business show up in ChatGPT?

Your business doesn't show up in ChatGPT for one of six specific reasons: the model never learned you exist, the search indexes its live retrieval draws on have barely crawled you, your site turns away the crawler behind ChatGPT Search, your pages contain nothing an answer can quote, no third-party source corroborates you, or your identity online is inconsistent — or shared with someone else. Nothing here is punishment, and nothing is permanent. Each cause has a tell you can spot yourself, and the four-test diagnostic below takes about ten minutes to run.

What “showing up in ChatGPT” actually means

ChatGPT can know about your business through two separate routes: what the model absorbed during training (a snapshot of the public web, months old), and what its live web search retrieves the moment someone asks. The two routes read different material, age differently, and fail differently — the full mechanics are in how ChatGPT recommends businesses. When you're missing from a recommendation, it means one or both routes found too little about you to name you with confidence.

This matters more than it did a year ago, because the asking has moved:

And the ranking you already earned on Google does not carry over. Being findable by Google and being nameable by ChatGPT are different contests:

So the question isn't whether you've done marketing. It's which of the six specific failure points below applies to you.

The six real reasons, each with its tell

1. The model never learned you exist

ChatGPT's underlying model learned about businesses from its training data — the public web as it stood months before you asked. A business that's new, renamed, or thinly documented simply isn't in that memory, no matter how good the work is.

The tell: ask ChatGPT “What do you know about [your business name]?” and it hedges, answers generically, or admits it has nothing. (The deeper version of this check, with all five prompts and the three outcomes they produce, is in does ChatGPT know my business.)

The fix direction: publish consistent, factual material about the business everywhere the engines read — your own site, profiles, directories — and let the corroboration build. This route moves in months, not days. The good news: the live-search route (reasons 2–5) can start naming you much sooner, because it reads the web as it is today.

2. The search indexes behind live retrieval have barely crawled you

When ChatGPT searches mid-answer, it composes from pages a retrieval step returned. OpenAI documents the crawler for that surface — OAI-SearchBot — but not which search index or indexes sit behind it, and more than one provider may be involved. What has been measured from the outside is overlap: roughly 87% of ChatGPT's citations match Bing's top results for the same query (Seer Interactive, 2026). That is a correlation, not a pipeline anyone at OpenAI has confirmed — the distinction, and why it still changes what you do, is unpacked in how ChatGPT recommends businesses. Practically: years of Google-only SEO never checked whether Bing can see your site, and for many businesses it barely can.

The tell: search site:yourdomain.comon Bing. A handful of pages — or none — where Google shows dozens means a big share of the pages that overlap covers simply don't include yours.

The fix direction:verify the site in Bing Webmaster Tools and submit pages through IndexNow, which puts new URLs in front of Bing within minutes — Bing then decides what to index. The same index carries Microsoft Copilot's answers, which is documented rather than inferred, so the work pays twice. The Bing and Copilot guide covers the mechanics and the limits of the evidence.

3. Your site turns away the crawler behind ChatGPT Search

Every major AI provider runs separate crawlers for separate jobs, and conflating them is the most expensive mistake in this section. OAI-SearchBot is the crawler for ChatGPT Search: block it and you leave that surface. GPTBot is primarily training-data collection, and blocking it does not remove you from ChatGPT Search — which makes declining training a real, separable choice rather than a visibility decision. Anthropic splits three ways for the same reason: Claude-SearchBot (search appearance), ClaudeBot (training), Claude-User (a page a person asked Claude to open). Add PerplexityBot and Google-Extended and you have the list that matters.

The tell: your robots.txt names any of those agents under a Disallow rule — especially a blanket Disallow: /aimed at “AI bots” that swept up the search crawlers with the training ones — or your firewall's bot-blocking mode is on and has never been reviewed. This is the quietest failure of the six because software usually switched it on, not a person.

The fix direction:allow the search crawlers explicitly in robots.txt and in the CDN's bot settings, and decide about the training crawlers separately, on their own merits. It's usually a ten-minute change, and it gates everything else on this list.

4. Your pages give an answer engine nothing to quote

Answer engines compose recommendations from text that already reads like an answer: a page that says plainly what you do, for whom, where, and at what price — plus structured data that states those facts in machine-readable form. A homepage built around a slogan gives the engine nothing to lift.

The tell:no page on your site answers a real buyer question in its opening lines. If a stranger pasted your homepage into a summarizer, could it say what you sell and who it's for? If not, neither can ChatGPT.

The fix direction: answer-shaped pages aimed at specific buyer questions, and the correct schema subtype for your business — Dentist or HVACBusiness, not a generic tag. The page-shape specifics live in the ChatGPT playbook.

5. Nobody else corroborates you

Engines are cautious about naming a business on that business's own say-so. Citation studies — the analysis published by the monitoring vendor Octolens among them, its own dataset rather than independent research — put roughly nine in ten sources cited about a brand on third-party pages: review sites, directories, industry roundups, community threads. If none of those mention you, the engine has no independent reason to trust you exist, let alone recommend you.

The tell: search the roundups and comparison articles for your category, and the Reddit threads where your buyers ask for recommendations (why AI cites Reddit so heavily is its own story). You're not in any of them — but your competitors are.

The fix direction: reviews at a steady cadence, directory and profile consistency, and presence in the community conversations the engines already quote. When a rival keeps getting named, the citations behind their name are readable — reverse-engineering why AI picked your competitor shows the method.

6. Your identity is inconsistent — or shared with someone else

Engines assemble a picture of your business from every mention of the name. If your site says one name, your directories another, and your oldest profiles a third — or if a business three states away shares your name — the picture blurs. A blurred entity gets described cautiously, described wrongly, or skipped.

The tell: ChatGPT describes a different company when you ask about yours, blends two businesses into one answer, or gets your services, hours, or location wrong. Imagine a firm called Summit Injury Law — a fictional example, not a real scan — sharing its name with an unrelated Summit Law Group: every mention of either feeds one muddled entity. Wrong-answer cases have their own correction path in ChatGPT is wrong about your business.

The fix direction: pick one canonical name, category, and fact set; publish it identically everywhere; back it with Organization schema on your site. Consistency is the cheapest credibility there is.

The ten-minute self-diagnostic: four tests

You don't need software to find your failure point — you need ten minutes and the discipline to run all four tests before drawing conclusions. Swap your own details into the brackets. Where a worked example helps, we use Crestline HVAC in Tampa — a fictional business from our example library, not anyone's real scan data.

1. The existence test (about 3 minutes). In ChatGPT, without web search, paste: “What do you know about [your business name] in [your city]?” then “What does [your business name] offer, and who is it best for?” (For the example: “What do you know about Crestline HVAC in Tampa?”) Accurate and confident → the model knows you; move on. Hedged or empty → reason 1. Confident but wrong, or mixed with another company → reason 6.
2. The retrieval test (about 2 minutes). On bing.com, search site:yourdomain.com and count the results. Then, in ChatGPT with web search on, paste: “Who are the best [your category] in [your city] right now?” Note every business named and every source cited. Thin or empty Bing results → reason 2.
3. The crawler-access test (about 2 minutes). Open yourdomain.com/robots.txt in a browser. Search the text for the search crawlers first — OAI-SearchBot, Claude-SearchBot, PerplexityBot — then the training ones, GPTBot, ClaudeBot, and Google-Extended. Any search crawler under a Disallow rule, a blanket block that caught them all, or a CDN bot-protection mode nobody remembers enabling → reason 3.
4. The corroboration test (about 3 minutes). On Google, search “best [your category] in [your city]” and open the top three roundup articles. Then search [your category] [your city] site:reddit.com. Count how many times you appear anywhere. Zero → reason 5— and if your own site also never showed up among the citations in test 2's answer, reason 4 is likely riding along.

Before you trust one clean check: answers drift

One caution, offered calmly: a single manual check — pass or fail — is a weaker signal than it feels like.

The ground underneath moves too: 40–60% of the sources AI answers cite change every month (eMarketer, 2026 — the churn figure unpacked in the mechanism guide). A one-time diagnosis finds your failure point; only repeated measurement tells you whether a fix held.

None of this makes the ten-minute diagnostic pointless — it reliably finds structuralproblems like a crawler block or an empty Bing index. It means a good result isn't a finish line, and a bad one isn't a verdict.

What fixing it looks like, honestly

The fixes sort themselves by speed. Unblocking crawlers is a same-day change. Bing indexing moves in days once pages are submitted — and in Semrush-published practitioner testing (2026), the median time from a page getting indexed to its first ChatGPT citation was about 6.8 days. That figure is their observation from their data, not a schedule anyone can promise you. Quotable pages and schema are a weeks-scale project; third-party corroboration builds over months.

The step-by-step version, ordered by time-to-impact, is in how to get your business mentioned by AI. Running that loop — measuring, fixing, re-measuring — is what AEO agencies charge $2,000–6,000/month retainers to do (2026 market research), and what our own AEO service does for a flat monthly price instead. It is also work an owner can genuinely start alone, which is why the diagnostic above comes first: knowing your failure point turns a vague worry into a specific, finishable task.

Before you take our word for it

Know who benefits from this diagnosis: AEO Action wrote it, and sells the machine. The free scan is the diagnostic above, automated: it asks ChatGPT, Gemini, and Perplexity ten real buyer questions about your market — 30 live queries in about two to three minutes — and returns the three most revealing answers word-for-word with every business name highlighted, a short read of where competitors stand, and three moves plus a 30-day plan. No account, no email, no card; the PDF is permanent, and the shareable link lives seven days. The manual tests above are free forever either way — the mechanics on this page are true whether or not you ever run our scan.