Why isn't my business showing up in ChatGPT?
An 8-point diagnostic for brands invisible in AI answers — crawler access, JavaScript rendering, extractability, third-party corroboration — and how to test each one.
Why Isn’t My Business Showing Up in ChatGPT?
You typed your company’s category into ChatGPT, asked which businesses it recommends, and your competitors came back. You didn’t. The obvious next question — why? — usually gets answered with a generic list: publish more content, add schema, build authority. That advice isn’t wrong, but it skips the diagnosis and goes straight to the prescription.
Invisibility in AI answers has specific, testable causes, and they sit in four layers: whether engines can access your site, whether they can read it, whether they can extract an answer from it, and whether they trust your brand enough to name it. The layers are strictly ordered. A brand with an access problem can publish for a year and change nothing, because none of that content is ever reaching the model.
This is the diagnostic, in order, with a test for each cause and an honest estimate of how long the fix takes to show up.
Key Takeaways
- Check the diagnosis before the cure: 61.7% of AI domain appearances are “ghost citations” — your page is used as a source but your brand is never named in the answer, so you look invisible while actually being cited.
- No major AI crawler executes JavaScript. If your content only appears after scripts run, ChatGPT, Claude, and Perplexity see an empty page — this is the single most common silent failure.
GPTBotandOAI-SearchBotare different controls. Blocking GPTBot doesn’t affect ChatGPT search visibility; blocking OAI-SearchBot removes you from it entirely.- Google rankings don’t carry over: only 12% of links cited by ChatGPT, Gemini, and Copilot appear in Google’s top 10 for the same prompt.
- Unlinked brand mentions correlate with AI visibility at 0.664 versus 0.218 for backlinks — roughly 3x — so the fix for the deepest layer is being talked about, not link-building.
Start Here: You Might Already Be Showing Up
Before touching your site, rule out the possibility that the diagnosis itself is wrong. Two things make brands look invisible when they aren’t.
Ghost citations. Research from Semrush and Kevin Indig analyzed 3,981 domain appearances across 115 prompts, 14 countries, and four AI engines, and found that 61.7% were “ghost citations” — the AI used the page as a source link but never mentioned the brand name in the answer text. Only 13.2% of appearances were both cited and mentioned. A further 25.1% went the other way: the brand was named in the answer with no citation at all, which no analytics tool on earth can detect.
That changes what “showing up” means. Your content can be actively feeding the answer a buyer reads while your name never reaches them. Feeding answers anonymously is not the same problem as being absent, and it has a different fix — one about brand-name proximity to your own claims, not about crawling.
Bad sampling. AI answers are non-deterministic: the same prompt produces different answers on different runs, for different users, in different sessions. Checking once and concluding you’re invisible is like checking one keyword ranking from one browser and declaring a rankings collapse. Three habits fix this:
- Ask unbranded buyer questions (“best commercial HVAC contractor for a small warehouse in Austin”), not “what do you know about [my company]” — the second is a navigational question no buyer asks.
- Use a logged-out or temporary chat, so account memory and history don’t personalize the answer toward or away from you.
- Run each prompt several times, across engines, and record how often you appear rather than whether you appeared.
That last shift — from a yes/no anecdote to a rate — is the difference between an opinion and a measurement. If the rate is genuinely zero, keep reading.
Layer 1: Access — Can the Engines Reach You?
Cause 1: You’re blocking the crawlers (often without knowing)
The most consequential thing to check takes two minutes and is frequently a surprise, because the block is rarely a decision anyone remembers making.
OpenAI runs three distinct agents, and its documentation is explicit about the difference:
| Agent | Purpose | Effect if blocked |
|---|---|---|
GPTBot | Crawls content for model training | No effect on ChatGPT search visibility |
OAI-SearchBot | Surfaces sites in ChatGPT’s search answers | You will not appear in ChatGPT search answers |
ChatGPT-User | Fetches a page when a user’s question triggers a live visit | Blocks user-initiated retrieval of your page |
Teams that decided “we don’t want our content training AI models” and disallowed everything OpenAI-shaped removed themselves from ChatGPT’s answers as a side effect. These are independent settings — you can disallow GPTBot and allow OAI-SearchBot to appear in answers without feeding training.
The block also may not live in your robots.txt at all. In July 2025, Cloudflare became the first major infrastructure provider to block AI crawlers by default, asking every newly registered domain up front whether to permit AI scraping. If your site was set up on Cloudflare after that change and someone accepted the defaults, your marketing team and your CDN are running opposite strategies. Similar rules can hide in WAF configurations, bot-management rules, and hosting-platform toggles.
Test: open yourdomain.com/robots.txt and search for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, and any blanket Disallow: / under a wildcard agent. Then check your CDN’s bot and AI-crawler settings separately — robots.txt looking clean does not mean traffic is getting through.
Time to effect: OpenAI states robots.txt changes take roughly 24 hours to register. This is the fastest fix in the entire diagnostic.
Cause 2: Your content only exists after JavaScript runs
This is the quiet one, and it invalidates enormous amounts of otherwise good work.
Vercel’s analysis of AI crawler behavior found that none of the major AI crawlers render JavaScript — the list includes OAI-SearchBot, ChatGPT-User, GPTBot, ClaudeBot, Meta-ExternalAgent, Bytespider, and PerplexityBot. They fetch JavaScript files (ChatGPT in 11.5% of requests, Claude in 23.8%) but never execute them. Googlebot renders JavaScript; the AI crawlers do not. So a single-page app, a client-side-rendered pricing table, a tabbed FAQ that loads on click, or reviews injected by a third-party widget can be perfectly visible in Google and completely blank to every AI engine.
The same study found ChatGPT hitting 404s on 34.8% of its fetches and spending 14.4% of requests following redirects — a reminder that crawl efficiency matters too. Broken links and long redirect chains burn the budget an engine spends on you.
Test: disable JavaScript in your browser and reload your key pages, or run curl -s https://yourdomain.com/your-page | less and read what comes back. Whatever isn’t in that raw HTML doesn’t exist as far as an AI engine is concerned.
Fix: server-side render or pre-render the content that carries your answers — pricing, service descriptions, FAQs, comparison tables, reviews. This is usually an engineering ticket rather than a marketing one, which is exactly why it survives so long undetected.
Layer 2: Extraction — Can It Lift an Answer Out of Your Page?
Cause 3: The answer is buried
Most business writing builds toward its conclusion: context, then explanation, then the answer in the last paragraph. AI engines retrieve passages, not pages. A passage that only makes sense after three paragraphs of preamble is a passage that can’t be quoted.
Test: take any section of your page and read only its first two sentences. If those two sentences don’t answer the heading above them, the section isn’t extractable.
Fix: lead every section with its answer, then justify it. Make each passage self-contained — no “as we discussed above,” no pronouns pointing at content the retriever never received. Include the concrete specifics models prefer to quote: numbers, ranges, named conditions, dates.
Cause 4: You never answered the question anyone actually asks
A page can be extractable and still never surface, because it isn’t organized around a question a buyer poses. Assistant prompts average about 23 words against roughly four for a Google query, and they carry situation, constraint, and persona. “Enterprise cloud solutions” is a service page. “What does it cost to migrate a 40-person company off on-prem, and what usually goes wrong?” is a prompt.
Test: list the ten questions your sales team hears most, ask each one to two assistants, and note whether you appear. Then check whether you have a page whose heading is that question.
Fix: organize content by question rather than by topic, and confirm the questions are real before you write. Guessing at prompts is how content budgets evaporate — our guide to finding out what people actually ask AI about your industry covers the discovery methods and the validation gate to run first.
Layer 3: Trust — Does the Model Have Any Reason to Name You?
Cause 5: Nothing outside your own website mentions your brand
This is the deepest layer and the one most invisibility ultimately traces back to. A model’s confidence in a brand comes from encountering it repeatedly, across independent sources, in the right context. A site that only talks about itself gives it nothing to work with.
Ahrefs studied 75,000 brands against millions of AI Overview responses and ranked the factors correlating with brand mentions:
| Factor | Correlation with AI visibility |
|---|---|
| Unlinked web mentions | 0.664 |
| Branded anchors | 0.527 |
| Branded search volume | 0.392 |
| Domain Rating | 0.326 |
| Backlinks | 0.218 |
Unlinked mentions correlate roughly 3x more strongly than backlinks. Brands in the top quartile for web mentions averaged 169 AI Overview mentions; the next quartile down averaged 14 — a 10x gap. And 26% of the brands studied had zero AI Overview mentions at all, so being invisible is the normal condition, not a unique failure.
Where those mentions live matters too. Semrush’s study of the most-cited domains in AI answers confirms what anyone reading citation lists already notices: community platforms, reference sites, video, and editorial media dominate the sources engines pull from. Your own domain is one voice among them, and the least independent one.
Test: search your brand name and your category together, excluding your own domain. Count the independent sources — reviews, forum threads, directories, podcasts, editorial mentions, comparison write-ups — from the last twelve months. If the honest answer is “a handful,” you’ve found your bottleneck.
Fix: this is PR and participation, not link-building. Earn mentions in the communities and publications your buyers read, get listed in the directories and review platforms your category has, and answer questions publicly where your buyers ask them. The mention counts even without a link — that’s what the 0.664 figure means.
Cause 6: Your entity is ambiguous
Models resolve a brand to an entity before they can recommend it. If your company name, category description, location, and founding details differ across your site, LinkedIn, directories, and press coverage — or if your name collides with a larger brand in a different industry — the model can’t confidently decide who you are, and confident answers beat ambiguous ones.
Test: ask an assistant “who is [your company], and what do they do?” A vague, hedged, wrong, or conflated answer means the entity is unresolved.
Fix: make the description of what you do identical everywhere — same category words, same locations, same scope. Add Organization schema, and keep third-party profiles consistent with it.
Layer 4: Strategy — You’re Fighting the Wrong Battle
Cause 7: You assumed your Google rankings would carry over
They largely don’t. An Ahrefs analysis of 15,000 long-tail queries found only 12% of the links cited by ChatGPT, Gemini, and Copilot also appeared in Google’s top 10 for the same prompt (Perplexity was higher, at 28.6%). Nearly nine out of ten AI citations come from outside the page-one set. That’s why strong SEO teams are shocked to find themselves absent — the rankings are real, they just don’t transfer. This is a separate channel with a separate scoreboard, which is the case for treating AI visibility as its own investment.
Cause 8: You’re contesting prompts that already have a settled answer
Some prompts have a strong, recent, well-corroborated incumbent answer. Displacing it is expensive. Others — usually the specific, situational ones — are answered from a three-year-old blog post and a forum thread. Same effort, wildly different odds.
Test: for each prompt you care about, look at what the engine cites today. Weak, stale, or absent sources mean open territory. A well-maintained authoritative source means you’re buying an expensive fight.
Fix: rank your target prompts by demand, your current visibility, and incumbent strength, and start where the vacuum is. Which engine you prioritize is part of the same calculation.
The Diagnostic, in Order
Work top to bottom. Nothing below matters until the layer above it passes.
| # | Cause | Test | Time to effect |
|---|---|---|---|
| 1 | Crawlers blocked | robots.txt + CDN/WAF bot settings | ~24 hours |
| 2 | JS-only content | Load page with JavaScript disabled | Days–weeks |
| 3 | Answer buried | Read the first two sentences of each section | Weeks |
| 4 | Wrong questions | Ask your top 10 sales questions to two assistants | Weeks |
| 5 | No third-party mentions | Count independent mentions, last 12 months | Months |
| 6 | Ambiguous entity | ”Who is [company] and what do they do?” | Weeks–months |
| 7 | Assuming SEO transfers | Compare your page-1 rankings to actual citations | Ongoing |
| 8 | Contested prompts | Inspect cited sources per prompt | Ongoing |
Two things fall out of this table. First, the fast fixes are also the cheapest — an afternoon of access-layer work can undo a year of accidental invisibility. Second, the layer that decides whether an assistant recommends you is the slowest one, and no technical change substitutes for it. When you see claims of citations within 48 hours, they’re describing rows 1 through 3 and skipping row 5.
After You Fix It, Measure the Right Thing
The metric that tells you whether the diagnostic worked is not traffic. Given ghost citations and the 25.1% of appearances that are mentions with no link at all, a large share of your influence in AI answers will never appear in an analytics report. Track two things monthly: your visibility rate per prompt — how often you appear across repeated runs on the questions you chose — and, separately, AI-referred sessions and the revenue behind them. The first moves weeks before the second does, which makes it the leading indicator worth reporting on. Our guide to online visibility metrics covers the full measurement stack, and the playbook for getting recommended by AI engines covers the offensive side once the diagnostic is clean.
Frequently Asked Questions
Why is my website not showing up on ChatGPT?
There are four possible failure layers, and they need to be checked in order: access (your robots.txt, CDN, or firewall blocks AI crawlers), rendering (your content only exists after JavaScript runs, and no major AI crawler executes JavaScript), extraction (the answer is buried in a page organized around a topic rather than a question), and corroboration (nobody else on the web mentions your brand, so the model has no reason to trust it). Most brands assume the problem is the last one and spend months on content while an access-layer issue makes all of it invisible.
Does blocking GPTBot stop my site appearing in ChatGPT?
No — those are separate controls, and confusing them is a common self-inflicted wound. GPTBot crawls content for model training. OAI-SearchBot is the crawler that surfaces sites in ChatGPT’s search answers. OpenAI’s documentation states that sites blocking OAI-SearchBot will not be shown in ChatGPT search answers, while blocking GPTBot alone does not affect search visibility. You can legitimately disallow GPTBot and allow OAI-SearchBot if you want to appear in answers without contributing to training.
How long does it take to show up in ChatGPT?
It depends entirely on which layer you fixed. Access fixes are fastest — OpenAI states that robots.txt updates take roughly 24 hours to take effect in its systems. Rendering and content-structure fixes surface over the following crawl cycles, typically weeks. Corroboration — third-party mentions, reviews, and citations that make a model confident about your brand — is the slowest layer and works on a horizon of months. Any promise of citations within 48 hours is describing the easiest layer and quietly ignoring the one that actually decides recommendations.
Does ChatGPT use Google rankings?
Not directly, and the overlap is much smaller than most teams assume. An Ahrefs analysis of 15,000 long-tail queries found that only 12% of the links cited by ChatGPT, Gemini, and Copilot appeared in Google’s top 10 for the same prompt. Ranking well on Google is helpful but is neither necessary nor sufficient for being cited in an AI answer — which is why a brand can hold page-one rankings and still be entirely absent from assistants.
How do I check whether AI mentions my business?
Ask a buyer’s unbranded question — not your company name — in a logged-out or temporary chat, and repeat it several times, because AI answers are non-deterministic and a single sample proves nothing. Check the cited source links as well as the answer text: 61.7% of AI domain appearances are citations with no brand mention in the answer, so your site can be feeding an answer that never names you. Tracking the same prompts over time across engines is what turns those anecdotes into a measurable visibility rate.
Find out where you stand — free. Run your site through the free AI visibility checker to see how you appear across ChatGPT, Gemini, Perplexity, and more in seconds, no credit card required. When you’re ready to track your visibility rate prompt by prompt, start a 7-day free trial.
