Why isn't my business showing up in ChatGPT?

Before you change anything: there are three gates, not one

Almost every conversation about this goes wrong in the same place. The owner assumes there is one switch, and that flipping it puts the business in ChatGPT. There are three separate gates, and clearing one does nothing for the next.

Gate one is reach. Can the crawler fetch the page at all? Gate two is retrieval. When somebody asks the question, does your page come back among the results the assistant reads? Gate three is citation. Out of everything it read, does it quote you?

The six reasons below are gates one and two, which are fixable and checkable. Gate three is nobody's to promise, and OpenAI says so in its own documentation: the web search guide describes sources as "the complete list of URLs the model consulted" while inline citations show "only the most relevant references", and notes that "the number of sources is often greater than the number of citations." Being read is not being quoted. Any provider who tells you otherwise is selling gate one and calling it gate three.

Reason 1: your robots.txt blocks the crawler that feeds ChatGPT

This is the first thing to check because it is free, it takes thirty seconds, and it is the one people get backwards most often. Open yourdomain.ca/robots.txt in a browser.

The trap is that OpenAI runs several crawlers and they do different jobs. Blocking GPTBot is the popular advice, and it stops your content being used as training data. It does not remove you from ChatGPT's answers. The crawler that decides that is OAI-SearchBot, and OpenAI's bots and crawlers page is explicit: sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers." So a site can block GPTBot, think it has protected itself, and stay perfectly visible; or block OAI-SearchBot on a designer's default and vanish from ChatGPT while every Google ranking stays exactly where it was.

One more thing worth knowing, because it explains confusing results: a third crawler, ChatGPT-User, fires when a live person asks a question that needs your page fetched, and OpenAI states that "because these actions are initiated by a user, robots.txt rules may not apply." So someone can paste your URL into ChatGPT and get a perfect summary while your business never appears in an unprompted answer. OpenAI says plainly that ChatGPT-User "is not used to determine whether content may appear in Search." Testing by pasting your own link proves nothing.

Reason 2: your firewall turns the crawler away even though robots.txt invites it

Permission is not access, and this is the reason no ranking tool will ever show you. Your robots.txt can say yes while a bot-protection rule, a WAF or a challenge page says no. The crawler gets a 403 and moves on.

This is not our inference. It is the one sentence in which OpenAI states what makes a site eligible for ChatGPT search at all, and it names two requirements rather than one:

"To make a website eligible for inclusion, allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses."

OpenAI, Searching the web with ChatGPT, read 19 September 2026

Reason one is the first half of that sentence. This is the second half, and it is the half almost nobody checks. Google's guidance on AI features says the same thing in its own words, putting allowing crawling "in robots.txt, and by any CDN or hosting infrastructure" at the top of its list.

There is a sharper version of this on Cloudflare, and it is worth spelling out because the file you wrote is not necessarily the file the crawler reads. A setting called managed robots.txt can insert Cloudflare's own block above your origin's file. We hit this on a client site in September 2026: while the domain was still moving onto Cloudflare the dashboard reported the feature off, and the moment the zone went live the served file carried a Disallow: / for nine AI crawlers, stacked on top of a perfectly good allow-all file sitting in the repository. Nothing in the project had changed. That is our own finding on our own client's domain, not a vendor statement, and the lesson from it is one line: read the live file over the internet, never the one in your project.

The test is one command per crawler, from any computer with a terminal:

curl -sS -o /dev/null -w "%{http_code} %{size_download}\n" -A "OAI-SearchBot/1.4" https://yourdomain.ca/

Run it once with no -A flag for the browser's numbers, then once per crawler name. You want the same status code and roughly the same byte count every time. A different status, or a body a fraction of the size, means that assistant is being turned away at the door.

Reason 3: Bing has never indexed your site

You will read everywhere that ChatGPT runs on Bing, so getting into Bing is how you get into ChatGPT. That was true when ChatGPT search launched and it is not what OpenAI documents now. We rendered OpenAI's ChatGPT search page on 19 September 2026 and counted: it mentions Bing zero times. It says only that ChatGPT search "sometimes partners with other search providers", and names none of them. We are flagging that because most of the advice you will find on this question is repeating a fact that has quietly expired.

Bing still belongs on this list, just for honest reasons rather than that one. Microsoft Copilot grounds its answers on Bing, so a site Bing has never seen is invisible there. Bing's published guidelines are the most specific either search engine gives about what breaks AI grounding. And IndexNow gets a new page in front of Bing in minutes rather than weeks. Being absent from Bing costs you real visibility; it just is not the ChatGPT switch it is usually sold as.

The test takes ten seconds: search Bing, not Google, for site:yourdomain.ca. If it returns nothing, that is your answer and the fix is registration plus a sitemap, which is free and takes an afternoon. We wrote that up separately in why your website needs Bing to appear in ChatGPT, including IndexNow, which pushes new pages to Bing in minutes instead of waiting to be crawled.

Reason 4: a noarchive or nosnippet tag is quietly removing you

These two are the cruellest fault on the list, because every ranking tool stays green while they are on. A page carrying noarchive or nosnippet, in a meta tag or an HTTP header, still ranks, still gets crawled, still reports healthy, and is removed from Copilot citations and from Google's AI answers. The robots meta tag reference documents what each one does. A related attribute, data-nosnippet, does the same job on one block of text and is even easier to leave behind.

They are usually accidental: a staging setting that shipped, a plugin default, a developer protecting a page in 2019 that nobody has looked at since. The test is to open the page source and search for both words. The longer version is in the two settings that hide a website from AI answers.

Reason 5: your pages are empty without JavaScript

If your site is a modern single-page app, there is a real chance the crawler receives an empty shell and a pile of script tags where your sentences should be. Right-click, View Source, and look for your own words. If they are not in there, they are being painted in after the page loads.

Bing names this outright in its Webmaster Guidelines: "hiding critical content behind client-side rendering" is on the list of things to avoid, and "content that cannot be reliably rendered may not be indexed or selected for grounding results." Google is gentler, because Googlebot does render JavaScript on a second visit, but its JavaScript SEO basics still recommends server-side or pre-rendering because "not all bots can run JavaScript."

Here is the part that matters for ChatGPT specifically, and we are labelling it as reasoning rather than a quote, because no vendor has published an answer either way: not one AI company documents whether its crawler renders JavaScript at all. OpenAI's bots page says nothing about rendering, timeouts or page-size limits. An undocumented capability is not one to build a business on, so the safe reading is that your words need to be in the HTML on the first response.

Two size limits are worth knowing while you are in there. Google publishes that Googlebot crawls the first 2MB of a file, that the limit "is applied on the uncompressed data", and that past the cutoff it "stops the fetch and only sends the already downloaded part of the file for indexing consideration." Every CSS and JavaScript file gets its own 2MB ceiling too, so one oversized bundle can break a render on a page that looks perfect in your browser.

Reason 6: nothing you have published answers the question being asked

This is the only reason on the list that is not a bug, and it is the one most businesses are actually failing. Once the first five pass, the assistant can read every word you have. The question is whether any of those words answer what somebody typed.

Most small-business websites are written as a brochure: a homepage that says what the company values, a services page that lists categories, a contact page. None of that answers "how much does X cost in Hamilton", "do I need a permit for Y", or "what is the difference between A and B". An assistant answering a question needs a page that answers that question, in the words a person types, in text it can lift in one piece. Brochure copy gets crawled and never quoted, which looks identical from the outside to being blocked.

The fix is the slowest of the six and the only one with a lasting advantage in it, because the other five are things your competitors can fix in a morning too. It is writing real answers to the real questions, one page per question, with the answer at the top rather than buried under a story.

The six reasons, and how to test each one yourself

Ordering is by mechanism, cheapest to check and most total in effect first, not by how often we have observed each one. Sources: OpenAI, Bots and crawlers; Google, AI features and your website; Bing Webmaster Guidelines; Google, robots meta tag; Google, Googlebot.

Can I just have all six checked at once?

Yes, and free. We built the same six tests into a checker that fetches your homepage the way a browser does and the way OAI-SearchBot, ClaudeBot, PerplexityBot and bingbot do, reads your robots.txt one crawler at a time, looks for the silent tags, and tells you whether your text is readable as plain text. It takes about ten seconds, there is no account and no call, and if it comes back clean we tell you that too. Run the free check.

What if all six pass and ChatGPT still does not mention my business?

Then you have cleared gates one and two, and you are in gate three, which nobody controls. This is the honest part that most of this industry skips, so here it is plainly: we will not promise you a citation, and neither can anyone else. Bing says it in writing in its own guidelines: "SEO does not guarantee rankings or traffic, and GEO does not guarantee grounding or citations in AI experiences." OpenAI says it a different way, in the gap between the sources it consults and the ones it names.

What is left after that is not nothing. It is being the page most worth quoting on the question you care about, and being reachable, readable and in the index so that you are in the running at all. That is a real job with a real ceiling, and stating the ceiling first is the difference between a supplier and a salesman.

How would I know if any of this actually worked?

Until recently the honest answer was that you could not tell, and we said so. That changed. Microsoft Clarity now reports AI citations free: which of your pages were referenced in AI answers, with citation counts, and the queries that retrieved them. It also shows the grounding queries, which Microsoft notes "may differ from the exact phrases users type" and which are the most useful signal on the page, because if a page is being retrieved for the wrong question the fix is almost always its heading and its opening paragraph.

Two caveats to carry, both Microsoft's own. Citation counts "reflect how often a page was referenced, not its ranking or prominence within an AI-generated answer." And the AI referral traffic figure counts ChatGPT, Copilot, Gemini, Claude and Perplexity used directly, so Google AI Overviews traffic is not in it. The number is a floor, not a total.

Turn it on before you change anything. Microsoft does not backfill, so a site measured after the work has no before to compare against.

Where to go next

Each of the six has a longer write-up where it earns one. Start with the five-minute check if you want to run reasons one through five yourself today, the Bing article if reason three is yours, and the noarchive and nosnippet article if reason four is. If the whole idea is new, what AI visibility actually means is the place to start, and your Google Business Profile covers the part of this that is not on your website at all.