Why your business is not showing in ChatGPT, and how to diagnose it

How to run the access test step by step

The existing diagnosis starts with crawler access because it is quick and decisive. Here is the exact routine we follow, so you can repeat it on your own site in under an hour.

1. Open robots.txt. Type your domain followed by /robots.txt. Look for any line that disallows everything for all agents, and for named rules about OAI-SearchBot, GPTBot and ChatGPT-User.
2. Read the order of rules. A specific rule for a named bot overrides the general one for that bot. A general block with no named exception stops every bot that does not have its own group.
3. Check your security layer. Open your CDN, firewall or security plugin settings. Look for bot fight modes, country blocks, rate limits and managed rules that treat unknown bots as threats.
4. Test a real request. From a command line, request a key page while sending the bot's user agent string, and note the status code. A 200 is good. A 403 or 429 means a layer is refusing.
5. Read your server logs. Search for the bot names over the last month. If they appear with 200 responses, access works. If they never appear, either nothing is attempting to crawl you or something is stopping them before the log.
6. Check indexability on the same pages. Look for noindex tags, canonical tags pointing elsewhere and redirects.

Our free AI crawler access checker automates the first checks, and the guide to AI crawlers and robots.txt explains what each bot does. Record the date of your test. Settings change, and an old result is not evidence.

What should you do when the symptom is wrong details, not absence?

Being named wrongly is a different problem from being missing, and it is often easier to fix. The engine knows you, but one of its sources is off. Follow the trail.

1. Copy the exact wrong sentence from the saved answer.
2. Search for that detail on the open web: an old address, a former service, a merged brand name.
3. List every page that carries the wrong version. Common homes are directories, old press pages, social bios and cached copies on partner sites.
4. Correct the ones you control. Ask for edits on the ones you do not.
5. Publish a clear, dated statement of the right fact on your own site, on a page that is easy to find and crawl.
6. Rerun the prompt after a few weeks and see whether the answer has moved.

Our guide to fixing wrong information about your brand in AI goes deeper, including what to do when the source is a page you cannot edit.

Which cause applies to your kind of business?

The seven causes appear in every sector, but their weight differs. This table shows where we would look first, based on patterns in our own work, not a measured survey.

A worked example, hypothetical: a Dubai property consultancy

Example, hypothetical: a property consultancy in Dubai tests 15 prompts and finds it is never named, while three competitors appear often. This shows the order of diagnosis, not a real client.

• Hour one. It asks ChatGPT about its company name directly. The answer is vague, so the engine knows little. It checks robots.txt, which is open, but the firewall returns errors to unknown bots. Cause 1 is confirmed and fixed.
• Day one. It reads its own pages. They list services and awards, but no page answers what a first time investor should check before buying. Cause 3 applies.
• Day two. It compares facts. A former office address still appears on two directories. Cause 4 applies and is corrected.
• Week two. It reads the sources cited in the competitor answers. They are two portals and a review site where the firm has no listing. Cause 5 applies and a listing plan starts.
• Month three. It reruns the same 15 prompts and compares the text.

Four causes were present at once. That is common. Fixing only the loudest one often leaves the rest in place, which is why a full order of checks beats a single quick fix.

How long should you wait before deciding a fix did not work?

Patience has to be balanced against drift. There is no published timeline, so use a plan instead of a guess.

These windows are our working judgement, not documented by any engine. Use the same prompts, the same settings and a similar day of the week, so the comparison is fair.

Things that look like causes but are not

• Your logo or design. Visual style does not decide whether text is retrieved.
• A missing social media following. Social presence can contribute mentions, but its absence alone does not hide a well sourced page.
• Page speed on its own. Extremely slow pages can time out, but ordinary speed differences are not a known reason for exclusion.
• The lack of an llms.txt file. No major engine has confirmed it as a factor. Read our view on llms.txt before spending time on it.

When to get outside help

Do the first test yourself. If you find a block, fix it. If you find several causes at once, or you cannot change your site's rules, templates or markup, help saves time. Our GEO Pilot runs a baseline across five AI engines, checks crawler access, implements the technical fixes, rewrites up to 10 priority pages and retests, for a fixed $999 delivered inside 30 days. We do not guarantee that your business will appear, because nobody controls that, but we do give you a before and after on the same prompts. You can also start with the free AI visibility checker.

Start by confirming that you are truly missing

Before you fix anything, test properly. One prompt typed once is not a diagnosis. ChatGPT may answer the same question differently on different days, with search on or off, and for different users and places.

1. Write 15 prompts a buyer would use. Include three that name your category and city, three that compare options, and three that describe a problem without naming a product.
2. Run each twice, on two different days, with search enabled.
3. Record whether your brand is named, whether your site is linked, and who is named instead.
4. Also ask one direct question: what do you know about your company name? This shows whether the engine knows you at all.

You now have a map. If you appear in some prompts and not others, you have a coverage problem. If you never appear, even when asked by name, you have an awareness or access problem. Those need different fixes.

Match the symptom to the likely cause

Cause 1: crawlers are blocked

OpenAI documents OAI-SearchBot as the crawler used for ChatGPT search, GPTBot for training and ChatGPT-User for actions a user requests. If your robots.txt, your CDN or your firewall blocks them, retrieval cannot reach your pages.

• Read robots.txt for wildcard blocks and bot specific rules.
• Check your CDN and security plugin settings for bot blocking or rate limits.
• Test a real request with the bot user agent and look for 403 or 429 responses.

This is the one cause that can be fixed in an afternoon and removes the problem entirely. Start here. Our technical SEO work always begins with this check.

Cause 2: the page cannot be read or indexed

Even with open access, a page may hide its content. Common culprits are a noindex tag, a wrong canonical, a redirect chain, a login wall, or text that only appears after scripts run.

Open the page source and search for your main paragraph. If it is not there, many crawlers will not see it. Also check Search Console for the page's index status. Google says its AI features need pages that are indexed and snippet eligible, so index problems limit more than one engine.

Cause 3: your pages describe services, not buyer questions

Many business sites say what the company offers but never answer what a buyer asks. A model searching for the cost of a service or a comparison finds nothing to quote on your site, so it quotes someone else.

• List the ten questions your sales team hears most often.
• Give each one a heading and a direct first sentence answer.
• Add the detail that makes the answer trustworthy: ranges with dates, steps, limits.

The panel is a sample content brief layout. A good brief starts from real prompts and sales calls, not keyword lists alone.

Cause 4: your facts disagree across the web

If your site says one thing, your Google Business Profile says another and an old directory says a third, an engine has no way to know which is right. It may drop you, mix details or pick an outdated version.

Make a fact sheet: name, description, services, locations, phone, hours, founding date if you publish it. Compare it against your top ten profiles. Correct each difference. Details such as a new phone number or a closed branch are the typical offenders.

Cause 5: the sources engines trust do not mention you

When ChatGPT searches, it often cites review platforms, directories, news and trade sites, forums and comparison pages. Look at the links shown in your test answers. If every cited source is a site where you have no listing, that is your gap.

Sample data again, this time showing how a brand compares with named competitors. Build your own source list from your saved answers. Then work through it honestly: complete the listing, ask real customers for reviews where the platform allows it, and contribute something useful. Do not buy reviews or mass produce profiles.

Cause 6: the brand is small, new or local

Language models learn from a vast amount of text. A brand with little text about it, or a very new one, may simply not be known. Search can compensate, because it retrieves fresh pages, but a business with almost no web footprint still has little to retrieve.

This cause is real and we say so plainly. Nobody can make a four month old company appear as prominent as an established one with a contract. What you can do is build clear owned pages, a few trusted profiles and a steady record of proof. Narrow questions are the realistic target: a specific service in a specific city, not the whole category.

Cause 7: answer variation, not a real problem

Sometimes nothing is wrong. Answers differ by user, location, time and model version. A brand can appear in one run and not in the next with no change on your side. If you appear in around half of your repeated runs, chasing the missing half may waste effort better spent elsewhere.

Special case: UAE buyers and bilingual answers

In the UAE, the same buyer may ask in English at work and in Arabic on a phone. An engine can name different companies in each language. Run your prompt test in both, and note which language gives you the weaker result. If Arabic is weaker, the likely causes are the same ones in this guide, but with an extra check: do your Arabic pages exist, are they indexed, and were they reviewed by a native speaker?

Also include place in your prompts. Questions that name Dubai, Abu Dhabi or Sharjah often pull in local directories and map data, so your Google Business Profile and listing details matter more there. Our guide on how UAE buyers use AI search goes deeper.

A diagnostic order that saves time

1. Run the prompt test and save the answers with dates.
2. Check crawler access and indexability on your top five pages.
3. Compare your facts across your top ten profiles and fix differences.
4. Rewrite your three most important pages around real buyer questions.
5. Audit the sources cited in your answers and close the biggest gaps.
6. Repeat the prompt test after a few months and compare like with like.

Step three and four often overlap with what you already planned for ordinary SEO, so the extra cost is small. If you want the whole diagnosis done for you, our free LLM visibility audit covers prompts, access, facts and sources, and the GEO service covers the fixes.

When it is not worth chasing

Be selective. If your buyers rarely ask AI tools for vendors, or you win mostly through referrals and repeat work, ChatGPT presence may rank below faster channels. A quick reply system, clean local listings and sound SEO might return more this quarter. We would rather tell you that than sell you a retainer you do not need.