GUIDE · PART OF AI VISIBILITY TRACKING
AI visibility diagnosis:
why AI isn't naming you.
Finding out you're missing from an answer is the easy part. The useful part is working out why, because the cause decides the fix, and there are only four causes. This guide gives you the order to check them in, and what each one looks like.
What is AI visibility diagnosis?
AI visibility diagnosis is working out why an AI engine left a brand out of an answer, or described it wrongly. There are four usual causes. The engine's crawler couldn't read the page. The page didn't answer the question in a form the engine could use. The sources the engine trusts on that subject don't mention the brand. Or the engine isn't sure what the brand is. Each has a different fix, which is why the diagnosis has to come first.
Key facts
- There are four usual causes: access, content, sources and identity
- Check them in that order. The early ones are quicker to rule out, and they block the later ones
- A page that is consulted and never cited points to a content problem
- If the cited sources all mention rivals and leave you out, the problem is the sources
- If engines disagree about what kind of company you are, the problem is identity
- Diagnose prompt by prompt. Two lost questions can have two different causes
Last reviewed 15 October 2026 by the SEORCE team.
Guessing the cause is how months get wasted.
The natural reaction to a missing mention is to write more content. Sometimes that's right. Often the page was fine and something else was in the way.
A rewrite won't help a blocked page
If the crawler can't reach the page, nothing you do to the words will matter.
One prompt, one cause
A site that seems invisible to AI is usually missing on particular questions, each for its own reason.
The answer doesn't say why
An engine names the brands it chose. It never tells you why you weren't one of them.
Four causes, checked in order.
Work down the list. Stop at the first one that fits, fix it, and measure again before going further.
The engine couldn't read the page
Rule this out before anything else, because it takes minutes to check and it makes every other fix pointless if it's the problem.
- Is the search bot allowed?Each AI company runs separate bots. Check the one that builds the search index.
- Is the content in the HTML?Many AI crawlers don't run JavaScript.
- Does the page appear among the sources at all?If it's never even consulted, suspect access first.
The page didn't answer the question
This is the commonest cause among pages that can be read. The engine found your page, looked for a usable answer and took one from somewhere else.
- Consulted, never citedThe clearest sign. The page was read and not used.
- The answer is buriedIt's there, in the fifth paragraph or inside a table.
- A rival says it more plainlyCompare your page with the one that was cited.
The trusted sources leave you out
An engine leans on reviews, directories and forums as well as brand sites. If those leave you out, your own page can be perfect and you'll still be missing.
- Third-party sources name rivals onlyA directory or roundup that lists everyone else.
- You have few links from sources engines citeLink Impact Score shows which linking domains carry Citation Potential.
- A rival is named on one engine onlyUsually a single source is doing the work there.
The engine is unsure what you are
The slowest cause to fix, and the easiest to miss. If an engine has you down as the wrong kind of company, it won't consider you for the right questions.
- The wrong categoryYou're described as something you aren't, or used to be.
- Engines disagreeOne has you right and another doesn't.
- Confusion with another nameA similar brand, or an old one of your own.
One lost prompt, diagnosed.
Take a single question you care about and work through it. The habit matters more than the tool.
Pick the prompt
One buyer question where you're missing across repeated runs.
AI mention tracking →Check access
Can the search bots read the page you'd expect to be used?
Free browser extension →Compare with the winner
Put your page beside the one that was cited. Look at where each states the answer.
Prompt diagnosis →Look at the other sources
Which third-party pages were cited, and are you in them?
AI citation tracking →Check your identity
Ask each engine what your company does. Then fix the first cause that fits, and re-run.
Rule out access first.
The first cause on the list is the quickest to test. Check that your robots.txt isn't blocking pages in your sitemap, then open the page with the extension to see what its raw HTML holds.
From symptom to cause to fix.
| What you see | Likely cause | What to do |
|---|---|---|
| Your page never appears among the sources | Access | Check bot rules and whether the content is in the raw HTML. |
| Your page is consulted and never cited | Content | State the answer plainly, near the top. |
| A rival's guide is cited every time | Content | You have no page that answers this. Write one, or improve the nearest. |
| Cited third-party pages list rivals only | Sources | Get listed or mentioned in those sources. |
| You're described as the wrong kind of company | Identity | Correct your own pages and profiles first. |
| Named on one engine and not the others | Sources | Compare what each engine cites. |
Your page, beside the ones that won.
Prompt diagnosis in AI Beacon sets your page next to the sources an engine used for the same question, and says what would need to change. Technical Suite checks access for each bot, so you can rule that out in the same project.
Frequently asked questions.
There are four usual reasons. ChatGPT's search bot can't read your page, the page doesn't answer the question in a usable form, the sources it trusts on the subject don't mention you, or it has the wrong idea of what your company is. Check them in that order.
Ranking and being cited are related and they aren't the same. A page can rank well and still bury its answer, or load it by script. Compare it with the page that was cited.
Open the page with the free Browser extension, which shows whether AI crawlers are allowed in. Then check that the content you care about is in the raw HTML.
No. Diagnose one prompt at a time. Different questions are lost for different reasons, and a site-wide verdict hides that.
Weeks more often than days, and it differs by engine and by cause. Access and content fixes tend to show sooner. Identity is the slowest.
It shows you the evidence: your page beside the sources that won, whether each bot could read it, and how each engine describes you. That is usually enough to name the cause, and the decision about the fix stays with you.
You can rule out the first cause for free. The robots.txt and sitemap conflict checker shows pages your rules block, the AI crawler log analyser shows which bots visit, and the browser extension shows what the raw HTML contains.
Find out why you were left out.
Add your site and one buyer question you're losing. You'll see the sources that won it, set beside your own page.
Not ready for an account? Use the free browser extension or any of the free tools.