This article is part of our comparison of the best AI visibility tools.
Your site ranks. Your paid campaigns run. Your team ships pages every month. Then someone asks ChatGPT for the best options in your category, and the answer names three competitors and not you.
Nobody on the team can say why. Your SEO tools report positions in Google, not what an AI engine reads when it builds an answer. And without an SEO in house, the question lands on the marketing director's desk with no obvious owner.
Here is the good news: AI invisibility is rarely a mystery. It comes from a short list of causes, most of them on your own site, and each one can be checked. This article goes through the six we see most often, how to spot each one, and what to fix first.
Why does a good website get skipped by AI engines?
Because an AI engine does not read your site the way your visitors do, nor the way Google's ranking does.
To cite a page, the engine has to reach it, read it, understand what you sell and to whom, and find in it an answer to the question it was asked. Then it checks the wider web: do other sites name you for the same thing? A break at any step and the engine moves on to a competitor whose page passes all of them.
The first five causes below sit on your site. The sixth sits on other people's sites. Most brands that stay invisible have one problem on each side.
Reason #1: AI crawlers cannot reach your pages
The most common cause is also the most frustrating, because the page can be excellent.
If your robots.txt blocks OAI-SearchBot, ChatGPT's search crawler, ChatGPT cannot quote your page. If your firewall or bot protection turns AI crawlers away, they get an error where your visitors get your offer. Many brands set these rules years ago, or inherited them from a security plugin, and nobody has looked since.
How to spot it: open your robots.txt and look for rules on the AI crawlers. Then check whether your CDN or firewall has a bot protection mode switched on. The page works in your browser; the question is what an AI crawler gets.
The fix: allow the AI search crawlers you want to be cited by, and make sure your bot protection lets them through. Nothing else in this article matters until this is done. Once crawlers get in, a short llms.txt file can point them to your most useful pages: our guide on how to create an llms.txt file walks through it.
Reason #2: Your key content only exists after JavaScript
Your prices, your product names, your comparison tables may only appear once the page's scripts have run. A browser runs them. Many AI crawlers do not, or not fully.
The result: the crawler reads a page with a header, a footer and very little in between. Your offer is there for humans and missing for machines.
How to spot it: compare what your key pages say with JavaScript turned off and once fully rendered. Headings, prices and structured data that only exist after the script runs are the ones at risk.
The fix: render the important content on the server, or at least the parts an AI needs to understand the page: what it is, what it costs, who it is for. This is a developer task, but a short one once you know which pages are affected.
Reason #3: The AI cannot tell what you sell, to whom and where
"The platform that powers growth." "Solutions for modern teams." Your brand team signed off on it. An AI engine trying to match your site to "best invoicing software for construction companies in France" finds nothing to hold on to.
AI engines look for plain facts: what the offer is, for which buyers, in which countries, at what price or price logic, with what proof. When a page never states them, the engine cannot recommend you for a question that hinges on them. A missing "where" alone can keep you out of every question that names a country or a city.
How to spot it: read your main offer page as a stranger. Could you write one line with who, what, for whom and where, using only what the page says? If not, an AI engine cannot either.
The fix: say it, in words, on the pages that sell. Then back it with structured data (JSON-LD) that matches the visible text: the type of the page, the organization, the product or service, an FAQ where the page has questions and answers.
Reason #4: You have no page for the questions buyers ask
Your site describes your offer. Your buyers ask AI engines something else: "best X for Y", "X versus Y", "how to choose an X", "how much does X cost". Those questions are answered from comparisons, buying guides and best-of lists. If you have none, the AI quotes someone else's, usually one that names your competitors.
How to spot it: list the ten questions your buyers would ask an AI before shortlisting you. For each one, do you have a page that answers it directly? Most brands find two or three.
The fix: create the purchase content that is missing, starting with the questions that matter most for your business, and link it to your offer pages. This is where most of the writing effort goes, and where a team without an SEO usually stalls.
Reason #5: Your site contradicts itself
One page says the starting price is X, another says Y. The brand name is spelled two ways. Three pages target the same question, so the engine picks one, and not necessarily the one you wanted. An old page still describes a product you stopped selling.
AI engines aggregate what they read. When your own pages disagree, the engine's confidence drops, and it prefers a source that gives one clear answer.
How to spot it: search your site for your prices and your product names, and list the pages that compete for your main questions. Inconsistencies tend to hide on old pages nobody owns any more.
The fix: pick one version of each fact and apply it everywhere. Merge or redirect pages that compete for the same question. Keep your public profiles (directories, review platforms, your Google Business Profile) aligned with the site.
Reason #6: Nobody else talks about you
Even a perfect site is your own word about yourself. Before recommending a brand, AI engines read what others say: comparisons, rankings, review sites, trade media. If those pages name your competitors and not you, the answer follows them.
How to spot it: look at the sources cited under AI answers on your main questions. Count the pages that name a competitor and not you.
The fix: get mentioned on those exact pages. That is the Off-site side of AI visibility, covered in detail in AI mentions are the new backlinks.
How do you diagnose all this without an SEO on the team?
You can work through the six reasons by hand once. Doing it every month, on every page that sells, while your site keeps changing, is another story. That is what the On-site side of Mentionable does.
Start from where you stand. Mentionable asks your buyers' questions every day to the AI engines you choose among ChatGPT, Claude, Perplexity, Gemini, Google AI Mode and Google AI Overview (up to 3 on Presence, 4 on Influence, all of them on Reference), and shows where your brand is cited, by question and by country, next to the competitors cited instead. That tells you which pages should be cited and are not.
Let the crawl find the problems. Site health crawls your site every month at no extra cost, up to your plan's page limit, and sorts what it finds into four levels: Blocking, AI impact, Consistency and SEO hygiene. Each crawl is compared with the previous one, so you see what was fixed and what is new.
Fix page by page. On any page that sells, a page fix reads the page live as a browser, as OAI-SearchBot and fully rendered, shows what an AI understands from it (who, what, for whom, where), and lists every fix by severity with the JSON-LD ready to paste. Fixes are free on every plan, with no monthly limit. Mentionable does not rewrite your sales copy: your offer pages stay in your words.
Follow one plan. The Action plan puts everything in one ordered list over 6 months: blocking fixes, pages that sell, pages to create or rewrite, sources to get. Pages to create arrive drafted from the Google top 3 and the pages AI engines cite, in your brand voice, with the facts only you know highlighted for your team to complete. Presence includes 4 articles per month, Influence 8, Reference 15.
Your team reviews, publishes and applies the fixes (or hands them to a developer). Mentionable never publishes anything on your site.
How long before AI engines notice the changes?
Not overnight, and not at the same speed everywhere.
Mentionable shows the usual delay next to each engine: a few days for Perplexity, one to three weeks for ChatGPT, four to eight weeks for Google AI Overview. Judge a fix too early and you will conclude it did nothing.
Mentionable closes the loop for you. At the next crawl, each fix is checked again, and a fully applied page is checked off in your plan on its own. Then the Progress page compares AI answers on the page's questions four weeks before and after, against the questions you did not touch, so the general movement of the market does not pass for the effect of your work. That is the number your leadership will ask for.
Where to start
Check your robots.txt and your bot protection today. That one is free and takes ten minutes.
For the rest, the 7-day trial needs no credit card. Enter your URL, keep the questions that matter, and within a few days you see where your brand is cited, what the first crawl found on your site, and the Action plan that goes with it.
If your team would rather delegate the whole site project, Mentionable Agency's GEO service goes further than the tool: its team runs the technical audit, the content strategy and the writing with its own methods, and the tool measures your visibility in AI answers before and after.
Related articles
- Your Competitors Are Cited by AI Instead of You: how to see it and what to do about it.
- AI Mentions Are the New Backlinks: the Off-site side, getting cited on the sites AI engines read.
- AI Visibility for Multi-Location Brands: what changes when every city is its own ranking.
