Retrieval: can AI systems reach you at all
The cheapest failure to fix and the most frequently missed, because the site looks perfect to the person checking it.
What retrieval means
Retrieval is the first of the four things an AI visibility evaluation resolves: whether AI systems can reach your pages, read them, and extract meaning from them. Everything downstream depends on it. A business that fails retrieval cannot be described correctly, cannot be corroborated, and cannot be selected, because nothing was ever read.
It is also the failure most likely to be invisible internally, because a browser with JavaScript, cookies and a human behind it sees a working site. The systems doing the reading arrive with none of those.
The five common causes
Blocked crawlers. Many sites block AI user agents in robots.txt, sometimes deliberately, frequently because a directive was copied from a template or added by a plugin nobody audited. This is the single most common cause and it takes minutes to check.
Substance rendered only in script. If the meaningful content arrives after client-side rendering, some systems will see an empty shell. The page works perfectly for a person and is blank to a reader that does not execute the script.
Content behind interaction. Accordions, tabs and modals that load their contents on click. A person clicks. A retrieval system usually does not.
Bot mitigation. Security products that challenge unfamiliar agents. Legitimate protection that also excludes the systems you want reading you, and it is configured by someone who was solving a different problem.
Slow or unstable responses. A retrieval attempt that times out is a page that does not exist, and the failure leaves no trace anyone looks at.
The check anyone can run in a minute: fetch your most important page with scripting disabled. What remains is close to what a retrieval system gets. If your value proposition is not in it, you have a retrieval problem regardless of how the site looks.
Why blocking is a real decision, not always an error
Some businesses block AI crawlers deliberately, and there are honest reasons: protecting original work from being reproduced without attribution, or a legal position on training data. That is a legitimate choice and it has a cost, and the cost is that the business cannot be recommended by systems it excluded.
What an evaluation should never do is treat a deliberate block as a mistake. What it should do is state plainly that it is in place, so the decision gets made on purpose rather than discovered eighteen months later.
Retrieval is not the same as indexing
Being in a search index and being retrievable by an AI system are different conditions with different mechanisms. A business can be well indexed and poorly retrievable, and the reverse also happens. Assuming search visibility implies AI retrievability is one of the most common reasons a business is confident and invisible at the same time.