Why AIs sometimes ignore your site
Blocked crawlers, generic content, lack of evidence, a brand barely mentioned elsewhere: the most common causes of invisibility in AI answers.
A site can be perfectly well-ranked on Google and yet be completely absent from answers generated by ChatGPT, Perplexity, or Gemini. This is almost never random: there's a limited set of recurring causes, often compounding one another, that explain why an AI ignores a site that's otherwise legitimate on its topic. The good news is that these causes are, for the most part, identifiable and fixable, once you know precisely where to look rather than simply noting the absence of results.
The crawler simply doesn't have access to the content
The most basic, and yet common, cause is a technical block. Some sites explicitly block, via their robots.txt file, the crawlers used by generative AIs to index content, sometimes out of unfamiliarity, sometimes as a deliberate choice. Others make their content inaccessible unintentionally, for instance by rendering it entirely client-side in JavaScript with no version accessible to bots that don't execute it. In both cases, the result is the same: the content is never retrieved at the search step, and so can never appear in an answer. The lesson The robots.txt file and controlling AI crawlers explains how to check and adjust these settings. A slow site, with high load times, or one whose crawl budget is already largely consumed by low-value pages, may also see its most relevant content crawled less often, or later, than a technically leaner site. These technical performance factors, often overlooked once content is judged satisfactory on substance, deserve regular checking just as much as editorial quality does.
Content too generic to be selected
Even when accessible, content that just rephrases what already exists elsewhere, with no distinct angle, no hard data, no concrete example, has little chance of being kept during passage sorting. Faced with several sources saying roughly the same thing in a similar way, a reranking system generally favors the one that provides more precise or more directly useful information for the question asked. Content that stays on the surface, describing a topic without ever answering it in an operational way, often gets overtaken by more concrete sources. Duplicate, or near-duplicate, content across several pages of the same site poses a similar problem: when several passages say roughly the same thing in an almost identical way, a sorting system has no reason to favor one over the other, and may simply discard both in favor of a more distinct third-party source.
Lack of evidence and trust signals
Generative AI systems, like human readers, need reasons to trust a piece of information before repeating it. Content with no identified author, no date, no verifiable data, no links to external sources, sends weaker trust signals than content that clearly states who wrote it, with what expertise, and on what basis. This isn't a matter of content length: a short but precise article, with an identifiable author and a verifiable source, often inspires more trust than a long but anonymous article that never takes a real position. This bundle of signals is grouped under the acronym E-E-A-T (experience, expertise, authoritativeness, trustworthiness), covered in the lesson E-E-A-T: experience, expertise, authority, trust. The importance of concrete evidence and sourced figures is also covered in the lesson Citations, figures, and evidence: why they matter.
A structure that prevents good fragment-level reading
As seen in the lesson on chunking, an AI rarely reads a whole page: it retrieves fragments of it. Content written as one long continuous text, with no clear subheadings, no self-contained paragraphs, is harder to split correctly, and some of its most useful passages may simply never get isolated or kept. Well-structured content, with subheadings that explicitly pose a question followed by a direct answer, is mechanically easier for these systems to process. The same goes for content locked away in formats that are poorly machine-readable, such as essential information presented only in an image, a video, or a scanned PDF with no extractable text: the content can be visible to a human and yet completely invisible to a system that only reads text.
A brand barely mentioned elsewhere on the web
Finally, a less visible but equally decisive cause: your brand's presence outside your own site. A generative AI doesn't rely solely on what you say about yourself; it also takes into account, directly or indirectly, how your brand is discussed elsewhere — on third-party sites, comparison articles, reviews. A brand that's rarely or never mentioned outside its own site is structurally less likely to be recognized as a relevant answer. It's to pinpoint exactly this kind of gap, and know on which engines and which topics a brand is being ignored, that tools like Vurto measure citation rates per AI engine rather than leaving it to guesswork.
Keep in mind that invisibility in AI answers is almost never a uniform or mysterious phenomenon: it generally breaks down into one or more specific blockers, technical or editorial, that can be identified and fixed one at a time. Addressing these causes one by one, rather than searching for a single sweeping explanation, remains the most effective way to restore lost visibility. Working through this list methodically, rather than assuming a single root cause, is usually the fastest way to move from invisibility to a first citation. A quarterly check across these five areas is usually enough to keep a growing site from quietly drifting back into invisibility. Fixing even one of these blockers can be enough to unlock a first meaningful round of citations.