Cited vs mentioned vs absent

When an AI engine answers a question that touches your brand, the result is not simply “you showed up” or “you did not.” There are three distinct outcomes, and treating them as one number hides the most useful signal you have. An answer can cite you with a clickable link, mention you by name without a link, or leave you absent entirely.

Each state means something different for traffic, for trust, and for what you should do next. Knowing which one you got, prompt by prompt and engine by engine, is the difference between guessing and managing.

What cited, mentioned and absent actually mean

These are not shades of the same thing. They are three separate results produced by the same answer, and the gap between them is where most of your opportunity sits.

• Cited means your URL appears in the answer’s list of sources. This is the strongest result. It carries a clickable link that can send real traffic, and it carries the implied endorsement of being one of the sources the engine chose to stand behind.
• Mentioned means your brand or product is named in the prose, but the engine did not attach a link to you. Your name is in the conversation, which is genuinely valuable for awareness, but the reader cannot click through, and the position is fragile.
• Absent means you do not appear at all for that prompt on that engine. The answer was composed entirely from other sources, and the reader never encountered you.

The subtle middle state is the one people miss. A mention without a link looks like a win in a casual read of the answer, but it behaves much more like a near-miss. The engine knows you exist and considered you relevant enough to name. It simply did not trust a specific page of yours enough to point at it as a source.

Three outcomes, three different next moves.

Why the distinction changes what you do next

If you collapse all three into “visible or not,” you lose your roadmap. A page that is absent and a page that is mentioned-but-not-cited need completely different work, and confusing them wastes effort.

An absent result usually points to a foundational gap. The engine either could not retrieve a relevant page from you, or did not consider any of your content an answer to that question. That is a content and retrievability problem, often upstream of any single page.

A mentioned result is the opposite kind of problem, and a far happier one. The engine already accepts you as part of the answer. It is one step away from linking you. Closing that gap is usually about giving it a clean, liftable, clearly attributable passage to point at rather than inventing relevance from scratch. That is why mentioned-but-not-cited prompts are frequently the fastest wins on the board.

And a cited result is not a finish line either. Citations move as models refresh and as competitors improve. The state you hold today is a position to defend, not a trophy to file away.

How each outcome is detected and recorded

Determining which of the three states you got means reading the actual answer an engine produced for a specific prompt, then checking two things separately. First, is your brand named anywhere in the generated text? Second, does the answer’s source list contain a URL on your domain? Those two checks combine into the three states above.

The detail that matters here is honesty about failure. If a tracking run hits a network error or a rate limit, the correct answer is “unknown,” not “absent.” Writing a failed request as an absence turns a temporary blip into what looks like a lost citation, and you end up chasing a problem that never happened. A trustworthy measurement system separates “we checked and you were not there” from “we could not check.”

The factors that move a result between states

• Retrievability, whether the engine’s crawler can reach and parse a relevant page at all. If nothing is fetched, absence is the only possible result.
• Answer-shaped content, a passage that states the answer plainly and can be lifted without rewriting is far more likely to earn a link than a paragraph that buries the point.
• Clear attribution, unambiguous brand, author and entity signals make it easier for an engine to connect a named mention to a specific linkable source.
• Third-party corroboration, when reviews, communities and comparison pages echo your claim, an engine is more willing to cite you rather than just name you.
• Freshness, outdated pages are more likely to be mentioned from the model’s general memory than cited as a current source.
• Competing sources, when a rival page answers the same question more cleanly, you can slip from cited to mentioned even though nothing on your page changed.

The path from absent to cited, and the fix at each step.

When and where these states apply

The three outcomes are not fixed across the AI landscape. They vary by engine and they vary over time, and both kinds of variation are informative.

Across engines, the same prompt can produce a citation in one place and an absence in another. Engines differ in how aggressively they link sources at all, how much they lean on live retrieval versus model memory, and which domains they already trust. A citation in one assistant tells you little about your state in the next, so each engine deserves its own reading.

Over time, your state on a single prompt drifts as models are refreshed and as the cited set is reshuffled. A prompt that cited you last month can quietly demote you to a mention after an update. This is precisely why a single check is so misleading and why the same prompts need to be watched on a steady cadence.

Best practices for working the three states

1. Track the exact same prompts on a consistent schedule so you are comparing like with like and can see a state actually change.
2. Read your three outcomes per engine, never blended into one figure, because the right next action differs completely between an absence and a mention.
3. Prioritize mentioned-but-not-cited prompts first. They are the shortest path from visibility to a clickable link.
4. For absent prompts, ask the upstream question of whether any relevant page is reachable and answerable at all before tuning wording.
5. For cited prompts, note which page earned the link and protect it, since citations are defended, not banked.
6. Pair every finding with the page that produced it, so a state change points you at a specific fix rather than a vague worry.

The mistake that quietly costs you links

The most common error is reading the answer once, seeing your name in the text, and recording a win. The engine named you, so it feels like success. But a one-off, eyeball-level check cannot tell a durable citation from a fragile mention, and it cannot tell a real absence from a request that simply failed. You congratulate yourself on a mention while the linkable citation goes to a competitor, and you never notice the gap.

The way around it is consistency and precision. Measure the same prompts repeatedly, classify each into the correct one of the three states rather than a yes or no, and never let a failed check masquerade as an absence. You cannot manage a result you only glance at once, and you cannot improve a number you are not separating into its real parts.

Where to start and how to keep it honest

Begin with a clear picture of whether your pages are even built to be cited. The free AI-Readiness Grader scores a page on the retrievability, answerability and attribution signals that decide whether an engine can move you from absent to mentioned, and from mentioned to cited. It tells you, before you spend a single tracking run, which pages are one fix away from a link and which need foundational work.

From there the job becomes continuous, because the three states keep shifting. CiteTrack’s citation tracking records which of the three outcomes you got for every prompt and every engine, refuses to write a network blip as an absence, and surfaces your position among the sources so a demotion from cited to mentioned is caught early. If you want the mechanics behind which passages engines reach for, the lesson on how AI decides what to cite is the natural next read.