Core concepts

Mentions, citations & share of voice

There are several distinct things that can happen to your brand inside one AI answer. Knowing which one happened — and why — is the foundation everything else in OptimizeCamp is built on.

Five ideas, one answer

Every time an answer engine responds to a tracked prompt, OptimizeCamp looks for five distinct things: whether you were mentioned, whether you were cited, what sentiment you were mentioned with, how prominent your position in the answer was, and — once you compare yourself to competitors — what share of voice you hold. They sound similar but they’re measured completely independently, and mixing them up is the single most common way people misread their own dashboard.

Being mentioned

Mention

Your brand’s name — or one of its aliases — appears in the words of the answer itself. OptimizeCamp finds mentions with a word-boundary match against your brand name and every alias you’ve configured, case-insensitive, so “OptimizeCamp,” “optimizecamp,” and “Optimize Camp” all count as the same brand being named.

A mention is purely about the text. It doesn’t matter whether the answer also links to your site — being named is enough.

Being cited

Citation

One of your domains appears as a linked source backing up the answer. OptimizeCamp takes the answer’s citation list and matches each link’s domain against your domain and every tracked competitor’s domain. A match attributes that citation to that brand; a link to some unrelated third-party site — a review roundup, a forum thread — is a neutral citation that isn’t credited to anyone.

A citation is purely about the link list. Your brand name doesn’t need to appear anywhere in the prose for a citation to count.

Sentiment: how you’re mentioned

Sentiment

The tone of the sentence where you’re mentioned — positive, neutral, or negative. Unlike mentions and citations, sentiment can’t be worked out with pattern matching alone, so it’s the one piece of detection that uses a small, fast AI model: it reads the full answer plus the candidate brand matches, and returns a sentiment for each one. If that call fails for any reason, OptimizeCamp defaults every mention to neutral rather than guessing — sentiment never blocks or breaks a run.

Position: how early you show up

Position

also called prominence

How early your brand appears relative to every other brand named in the same answer. The first entity mentioned is assigned position 1, the second is 2, and so on — based on the character offset of each entity’s earliest name match in the text. Showing up first reads as meaningfully stronger than showing up fourth, even within the same answer, which is why position feeds directly into your score (more on that in Metrics & the Visibility Score).

Share of Voice

SoV

Your mentions as a percentage of all mentions — yours and every tracked competitor’s, combined. If you and three competitors are mentioned across a set of answers and you account for 40 of the 100 total brand mentions, your Share of Voice is 40%. It’s the headline competitive number because it answers one question directly: of all the attention AI answers give to brands like yours, how much of it is yours?

Why a mention and a citation aren’t the same thing

Because detection runs these two checks completely independently — one over the answer’s prose, one over its link list — you can absolutely have one without the other. That’s not a bug in the measurement; it reflects something real about how the AI actually used your content.

	Cited	Not cited
Mentioned	Named in the text and linked as a source — the strongest combination.	Named in the text, but nothing links back to you. Common when a model recalls your brand from training without pulling a fresh page.
Not mentioned	Your page is linked as a source, but your brand name never appears in the written answer.	You don’t show up in this answer at all.

A citation without a mention is worth paying attention to — it means the model leaned on your content but didn’t bother naming you in its answer, which is exactly the kind of gap the Optimize track exists to close.

How OptimizeCamp finds all of this in an answer

Detection runs in a fixed order, and it’s deliberately deterministic wherever it can be — an AI model is only called in for the one job that genuinely requires judgment.

Citations are read straight from the source

The platform’s own structured citation list is matched, link by link, against every domain you and your competitors track. This never involves guessing at links from prose.

Mentions are matched deterministically

A word-boundary search runs over the raw answer text for your brand name and every alias, and separately for each competitor’s name and aliases. Every match records where it happened in the text and how many times it occurred.

One small AI call judges tone — and catches false alarms

A lightweight model reads the full answer plus the list of candidate matches from step 2, and for each one returns a sentiment and a judgment call: does this match actually refer to the brand? That catches cases like a brand literally named “Notion” matching the ordinary English word “notion” — a false positive the deterministic step alone can’t tell apart.

This is also exactly why your brand profile’s aliases matter so much: step 2 can only find what it’s told to look for. A missing product name or an unlisted spelling variant means real mentions slip through undetected.

A worked example

Say you track OptimizeCamp against a competitor called RivalReach, and a tracked prompt — “what are the best AI visibility tools?” — comes back with this answer:

“There are several strong options. OptimizeCamp stands out for its closed-loop reporting, and RivalReach is a popular pick for larger teams. Reviewers on G2 tend to rate both highly.”

Citations returned with this answer: a link to rivalreach.com/features and a link to g2.com/best-ai-visibility-tools.

Brand	Mentioned?	Sentiment	Cited?
OptimizeCamp	Yes — position 1 (named first)	Positive	No — no optimizecamp.com link in this answer
RivalReach	Yes — position 2	Neutral	Yes — rivalreach.com/features
g2.com link	—	—	Neutral citation, not credited to either tracked brand

Notice OptimizeCamp is mentioned and mentioned first, with positive sentiment — but earns no citation in this particular answer. That single answer would still count fully toward Mention Rate and Position Score, but not toward Citation Rate. Multiply this pattern across every tracked prompt and platform, and you get the metrics covered next.

Key takeaways

A mention (named in the text) and a citation (linked as a source) are detected independently — you can have either without the other.
Sentiment is the one detection step that needs judgment, so it's the only one that calls a (small, fast) AI model — with a safe neutral fallback if that call fails.
Position is based on how early your brand's first mention appears relative to every other brand named in the same answer.
Share of Voice is your mentions as a percentage of all tracked mentions — yours plus every competitor's — the headline competitive number.
Detection can only find aliases it's been told about, which is why your brand profile's alias list directly controls measurement accuracy.

Next step

Ready to see how these become the numbers on your dashboard? Metrics & the Visibility Score explains what mentions, citations, position, and sentiment each mean for the scores you see.