AI BEACON · SENTIMENT
How an answer describes you,
scored from 0 to 100.
Presence gets you into the answer. The sentence around your name decides what the reader takes from it. AI Beacon scores that sentence for every prompt you track, so a description that is quietly getting worse shows up as a number rather than a hunch.
Named in every answer. Still losing the deal.
Presence on its own can be misleading, and it tends to mislead in the same few ways.
You appear in the answer as the expensive option, or the one for bigger teams. The mention is real and it is working against you.
An answer repeats a characterisation of your product that is out of date, and there is no obvious page it could have come from.
Somebody has a feeling the tone has hardened over the last few months. Without a score there is nothing to test that against.
Score the sentence, then watch the line.
The score is taken from the answer text itself, which is the only description that reaches the reader.
The prompt is answered
On each engine, because wording varies more between engines than most people expect.
The description is scored
A single number from 0 to 100 for how your brand is described in that answer.
Read across prompts and engines
A score that holds on one question and falls on another tells you where to look.
Follow it to a source
The sources behind the answer are listed, so a description can be traced to what fed it.
See citation tracking →One number, read four ways.
The score is the start. What makes it useful is what you can line it up against.
A single reading, from 0 to 100.
One number per prompt, so it can be tracked, compared and put in a report without needing a paragraph of explanation.
- Per promptEach question you track carries its own score.
- From the answer textTaken from how the answer described you, in that answer.
- ReportableA number a dashboard can hold and a client can follow month to month.
The same question, described differently.
Engines build their answers from different sources, so they often reach different characterisations of the same product.
- Seven enginesPlus Google AI Overview, AI Mode and Search, each read in its own right.
- Side by sideA low score on one engine against a high one elsewhere is a lead worth following.
- With the sources alongsideWhich makes the difference explainable rather than mysterious.
The sentence, not only the number.
A score tells you something changed. The text tells you what to do, because the words usually point straight at the page that produced them.
- What the answer saidThe description as written, for the prompt and engine in question.
- Repeated characterisationsA phrase that turns up in several answers has a source behind it.
- Over timeThe description from last month beside this month's.
A description comes from somewhere.
Sentiment is a diagnostic. When a characterisation is wrong or out of date, the work is usually a page, and occasionally a correction.
- Traced to its sourcesThe pages, threads and videos the answer drew on.
- Written againstContent Studio writes the page that answers the characterisation properly.
- Checked for stalenessOld facts on your own pages are a common cause worth ruling out first.
Presence is a yes or no. Description is the detail.
Tracking one without the other leaves you with a number that goes up while the sentence around your name gets worse.
A typical stack
- –Mentions counted, wording ignored
- –Tone judged from whichever answer someone read last
- –No way to tell a good mention from a damaging one
- –A characterisation with no traceable source
With brand sentiment
- ✓A score from 0 to 100 on every prompt you track
- ✓Read per engine, where the wording often differs
- ✓The answer text kept alongside the score
- ✓Sources listed, so a description can be traced
Sentiment is reported as a single score from 0 to 100. It is not broken into positive, neutral and negative bands.
The rest of AI Beacon, and where this one sits.
Sentiment is most useful read next to presence and sources. These are the pages that hold those.
What is AI brand sentiment?
AI brand sentiment is a measure of how an AI answer describes a brand, as distinct from whether it names it. In AI Beacon it is a score from 0 to 100 for each prompt you track, taken from the answer text on each of 7 AI engines plus Google AI Overview, AI Mode and Search. The score sits beside the wording itself and the sources the answer used, so a characterisation can be traced back to what produced it.
Key facts
- Sentiment is a single score from 0 to 100, per prompt
- It is not split into positive, neutral and negative bands
- It is scored from the answer text, on each engine separately
- The wording and the sources behind it are kept alongside the score
- Presence in the answer is still the primary visibility signal
- Scores are a guide to movement, and no score promises a citation or a ranking
Last reviewed 15 October 2026 by the SEORCE team.
On scoring a description.
No. It is a single score from 0 to 100 for each prompt. Bands sound tidier than they are in practice, because most descriptions of a product are mixed, and a number tracks movement better than a label does.
The answer itself. How the model described your brand in that answer, on that prompt, on that engine. It is not taken from reviews, social posts or customer feedback.
Because the answers are built from different sources. An engine leaning on an old forum thread will describe you differently from one leaning on your own documentation, and the sources are listed so you can see which is which.
Yes. A warm description on a prompt nobody in your market asks is worth less than a lukewarm one on the question that decides a purchase. Read the score next to the prompt, not on its own.
No. Nothing can edit a model directly. What you can do is change what it reads: publish the page that answers the characterisation, correct stale facts on your own site, and earn mentions on the sources the answers keep drawing on.
The Visibility Score is scoped to one project and its calculation is being reworked, so the site does not describe how it behaves. Sentiment is reported in its own right, per prompt, which is the reading to work from.
The mention is a start. The sentence decides.
Track a few prompts and see how each engine describes you on the questions that matter.