Skip to main content

​
Overview

Most of these metrics are rates, and a rate is only as clear as its denominator. This page states the denominator for every one of them, because several metrics that sound similar do not divide by the same thing.
Metrics are grouped into: Coverage, Mention, Source, Sentiment, Shopping, Fan-out and Social.
Every metric respects the filter bar. Filters that narrow the population of responses — Sources, Shopping, models, countries — also narrow the denominator. Filtering to “Sources” makes Source rate 100% by definition.

​
Coverage metrics

Type: NumberEvery AI response collected in the selected period, after filters. This is the denominator for Mention rate, Citation rate, Source rate and Shopping Coverage.
Type: NumberThe number of distinct queries that produced at least one response in the period — not the number of queries configured in the project, and not queries × runs.
A query that ran in three languages counts as three queries here, because each is a separate query row.
Type: PercentageSource rate=Responses with at least one citationTotal responses×100\text{Source rate} = \frac{\text{Responses with at least one citation}}{\text{Total responses}} \times 100How often the AI platforms bothered to cite anything at all.
Type: PercentageShopping Coverage=Responses containing shopping resultsTotal responses×100\text{Shopping Coverage} = \frac{\text{Responses containing shopping results}}{\text{Total responses}} \times 100The Overview card calls it Shopping rate; the Shopping page calls it Shopping Coverage. Same number.

​
Mention metrics

Type: PercentageMention rate=Responses mentioning the entityTotal responses×100\text{Mention rate} = \frac{\text{Responses mentioning the entity}}{\text{Total responses}} \times 100An entity mentioned five times in one response counts once. The denominator is every response in scope, not only those that mention some brand.
Scoped variants use a scoped denominator: on the Tags page the denominator is the responses carrying that tag; on the Platforms page it is that model’s responses; on a query it is that query’s responses.
The REST API calls this field visibilityPct, and some cards are still titled “Visibility score”. It is the same metric.
Type: Number (rank — lower is better)The mean of the position at which the entity appears, averaged over mentions.
When a model does not report a position, MentionLab stores 0 and that zero is included in the average — so an entity whose positions are frequently unreported will show an artificially low (apparently better) average. An entity with no mentions at all also renders as 0, displayed as -. Read this metric alongside the mention count, never on its own.
Type: NumberThe total number of times the entity’s name (or an alias) appears. Unlike Mention rate, repeats within one response all count.
“Mentions” means two different things depending on the page. On the Overview and Competitors entity tables it is the raw occurrence count. On the Sentiments page, “Total Mentions Analyzed” and the “Mentions” column count one per response per entity — a presence count, not an occurrence count.
Type: NumberAvg. mentions per result=Number of mentionsResponses mentioning the entity\text{Avg. mentions per result} = \frac{\text{Number of mentions}}{\text{Responses mentioning the entity}}How prominently a brand is featured when it is featured. The denominator is responses where the entity was present, not all responses. Shown as - when the entity was never mentioned.
Type: PercentageYour brand’s slice of all brand mentions. Three surfaces compute it differently — the numbers will not match, and all three are correct for their context:
Because one response can mention several brands, the denominator is a sum of per-brand counts, not a count of responses. It still sums to 100% across brands.
Type: PercentagePresence=Responses carrying the tagTotal responses×100\text{Presence} = \frac{\text{Responses carrying the tag}}{\text{Total responses}} \times 100Used by the Overview Response Tag Visibility card.
The card is described as “percentage of queries covered by each response tag”, but the denominator is responses.

​
Source metrics

Sources have three different units on one row — responses, pages and citation rows. Which one a metric uses decides what it means.
Type: NumberThe number of distinct AI responses that cited this domain. A response citing four pages from the same domain counts once.
Type: PercentageCitation rate=Responses citing this domainTotal responses×100\text{Citation rate} = \frac{\text{Responses citing this domain}}{\text{Total responses}} \times 100A per-domain reach metric: how much of your query volume this one site touches. Rates across domains do not sum to 100%.
The Platforms page has a differently-defined metric with the same label. There, “Citation Rate” is the percentage of a model’s responses that cite anything — a property of the model, not of a domain.
Type: NumberDistinct cited URLs from the domain. Shown in the domain detail sheet as Total page citations (a count of URLs, despite the word “citations”).
Type: NumberDistinct cited URLs from the domain on which your brand was found. Shown as Entity page citations. Always a subset of Pages.
Type: PercentagePresence=Brand pagesPages×100\text{Presence} = \frac{\text{Brand pages}}{\text{Pages}} \times 100Of this domain’s cited pages, how many mention you.
In the Sources grid this cell displays a ratio like 12/40 rather than a percentage. The colour band and the sort order both come from the percentage.
Type: PercentageTrue reach=Brand citation rowsTotal citation rows×100\text{True reach} = \frac{\text{Brand citation rows}}{\text{Total citation rows}} \times 100Of every time this domain was cited, how often the cited content mentioned you.
True reach counts citation rows, while Presence counts distinct pages. A domain can show high Presence and low True reach at once — its brand-mentioning pages exist but rarely get cited.
Only meaningful at domain level, not for individual URLs.
Type: PercentageCitation share=Responses citing this domainSum of that count across all domains×100\text{Citation share} = \frac{\text{Responses citing this domain}}{\text{Sum of that count across all domains}} \times 100Sums to 100% across domains.
The denominator is always the whole project. Filtering the grid by category or search does not rescale the remaining rows.
Type: PercentageEntity citation share=Brand pages on this domainBrand pages across all domains×100\text{Entity citation share} = \frac{\text{Brand pages on this domain}}{\text{Brand pages across all domains}} \times 100Your brand-mentioning cited pages, by domain. Note this is built on pages, not citations, despite the name.

​
Sentiment metrics

Type: Number (1–100)The mean of the per-mention sentiment scores the analysis model assigns, where 1 is most negative, 50 neutral and 100 most positive.
The Sentiments page renders this score with a % sign. It is an index, not a share — a score of 62 does not mean 62% of mentions were positive.
The thresholds that classify an individual mention as positive / neutral / negative are different (≤33 negative, ≤66 neutral, above that positive). So a brand with a 55 average score can still show a majority of individually-positive mentions.
Type: PercentagesThe split of mentions across positive, neutral and negative:Category %=Mentions in categoryPositive+Neutral+Negative×100\text{Category \%} = \frac{\text{Mentions in category}}{\text{Positive} + \text{Neutral} + \text{Negative}} \times 100
Mentions classified as mixed, and mentions with no score, are counted in Total Mentions Analyzed but in none of the three slices — so the three counts will not add up to that card.
Type: ComparisonPer-entity sentiment score and breakdown, so you can compare how AI platforms describe you versus competitors. This table always lists every entity, even when a View as pin is active.

​
Shopping metrics

Type: NumberDistinct brands, products and retailers detected in shopping results. Brands and products are counted after merging, so duplicates you merged collapse to one. Retailers are not merged.Shopping appearances that could not be resolved to a brand or product are excluded from these counts and from the All Products table — map them from Project settings → Shopping Settings → Shopping Products.
Type: PercentageVisibility=Shopping responses including this brand or productResponses containing any shopping result×100\text{Visibility} = \frac{\text{Shopping responses including this brand or product}}{\text{Responses containing any shopping result}} \times 100The denominator is shopping responses only — not all responses.
Type: PercentageA brand’s share of all shopping mentions.
Unresolved (“Not matched”) mentions are deliberately kept in the denominator, so the visible brand shares add up to less than 100%. The shortfall is the size of your mapping backlog.
Type: PercentagesEach retailer’s share of distinct product listings across all retailers. Merchants below 1% are grouped into Other.
Type: PercentagesOn the All Merchants table:

​
Fan-out metrics

Every fan-out rate and average divides by responses that produced at least one fan-out, never by all responses. Responses where the model suggested nothing are outside the denominator entirely.
Type: NumberEvery fan-out entry in scope. One response can produce several, so this exceeds the response count. Repeats of the same text within one response are collapsed at ingest.
Type: NumberDistinct fan-out texts. Deduplication is an exact text match — Best CRM 2026 and best CRM 2026 count separately.
Type: NumberAvg per response=Total fan-outsResponses that produced at least one fan-out\text{Avg per response} = \frac{\text{Total fan-outs}}{\text{Responses that produced at least one fan-out}}Because of that denominator this value can never fall below 1.0, and adding models that emit no fan-outs will not lower it. The REST API calls this field avgPerExecution.
Type: PercentageOccurrence rate=Responses suggesting this fan-outResponses that produced at least one fan-out×100\text{Occurrence rate} = \frac{\text{Responses suggesting this fan-out}}{\text{Responses that produced at least one fan-out}} \times 100How reliably a given follow-up gets suggested. A fan-out near 100% is a question the model treats as unavoidable in your category.

​
Social metrics (YouTube)

Type: NumbersVideos cited is the number of distinct videos cited. Citations is the number of distinct AI responses that cited any YouTube video. Citations / video divides one by the other.
Type: PercentageA video’s, channel’s or country’s share of YouTube-video citations in scope.
This column can sum to more than 100% down a table. One response often cites several videos, from several channels, in several countries, and counts once toward each. Individual rows are still ≤100%.
The YouTube Channel Share donut deliberately uses a different denominator so its slices total 100%. A channel’s donut percentage will not match its Citation Share in the table.
Type: PercentageHow often this video, channel or country appears across all AI responses in scope — the same denominator as the rest of the product.
Type: NumberDistinct home countries of the cited channels.
This count excludes channels with no country set, while the By Country table includes them as an “Unknown” row.