What the five measures mean

Reading your numbersUpdated 1 Oct 2026

BotSay reports five measures: Mention rate, Share of voice, Average rank, Your pages cited and Sentiment. Each counts something different about how AI search engines treat your brand, and Mention rate is the lead figure on This week.

What are the five measures? #

BotSay reports five measures, always in this order: Mention rate, Share of voice, Average rank, Your pages cited and Sentiment.

Mention rate. Of the answers read, the share that named you.
Share of voice. Of every brand mention in those answers, the share that was you.
Average rank. When you are named, how early. Lower is better.
Your pages cited. Of every link the answers cited, the share that pointed at your sites.
Sentiment. How favourably you are described, from 0 to 100. A model's estimate, not a person's reading.

The definitions follow the one-line summaries on Setup, How we measure, and How BotSay measures your numbers gives the full method. The worked examples use an invented brand, Tessera, against Novaly, Pinefield, Kestrel Pay and Orbita.

Which questions do the five measures count? #

The five measures are counted over the questions that do not name you. Evidence says so in one line: Counted over the questions that do not name you, the same set as your mention rate.

A question's class decides this. A Brand question names your brand, so an answer to it names you almost by construction. Counting those would lift your figures and leave every rival's where it was, so they are left out. Rival questions name a rival you track, and Open questions name no brand. Both count.

Each question is counted once per engine, however many times it was asked that week.

What is Mention rate? #

Mention rate is the share of answers that named you, counted out of the questions answered.

An answer here means one question answered by one engine. A question that never came back, was declined, hit an error or has not been read yet is left out of the count, not counted as though you went unmentioned.

Where a question was asked more than once in a week, BotSay averages within the question first and then across questions. A question named in one of two answers counts as one half. The engines are pooled, so an engine that answered more questions counts for more.

Mention rate is the one measure with a likely range, shown as Likely 19.8 to 31.0%. Why a number is blank or shows a range explains it.

What moves Mention rate? #

Five things move it, and only the first is a change in what the engines say about you.

What the engines say about you. Their own wording is the change you are tracking.
Your questions. Adding a question or stopping one changes which answers are counted. Every reading asks your whole set, so two readings compare like with like.
Names. You count as named under the names listed for your brand in Setup, Brands.
What Google shows. Google decides for itself whether to show an overview for a query. When it shows fewer, the rate is worked out over the questions it did answer, so the rate can rise with nothing about your brand having changed.
How many engines are in. A week still being read is marked Provisional until every engine has finished.

Work out Mention rate for a week #

Divide the answers that named you by the answers read: for Tessera, 57 of 228 answers is 25.0%.

In the week of 14 to 20 September, Tessera's 40 questions were read by six engines. That is 240 question and engine pairs, but Google showed nothing for 12 of them, so 228 answers came back. Tessera was named in 57, and 57 divided by 228 is 25.0%.

The likely range for 57 of 228 is 19.8 to 31.0%, so the true rate is probably inside it. A reading of 12 of 40 answers would give 30.0% with a range of 18.1 to 45.4%, which is why the range narrows as more answers are counted. This week shows the first figure as 57 of 228 answers. Likely 19.8 to 31.0%.

Share of voice is your share of all the brand mentions in the answers counted, out of the brands you track.

For each tracked brand, you included, BotSay counts its mentions per question, averaging within a question first. Your share is your total divided by the total for every tracked brand, so the shares of all tracked brands add up to 100%.

Share of voice answers a different question from Mention rate. You can be named in many answers and still hold a small share, if rivals are named in the same answers or named several times.

Your own mentions raise it and your rivals' mentions lower it, and the list of brands you track sets the total.

Your own mentions. More mentions raise your share.
Your rivals' mentions. When a rival is mentioned more, your share falls, even if your own count is steady.
Who you track. Add a rival and the total can grow. Remove one and the total shrinks, so your share rises with no change in your own mentions.
The questions counted. Brand questions are left out, as for every measure.

A rising Mention rate with a falling Share of voice means rivals are being named even more often than you are.

Divide your mentions by the mentions of every tracked brand: for Tessera, 30 of 130 is 23.1%.

Across the questions counted in one week, the tracked brands were mentioned this many times:

Brand	Mentions
Novaly	45
Tessera	30
Pinefield	25
Kestrel Pay	20
Orbita	10

The total is 130, so Tessera's share is 30 divided by 130, which is 23.1%, and Novaly's is 34.6%. If Tessera then stops tracking Orbita, the total falls to 120 and Tessera's share becomes 25.0%, though nothing about Tessera has changed. The figure is a share of the brands you track, so keep that list steady when you compare weeks.

What is Average rank? #

Average rank is where you come when an answer lists several brands, counting only answers that already name you. 1 is first, and lower is better.

It prints with a hash, for example #2.50. For each question BotSay takes the order in which the brands were named, averages your position within the question, then averages across questions and engines.

A change reads 0.12 better or 0.12 worse, never "up" or "down", because a fall in the number is good news. An answer that names you in prose, without a ranked list, gives you no rank, so Average rank can be blank in a week where Mention rate is not.

What moves Average rank? #

How early engines name you moves it, and so does how many answers name you at all.

How early engines name you. Being named first is the change you want.
How many answers name you. Average rank only looks at answers where you were already named, so it can get worse for a good reason. Get named in more answers, including harder ones, and you appear further down in more of them.
The questions counted. Brand questions are left out, as for every measure.

Read it beside Mention rate, because Average rank ignores every answer that left you out.

Work out Average rank for a week #

Add up your positions and divide by the number of answers that named you: for Tessera, positions 2, 1, 4 and 3 give #2.50.

The total is 10, and 10 divided by 4 is #2.50. The next week Tessera is named in six answers, at positions 2, 1, 4, 3, 4 and 4. The total is 18, and 18 divided by 6 is #3.00, so Average rank has moved 0.50 the wrong way.

Mention rate rose, because two more answers named Tessera. The two new answers were fourth-place mentions, which is why the average slipped. This is a good week with a worse-looking Average rank.

What is Your pages cited? #

Your pages cited is the share of links cited, for questions that do not ask about you by name, that point to a page you own.

A link is cited when the answer points the reader at it. A page the engine read without pointing at it is retrieved, and is not counted here. The links are pooled across engines.

A link counts as yours when its site is one of the sites listed as yours in Setup, Brands. The count uses the list as it stands today, so changing the list changes past weeks too. If BotSay finds a site that may be yours and is not listed, What they cited on Evidence can show a card, This is understated, with a Decide button.

What moves Your pages cited? #

Whether engines cite your pages moves it, along with what else they cite and which sites you list as yours.

Whether engines cite your pages. More citations of your sites raise the share.
What else the engines cite. More links to other sites lower your share of the total.
Your list of sites. Adding a site can raise the share in every week on record.
The questions counted. Brand questions are left out.

A rival with no site listed cannot have its pages told apart from anyone else's, and This week may offer Add your rivals' sites. What the Evidence page shows says which sites and pages make up the figure.

Work out Your pages cited for a week #

Divide the links that point at your sites by all the links cited: for Tessera, 36 of 400 is 9.0%.

In one week the answers to the questions counted cited 400 links across all engines, and 36 pointed at pages on sites Tessera has listed as its own. 36 divided by 400 is 9.0%.

Tessera then lists a second site, a documentation site, which accounted for 20 of the links already cited. The same week is counted again with 56 of the 400 links, which is 14.0%. The answers did not change, and the list did. That is why the definition carries the note: We count the sites listed as owned today, so changing a list changes past weeks too.

What is Sentiment? #

Sentiment is how warmly the answers describe you, on a scale from 0 to 100, estimated by BotSay's own model.

The model reads how an answer describes you and scores it, the same way every time. The anchors are fixed: 0 to 20 is an explicitly negative description, 40 to 60 is a neutral mention, and 80 to 100 is an explicit recommendation. Evidence calls a mention described well from 80 and poorly under 40.

The figure is the average of the scores, taken over mentions that carry a score. It prints as a whole number, and This week labels it Sentiment, estimated. Every score is backed by the quote it came from.

What moves Sentiment? #

The wording of the answers moves it, and so does which mentions get a score.

The wording of the answers. Whether an engine recommends you, lists you as one option, or adds a caution.
Which mentions are scored. A mention with no readable tone gets no score and is left out, so a few scored mentions can move the figure.
The model's reading. It is an estimate, and it does not compare across tools: another product's 0 to 100 score reaches its number with a different model and rubric.

Sentiment carries no likely range. It is a hint, so open Evidence and read the quotes behind it.

Work out Sentiment for a week #

Average the scores of the mentions that carry one: for Tessera, 80, 70, 60, 70 and 70 give 70.

In one week Tessera is named in five answers that carry a score. The total is 350 and the average is 70, so This week shows Sentiment, estimated 70.

In What they said on Evidence, the answer scored 60 reads as neither well nor poorly. Only a score from 80 counts as Described you well, and only a score under 40 as Described you poorly. Scores from 40 to 79 get no label, so the five quotes show more than the 70 does.

Tessera, Novaly, Pinefield, Kestrel Pay and Orbita are invented brands, used here as examples.