Measures
AI visibility results: how to tell if a page worked
How to compare AI answers before and after a page goes live, read the result with confidence, and know what a steady reading tells you.
Sentiment in AI answers: how engines describe you
What Sentiment measures in AI answers, how to read the quotes behind it, and the pages and listings to fix when an engine describes you poorly.
On this page
When an AI assistant names you, it rarely stops at the name. It adds a line about you: what you are good at, who you suit, and sometimes a catch. A buyer reads that line before they reach your site, so it is worth knowing what it says and whether it is getting warmer.
What an answer says about you
Ask an assistant which tool to pick and the answer usually lists a few brands, each with a short description. Here is one invented answer, the kind a buyer might get.
ChatGPT, asked this week
What is the best payroll software for a 20-person company?
For a team that size, Tessera is a strong choice: it handles multi-state tax filing and has clear pricing. Novaly suits contractors, and Pinefield is worth a look if you need time tracking.
You are named firstExample, using an invented brand
Being named, and named early, is covered by Mention rate and Average rank. Sentiment covers the words around your name. Two brands can sit side by side in the same list while one is "the easiest to set up" and the other "fine, if you can live with basic reports".
What Sentiment measures
How we measure lists five measures, in this order: Mention rate, Share of voice, Average rank, Your pages cited and Sentiment. Sentiment is the last of them: how favourably the answers describe you, on a scale from 0 to 100.
BotSay's own model reads how each answer describes you and scores it, the same way every time. The scale has fixed anchors:
- 0 to 20. An explicitly negative description.
- 40 to 60. A neutral mention: you are listed, with no praise and no warning.
- 80 to 100. An explicit recommendation.
The week's figure is the average of those scores, shown as a whole number and labelled Sentiment, estimated. Like the other four measures, it is counted over the questions that leave your brand name out. An answer to a question about you by name would describe you almost by design, so those answers stay out of the count.
The picture below shows one invented answer's description of a brand, beside that brand's Sentiment for the week.
Gemini, how it describes youSentiment this week: 78
Which payroll tool is easiest to set up?
“Tessera is easy to set up and its support answers quickly, though its reports are basic.”
Example, using an invented brand
The sentence praises two things and adds a limit. The limit is the part to work on.
A figure you can check against the words
Sentiment is read, where the other four measures are counted, so BotSay shows it with its evidence. Every score is backed by the quote it came from, and you can open that quote and judge the reading against the words that earned it.
That makes Sentiment a good pointer. When the figure moves, the quotes tell you why. Read it this way and you can report it with confidence, because you can show the sentence behind the number.
A few things help when you read it:
- It covers answers that name you. A brand named in very few answers has very few scores, and a handful of scores can move the average a long way. Read Sentiment beside Mention rate.
- Some mentions carry no score. A bare name in a list, with nothing said about it, has no tone to read. It is left out of the average, and in a week where no mention had a readable tone, Sentiment is blank.
- It belongs to BotSay's scale. Another tool's 0 to 100 score comes from a different model and a different rubric, so the two figures do not compare.
The full method, including what "estimated" means, is in the docs at How BotSay measures your numbers.
What moves Sentiment
The wording of the answers moves it. An engine that recommends you outright lifts the average. One that lists you as an option with no comment sits in the middle. One that adds a caution, about price or support or a missing feature, pulls it down.
Engines write those lines from what they read. If your own pages say little about setup, and a review site says setup is slow, the engine has one source to go on. If a page of yours is out of date, the engine may repeat the old facts. So the work on Sentiment is mostly work on the pages and listings the engines lean on.
Read the least favourable answers first
The Evidence page in BotSay has a tab called What they said. It lists every question whose answers named you, and it puts the least favourable first, so the worst thing said about you is at the top. Each row shows the quote, the engine, where you were named and the score.
Two of the four counts above the list are about Sentiment. Described you well counts answers scored 80 or more. Described you poorly counts answers scored under 40, with the note read these first when there are any. You can narrow the list to one engine or one theme of questions.
Open a single answer and you see the brands in the order they were named, the pages the answer linked to, and the web searches the engine ran before it wrote, where the engine reports them. The linked pages are the ones to look at first. Each is labelled as yours, a rival's or a publisher's.
What the Evidence page shows walks through each part of the screen.
Change how you are described
Once you know which line you would change and which page it came from, the fix depends on whose page it is.
- Your own page. Correct it. If an answer says your reports are basic and you shipped new reports last quarter, put that on the page the engine cited, in plain words, with the facts.
- A gap on your site. If the engine praises a rival for something you also do, and cites the rival's page for it, write your own page on that topic. BotSay's Work list ranks topics where a rival is named ahead of you, and a topic's page can carry a brief you can copy.
- A reference entry. Read what it says about you and correct it through that site's own process.
- A review site. Claim or complete your profile, then ask your customers to review you there.
BotSay's Where to get in list shows the pages on other people's sites that the engines read or linked in answers naming a rival, with a label for each: Check the entry, Claim your profile, Pitch them and others. Find pages to get your brand onto explains every label.
Compare it with your rivals
Your own figure is easier to read beside your rivals'. The Standing page shows every brand you track on the five measures, Sentiment included, and the By engine tab breaks your figures down engine by engine. An engine that describes you less warmly than the rest is a good place to start reading quotes.
If you want to try it by hand first, our post on checking what AI says about your brand shows how to copy the sentence that describes you from each answer and score it yourself. Keep the same scale every week, and keep the questions and their wording the same, so one week compares with the next.
What you gain with BotSay
- Sentiment scored on a fixed scale every week, across six engines, with the quote behind every score.
- The least favourable answers listed first, so you know which line to fix.
- The pages each answer linked to, so you know where the engine got its description.
- Pages to write and places to be listed, ranked, with a brief you can copy where a topic has one.
- One fixed monthly price with everything included and no add-ons. Plans differ only in the number of questions, from $59 a month for 60 questions.
Keep reading
Measures
Average rank in AI answers: how to move up the list
What Average rank counts in AI answers, why it can slip in a good week, how to read it beside Mention rate, and the work that moves you up the list.
Measures
Mention rate in AI answers: how often engines name you
What Mention rate counts in AI answers, how to read its likely range, why it differs by engine, and the work that raises it.
Your first reading runs right after setup
Find the page to write next.
Monthly plans from $59. Cancel at any time.