Glossary
What is sentiment in AI search?
Sentiment is the tone an AI answer takes when it describes your business. Reading it well means looking at the words, not trusting a single score.
What sentiment means in AI search
Sentiment analysis is the practice of judging the attitude in a piece of text. Google Cloud's documentation describes it as identifying a writer's attitude as positive, negative or neutral.
In AI search, the text is an AI answer and the subject is your business. A mention is an answer that names your business, and sentiment only applies to answers with a mention. The question is how the answer describes you when it names you.
Sentiment is about tone. An answer can praise you warmly and still give the wrong opening hours. Accuracy needs its own check against your business facts.
Groups or scores
Some tools give a numerical score. Google Cloud's documentation describes a score running from negative to positive, with a separate magnitude for the strength of feeling.
The same documentation notes that a score near neutral can hide mixed feelings, where praise and criticism cancel each other out. It suggests using magnitude to tell the two apart, and setting thresholds that suit your own use.
Microsoft's documentation for its sentiment service describes labels such as positive, neutral and negative. It says sentiment is judged for each sentence and for the whole document.
A fact is not praise
Many answers simply state something, such as where a business is or what it offers. That is neither praise nor criticism. Counting it as positive inflates the picture.
Keep plain statements in their own group. Do the same for answers that named you but said nothing that could be placed.
Where sentiment readings go wrong
Who a sentence is about matters. An answer may praise one business and criticise another in the same paragraph. A sentence about they or it could point at someone else, so read it in place.
Hedges, negation and irony are hard to read. A phrase such as not the cheapest but reliable carries two signals at once. A careful reading leaves such sentences unsorted rather than forcing a label.
How Carqo relates to sentiment
In Carqo, the sentiment view puts each answer that named your business into one group, Good, Mixed or Bad. Plain facts and answers that could not be sorted are counted separately, and there is no blended score.
The standard reading only looks at clauses about your business by name, and it stays silent on hedges, negation and irony. When answers were not checked for complaints, a note tells you so and warns that Mixed and Bad could be higher than shown.
In practice
- Collect answers that name you Use only answers with a mention. Keep the question you sent, known as the prompt, and the full answer text with each one.
- Read sentence by sentence Read the sentences about your business by name and place each answer in one group.
- Separate facts and unclear cases Put plain statements and answers you cannot sort in their own groups instead of forcing a label.
Common questions
Is sentiment the same as accuracy?
No. Sentiment is about tone, and accuracy is about facts. An answer can be warm and wrong, or cool and correct. Check each one separately. An error in a warm answer still needs correcting.
Why not use a single sentiment score?
A score blends different findings into one number. Google Cloud's documentation notes that a score near neutral can hide mixed feelings. Groups with the sentences behind them are easier to check.
What does it mean if there is no criticism?
It depends on how the reading was done. If the method never labels complaints, an empty group says nothing. Check what the reading looked for before treating silence as good news. Read a sample of the answers yourself to be sure.
Is sentiment the same as accuracy?
No. Sentiment is about tone, and accuracy is about facts. An answer can be warm and wrong, or cool and correct. Check each one separately. An error in a warm answer still needs correcting.
Why not use a single sentiment score?
A score blends different findings into one number. Google Cloud's documentation notes that a score near neutral can hide mixed feelings. Groups with the sentences behind them are easier to check.
What does it mean if there is no criticism?
It depends on how the reading was done. If the method never labels complaints, an empty group says nothing. Check what the reading looked for before treating silence as good news. Read a sample of the answers yourself to be sure.
Sources
- Google Cloud documentation, Natural Language API basics Link checked 29 October 2026
- Microsoft Learn, What is sentiment analysis and opinion mining in Azure Language Link checked 29 October 2026