What gets scored
Every inbound item is scored: public comments, inbound replies, and each incoming direct message (scored individually so you can see a thread’s tone change over time). Your team’s own replies are never scored, and items with no scorable text (a sticker or media-only comment) are left unscored. Emoji-only text is scored.
Reading the score
Each scored item gets a value from 0 to 100 and a label:
| Value | Label |
|---|---|
| 81–100 | Very Positive |
| 61–80 | Positive |
| 41–60 | Neutral |
| 21–40 | Negative |
| 0–20 | Very Negative |
In the list and thread, each scored item shows just its label — a small colored tag like Positive. The numeric 0–100 score appears in the Sentiment Analysis panel on the right when you open an item. Items that couldn’t be scored simply show nothing — no placeholder or spinner.
Conversation and post rollups
List rows show a conversation’s overall sentiment as a label, and a post’s sentiment rolls up across all of its comments. Rollups are recency-weighted — more recent messages count for more — rather than a flat average. You can filter the list by one or more sentiment labels, from Very Negative to Very Positive. The sentiment filter applies to comments only, and the list always stays in most-recent-activity order.
Warning
Sentiment is a triage aid, not a verdict. AI can misread sarcasm, dialect, and context — a blunt-but-fair message may score low, and a passive-aggressive one may score neutral. Always read the message yourself. Scores update if a comment is edited, and are excluded from public-records exports. Seeing sentiment requires the sentiment analysis add-on. With the add-on enabled, comment sentiment follows read-comments access and DM sentiment follows read-DM access.