Sentiment
How the commenter feels about whatever they are reacting to. Measured for every comment.
Comment-level social listening
Hear what the comment section really thinks.
Grapevine reads the public replies under posts and videos, not just the posts, and measures comment by comment how people feel and what they think of the person or brand you follow.
Reddit · YouTube · Bluesky · Disqus
Your target · last 30 days
Illustrative
How people feel
What they think of the target
Two questions, kept apart
Most comments react with feeling. Far fewer pass judgement on the person or brand they sit under. Grapevine measures both, and never fills in a judgement nobody made.
Stance on target
What they think of the person or brand you follow, recorded only when they actually say. “No view” is never counted as neutral, so averages are not dragged toward the middle.
Relevance
Whether a comment is about your target at all, so a reply about a competitor or another commenter does not count for or against you.
How it works
-
Name a target
A person or brand, with the aliases and nicknames people actually use.
-
Point it at the conversation
Follow keywords, or anchor on specific posts, videos and articles to capture every reply beneath them, including the ones that never repeat the name.
-
Collect and classify
Run collection on demand or on a schedule. Each comment is labelled against your target by a pinned language model, and the prompt and model version are kept with every label.
-
Read, drill down, act
Trends over time, reputation with the numbers behind it, themes, alerts and reports. Every point on a chart opens the comments that produced it.
Built for the comment section
Where people reply
Reddit, YouTube, Bluesky and Disqus today, with X and Facebook connectors in preview.
Context mode
Anchor on a post and capture the reaction to it, even from comments that never mention your target by name.
Reputation, decomposed
The reputation index always arrives with its comment counts, sentiment split and change since the last period. Never a lone number.
Themes
Group comments by what they are about, and see which themes carry the strongest feeling.
Alerts
Rules that fire when sentiment, stance or volume moves sharply against its own recent baseline.
Reports
Scheduled PDF, Excel and CSV reports for the people who need the picture without the dashboard.
Coordination signals
Identical comments repeated in a short window are flagged, so a pile-on reads as one campaign rather than a groundswell.
Corrections on the record
Analysts correct labels and the model's original stays on file, so a change in the model never looks like a change in public opinion.
Accuracy you can check
Draw a stratified sample to hand-label, score the published labels against it, and benchmark the classifier on public data.
About the subject, never the commenter
Grapevine analyses what is said about the person or brand you follow. The people writing the comments are not what it studies.
- Comment authors are irreversibly hashed the moment a comment is collected. Their names are never stored, logged or exported.
- Every comment keeps its public link, so you can read it in context on the platform where it was posted.
- Separate workspaces, role-based access, two-step verification and an audit log of every change.