Labels
Your own categories, applied to every incoming mention by AI. The description you write is the instruction the model follows.
Labels turn a feed into a queue. Buying signal goes to sales, Support risk goes to support, Product feedback goes to whoever owns the roadmap, and nobody has to read everything to find their share of it.
The setup wizard proposed a starting set based on your website. Edit them at Settings → Labels.
Name, colour, description
The name is what appears on a mention card. The colour is how you pick it out at a glance in a long feed. The description is the part that does the work: it is passed to the AI as the rule for when to apply the label.
Descriptions that work say when to apply the label, not what the label means:
- Good: “Apply when someone is evaluating us, comparing us to an alternative, or asking for a recommendation.”
- Weak: “Sales stuff.”
If a label is being applied too widely or not enough, the description is the thing to change. It is the highest-leverage edit in the product.
How many
Up to 10 per brand, on every plan. Fewer is usually better: a mention can carry several labels, and a long list makes the boundaries between them blurry for the model and for you.
Correcting a mention by hand
Click the tag icon on any mention in the feed to add or remove labels on that one mention. It is a direct correction rather than a training signal It fixes the mention in front of you; the description is what fixes the pattern.
Editing an existing label
Renaming or recolouring a label updates it everywhere immediately, on mentions already labelled with it. Changing a description only affects mentions scored after the change, so older mentions keep the labels they were given.
Deleting a label removes it from every mention that carried it. There is no undo, so a label you might want back is better renamed than deleted.
Competitors get their own labels, written for their company rather than copied from yours. Giving them matching names makes the two feeds much easier to compare.