Understanding your scores

Sentiment score

Every mention Reputably finds is scored for sentiment — broadly, whether the person is speaking positively, negatively or neutrally about the subject. Scoring is done by a language model reading the actual text, not by keyword matching, so sarcasm and context are handled far better than a simple word list would manage. Sentiment lets you triage: an angry post about your category is worth reading before a neutral passing mention.

Because scoring uses a model, it is a judgment, not a measurement — occasionally a mention will be read more positively or negatively than you'd score it yourself. Treat sentiment as a fast sort, then read the ones that matter. If a model call ever fails, Reputably falls back to a neutral default rather than dropping the mention, so a bad response never blocks a sync or hides a conversation from you.

Buying-intent score

Sentiment tells you the mood; buying intent tells you the opportunity. Reputably scores how strongly a conversation looks like someone actively looking to buy — a recommendation request ("can anyone suggest a good physio in Brunswick?") scores high, while general chatter about the category scores low. High-intent conversations surface first, because those are the ones where a timely, helpful reply can win a customer. This is the difference between monitoring for mentions and monitoring for leads.

AI-visibility results

AI-visibility isn't a single mention score — it's the outcome of asking ChatGPT, Gemini and Claude the buyer-style questions your customers ask, across many phrasings, and recording whether you're named and who's named ahead of you. The result you read is a pattern across those queries and models, not one answer. Because AI answers vary by phrasing, session and over time, the honest way to read it is as a direction of travel week to week, not an absolute grade. See Set up AI visibility tracking and the background guide what is AI visibility tracking?

Reading scores honestly

All of these scores exist to save you time, not to make decisions for you. They sort a large, messy stream of public conversation into "look at this first" order. The judgment — whether to reply, how to reply, whether a review needs a human touch — stays with you. Used that way, the scores are a genuinely powerful filter; used as gospel, any AI score will occasionally mislead.

Where to go next

• Set up lead monitoring — control what gets scored.
• Safe reply guidelines — act on high-intent conversations without spamming.
• Data sources & limitations — what's covered and what isn't.