Models read what others write about you

• A model repeats something negative about you and you do not know where it came from.
• A forum thread from three years ago is still the first source about your name.
• You find out about a small crisis when it is already a big one.

What a model says about you was written earlier by somebody else.

1. 1

   Your brand and its aliases are searched

   With the spelling variants from your entity record, misspellings included. And with your exclusions applied, so your dashboard does not fill up with mentions of the same-named company.

2. 2

   The tone is classified

   Positive, neutral or negative, always with the specific fragment behind it. A classification with no fragment beside it is the machine's opinion, and it is not shown that way.

3. 3

   Alerts fire on deviation

   The alert fires when volume or tone leaves your historic average, not when any mention appears. A brand with twenty mentions a day cannot receive twenty alerts.

• Each mention with its link, date, tone and the fragment justifying it.
• Volume trend against your average, to tell a spike from the usual.
• The sources that mention you most, usually the ones models end up citing.

• Consumer brands, where a forum thread weighs more than a press release.
• Anyone who has had an episode and does not want to be late again.
• Teams working on reputation who need to know which sources feed the models.

Available from Starter, with capture frequency growing by plan. Deviation alerts need a plan with alerts enabled, because an alert with no history to compare against is not an alert: it is noise.

See the full pricing

Does it cover social networks?

It covers the public indexable web: news, blogs, open forums and pages. Closed networks are not covered, and saying so matters: promising coverage of a platform that does not allow crawling is the kind of promise found out in the first month.

Is the tone classification reliable?

It is reasonable and not infallible, especially with irony. That is why each classification comes with the fragment behind it: you can check in two seconds whether it makes sense and correct it. Trusting a sentiment percentage you cannot open is the classic mistake.

How does this relate to what AI says about me?

Direct but not mechanical. A model with search enabled leans on public pages, so the sources that mention you most tend to be the ones it ends up citing. Which one it picks in each answer, though, nobody controls, and promising otherwise would be selling smoke.

Can I ask for a mention to be deleted?

From the dashboard yes, from the internet no. This observes what others publish; it has no ability to delete the original and would not have it even if it wanted to. What you can do is mark it irrelevant so it stops counting in your metrics.