Brand monitoring

AI brand monitoring across every assistant, continuously

Monitoring only works if somebody is watching when you are not, and nobody opens a dashboard every morning. We check all six assistants continuously, watch for six specific kinds of change, and email you once when one of them happens — with the answers behind it, so you can check our working.

Start 7-day free trialAll seven features

Card required, nothing charged for 7 days

The six things we watch for

Not a configurable rule builder. Six findings the product detects, chosen because each one has an obvious next action.

  • Visibility drop

    Your rate fell against its recent baseline on an engine or overall.

  • Visibility rise

    It went up. Worth knowing why, while you can still find out.

  • Engine blackout

    One engine stopped naming you altogether while the others carried on.

  • Competitor overtook

    A rival passed you on a question you were holding.

  • Cited, not named

    An answer used your page as a source and named somebody else.

  • Negative framing

    How you are described shifted, not just whether you were mentioned.

Once, not once a day

An alert you can ignore is worse than none

The failure mode for a tool like this is obvious: detect a drop, email about it every morning until it resolves, and train the recipient to filter your address. Each finding is written once and sent once. A drop that persists for three weeks is one email, not twenty-one, because the second email teaches people to stop reading the first. The same restraint runs through how we report everything else.

Deduplicated at the database.
A unique index on workspace, finding type and key. Not a best-effort check in code.
Email, not another inbox.
It arrives where you already read things, rather than in a notification centre.
Off is a real option.
Disable alerting per workspace and nothing is detected or sent at all.
States · ChatGPT · 30 answers

  • Named

    You were recommended by name.

    14 / 30

  • Cited, not named

    Your page was read. Someone else got the credit.

    7 / 30

  • Retrieved, not cited

    Fetched, then not used.

    4 / 30

  • Not retrieved

    The assistant never saw you.

    5 / 30
Illustrative figures from a demo workspace, not a customer's data.

Why an alert needs its evidence attached

A number that moved is a claim. The answers behind it are what make the claim checkable.

Assistants regenerate answers each time they are asked, so movement is normal and some of it is noise. An alert that just says your visibility fell eleven points asks you to trust an aggregate, and the reasonable response to that is to ignore it. Every finding links to the rate that moved, the answers that produced it and the pages they cited, so you can tell a real shift from a bad week in about ninety seconds.

This is also why we alert on a competitor overtaking you at the level of a single question rather than at the level of category share. Category share moves slowly and tells you little; one question flipping to a rival is specific enough to act on.

  • No recommendations. We say what changed. We do not invent a fix from a day of data.
  • No severity theatre. Six finding types, no red-amber-green scoring layered on top.
  • Included everywhere. On all four plans, at no extra cost, from the cheapest up.

Common questions about AI brand monitoring

What people ask before they let a tool email them.

What triggers an alert?
Six things: a drop in your visibility, a rise worth knowing about, an engine that stops naming you entirely, a competitor overtaking you on a question, an answer that cites your page without naming you, and a shift toward negative framing. Those are the six the product detects. There is no seventh hiding behind a bigger plan.
How often will I be emailed?
Only when something happens, and we will not promise you a rhythm. Each finding is recorded once and emailed once — a unique constraint on the workspace, the finding type and its deduplication key means a second detection of the same thing writes nothing and sends nothing, so a drop that persists for three weeks is one email and not twenty-one. A change that sits inside its own margin of error is suppressed entirely, so a genuinely quiet week produces no email at all. That is deliberate.
Can I turn alerts off?
Yes, per workspace. Alerting is a preference that can be disabled entirely, and when it is off nothing is detected or sent. We would rather you switch them off than learn to ignore them.
Are alerts included on every plan?
Yes, on all four plans at no extra cost, including Solo at 29 a month. Plans differ in how many questions and brands you track, not in which features you get.
Why alert on a rise as well as a drop?
Because a rise you cannot explain is information. Something you published, or something a competitor stopped doing, changed how a model describes your category. Finding out which is worth more than the alert, and you will only go looking if you know it happened.
Do alerts tell me what to do?
They tell you what changed and link to the answers behind it. We do not generate recommendations, because a plausible-sounding suggestion built on one day of movement is worse than no suggestion. The answers, the citations and the sentiment passages are there so you can reach your own conclusion.

The rest of what we measure

Six more, all on every plan.

Start monitoring your brand

Add your brand and your questions, and alerts start watching from the first run. Card required, nothing charged for 7 days.

Start 7-day free trialSee pricing