Crawl schedule and manual runs

Tracking is automatic. Once a day, at a fixed hour, every active prompt of every paid brand is sent to every engine that brand tracks, and the results become that day’s data point.

The daily poll

The check runs at 06:00 UTC. It is one run per prompt per engine per day, the same for every paid plan: a bigger plan buys more prompts and more engines, not more checks of the same prompt.

Free accounts are not polled automatically. On the free trial you get the same daily cadence as a paid plan while the trial lasts; when it ends, automatic polling stops and the dashboard keeps the history you already have.

Changing the frequency

Brand Settings → Crawl settings lets you choose Daily, Weekly, Monthly or Off, and for weekly or monthly you also pick the day. Less frequent checks make the trend line coarser but stretch your polling budget over more prompts, which is a reasonable trade in a slow-moving category.

Off is useful when a brand is parked: no runs, no spend, and the history stays.

Next check and crawl history

The dashboard’s “Next data refresh” card shows when the last check ran, when the next one is due, and links to the full crawl history. Each entry lists how many runs it produced and whether it completed, was partial, or failed.

A partial run usually means one engine was unavailable or rate limited at that moment. Nothing is lost: the prompts that did complete are counted, and the next day’s run picks up normally. If you see several failed runs in a row, contact support, that is not normal.

Run now

The Run now action on a prompt row polls that one prompt through all of your engines immediately, writing a real run. Use it when:

• you have just added a prompt and do not want to wait for tomorrow
• you have published something and want an early read on whether it changed anything
• you are demonstrating the product to someone and want live output

Manual runs are capped per plan: 10 in total during the free trial, then 30 a month on Starter, 100 on Pro and 300 on Team. One click can consume several engine calls, so spend them where the answer will actually change your mind.

Related: Prompts, AI engines and models, Plans and limits.