Check cadence

RankX AI evaluates scheduled work nightly at 02:00 UTC and runs whatever is due. Whether something is due depends on the cadence you chose for it, raised to a floor. Two floors apply and the higher wins: your plan's floor, and a seven-day floor while your account is on trial.

Nothing runs every night unless you set it to. Nightly is when RankX AI looks, not how often anything runs.

Cadence is also where most of your credit spend is decided, because everything recurring is a per-unit charge multiplied by how often it runs.

The four check families

Each has its own cadence setting, and each has its own floor.

AI visibility goes one level finer: per prompt

The other three families are set per Website. AI visibility is set per Website and then per prompt, and the per-prompt setting is what most accounts should actually use.

Every tracked prompt carries two settings of its own, both of which start out following the Website:

So one prompt can run on ChatGPT every night while another runs on Perplexity once a week. The effective cadence is the slower of the prompt's choice and the floor, so a per-prompt setting can move spend around under the ceiling your plan already sold you, never past it. A prompt's assistant list can only narrow the Website's, never widen it.

This is the control worth reaching for first. Pruning a panel and slowing the prompts that do not need watching costs nothing and changes the bill more than anything else on this page. See AI prompts.

The two floors

Your plan's floor is the fastest cadence the plan permits. On most plans and most families that is one day; the entry plan on each track is slower for rank checks and for audits. The per-plan figures are in plans and limits.

A seven-day floor applies while your account is on trial, on every plan and every family. It is not a throttle applied to you: a daily sweep of a full prompt list would consume more than the entire trial grant within the first few days, leaving the account at day four with one measurement and no credits. Seven days buys a scheduled re-check, which is the thing that actually shows a movement. See the free trial.

The higher floor wins, and your plan's floor applies again the moment you subscribe.

Choosing a slower cadence is always allowed

A floor is a lower bound on the interval, which is an upper bound on how often you spend. Nothing stops you checking less often than the floor, and for many Websites that is the right answer.

The arithmetic

Recurring spend is almost entirely this:

AI visibility: for each prompt, assistants it targets x runs per month. Rank tracking: keywords x checks per month. AI Shopping: shopping prompts x 2 surfaces x runs per month.

Twenty-five prompts on all six assistants, checked daily, is 4,500 charged checks a month. The same panel checked weekly is about 650. A mixed panel sits between the two: five prompts on two assistants daily plus twenty on all six weekly is 780 charged checks a month. Nothing else you can change moves the bill that far.

Match cadence to plan first. Each plan's monthly grant is sized for a cadence that answers real questions rather than for the fastest cadence on a full prompt list, which is why the floors exist as floors rather than as defaults.

Choosing a cadence that makes sense

Ask what decision the data changes. If a position moving today would change what you do today, check daily. If you review monthly, a daily check is 30 times the cost of a measurement you look at once.

AI visibility is noisy at any cadence. Assistants are non-deterministic, and two runs of the same prompt frequently return different brand lists. Daily checking does not reduce that noise, it produces more samples of it. A weekly rate across a panel is a better measurement than a daily rate on a small one.

Rank tracking rewards frequency more, because a position is a much more stable measurement than an answer. If you have to spend somewhere, spend here.

Audits reward it least. A site does not change technically between Tuesday and Wednesday unless you changed it. Run one when you have changed something, and on a slow schedule otherwise.

Prune before you speed up. Halving a prompt panel and doubling the cadence costs the same and measures better, because a smaller panel of prompts you care about beats a larger one you scroll past.

Manual runs

You can run a check at any time, as often as you like, for one prompt, one keyword or a whole panel. A manual run is charged exactly as a scheduled one, and it confirms the cost before it starts. It ignores the cadence, because a person asking for a check has said what they want, but it still honours the prompt's assistant list, so the same prompt costs the same whichever door it came through.

The floor governs the schedule, not your ability to ask. So an account on trial can still check something now; it just does not get a daily automatic sweep.

Worth running after a real change, and not worth running to see whether the number moved overnight.

What happens when a run is interrupted

Scheduled runs are built to survive interruption. A run that runs out of time records what it completed, marks itself partial, and dispatches the remainder rather than starting again from the top. A partial run is visible as partial, so the coverage gap is queryable rather than silent.

If the wallet runs dry mid-run, the run stops and says so, and it does not report the prompts it never reached as absences.

RankX AI also reports why a scheduled check was skipped when one was, which is usually one of: the cadence had not elapsed, the wallet was short, or the schedule was paused.

Where to go next

• Credits and metering, for the spend model.
• Plans and limits, for the floors.
• The free trial, for the seven-day floor.

Last updated 1 October 2026