Skip to main content
Browse documentation

Getting started

Tracking prompts

A prompt is a question Ranksify asks the models on your behalf, repeatedly, so the answers can be compared over time. Manage them at Your Prompts, grouped by topic.

The four states

Only active prompts are sampled. Everything else is a holding area, which is what makes it safe to generate a long candidate list without spending anything.

StateMeaning
proposedSuggested but not reviewed. Costs nothing.
stagedReviewed and queued, still not sampled.
activeSampled on schedule. Takes one of your plan's active-prompt slots.
archivedRetired. History is kept; no new samples are taken.

Sampling policy

Each prompt carries a sampling policy — daily, weekly or adaptive — which decides how often it is re-asked. A single run of one prompt against one engine is a sample, and samples are metered on a daily window, while prompt activation is metered monthly.

Getting prompts in

  • The onboarding wizard proposes a starter set from your domain and brand context.
  • Generation produces candidates in bulk for review, deduplicated against what you already track.
  • Search Console suggestions. If Search Console is connected, your top-clicked queries from the trailing 28 days are offered as prompts, already deduplicated against your existing list. Queries you rank for are the questions people are already asking.
  • Typing them. Mark a prompt branded if it names you explicitly — the insight filters split branded from unbranded, and the two behave very differently.

Topics and fanout

Topics group prompts into themes so visibility can be read per theme rather than one question at a time; they are the same topics the insight views filter on. Individual prompts can also be expanded into their query fanout — the related searches a model is likely to run while answering — which is surfaced in the Fanout view.

Quota

The prompt quota is a concurrent cap, not a monthly spend: it counts how many prompts are active right now. Activating one takes a slot and archiving it gives the slot straight back the same day — nothing resets at the start of a month, and a prompt you activate and archive twice has cost you one slot, twice, not two.

Activating is the action that spends the slot, so the prompts page shows live usage for both meters. If your active prompts need more samples than your daily allowance covers, the page says so directly rather than silently under-sampling. Per-plan limits are on the pricing page.

A prompt with no samples yet is not a bug — it is a prompt that has not come round in the schedule. Visibility numbers only become meaningful once a prompt has been asked enough times for its confidence interval to tighten, which is why every metric in Ranksify carries its sample size.