Credits & metering

Credits are the unit Lokrix uses to meter the real cost of measuring your AI visibility. The rule is simple: 1 credit ≈ 1 engine response. Every live run against an answer engine consumes credits from your workspace balance; the predictive model does not.

What a credit buys

One credit corresponds to roughly one response captured from one answer engine for one prompt, on one sample. Because Lokrix never reports a single-run yes/no — it runs each prompt K times across temperature, persona and locale to build a Monte-Carlo presence probability with a 95% confidence interval — a single prompt on a single engine already costs K credits. Read sampling and statistical honesty for why K samples are non-negotiable.

How seats and usage credits meter live runsPlan seats and usage credits feed a metered live-run step (cost equals engines times prompts times K samples), which writes to a per-tenant metering ledger.Plan seatstiered subscriptionUsage credits1 credit ≈ 1 responseLive runs meteredengines × prompts × KMetering ledgerper-tenant balance

Cost is multiplicative

The credits a scan consumes are the product of four factors, not a sum:

credits ≈ prompts × engines × samples (K) × cadence

  • Prompts — the size of your prompt universe.
  • Engines — five run live by default (ChatGPT, Claude, Perplexity, Grok, Gemini); AI Overviews and AI Mode need a search-provider key to observe.
  • Samples (K) — repeats per prompt for the confidence interval.
  • Cadence — how often the scan repeats (daily, weekly, monthly) multiplies the whole thing over time.

Doubling any one factor doubles the credit draw. See scan cost for a worked estimate before you launch.

Live runs spend; predictions do not

Only live engine responses are metered. The calibrated predictive scoring model returns everyday scores and powers what-if simulation without touching your balance, which is why Lokrix defaults to it and escalates to live runs deliberately — when you need fresh, ground-truth evidence rather than an estimate. Data is always labelled by provenance: live, cached or mock (sample); mock is never presented as live, and cached results re-use a prior live response without re-charging. See live vs. predictive.

Balance and metering

The credit balance is held per workspace, shared by every property inside it. As a scan runs, each captured engine response is metered against that balance in near real time. If a scan would exceed the available credits — or a configured cap — Lokrix stops before overspending rather than silently running up a bill. To bound spend up front, configure budgets & spend caps. For the plan and seat context around credits, see plans & seats.