How should buyers normalise prompt volumes and usage costs?

Count observations, not just saved questions. Ahrefs explicitly defines a check as one prompt execution on one model in one location. Its documentation also distinguishes custom tracking from access to a broader prompt database. (help.ahrefs.com)

For procurement, use this desk calculation:

Scheduled observations = prompts × engines × locations × collection occasions × repeats.

Illustrative workload, calculated 16 October 2026—not observed customer demand:

• 100 prompts × four engines × one location × 30 daily collections = 12,000 observations per month.
• Three repetitions per collection would increase that to 36,000 observations.

This does not mean every platform bills those workloads identically, or that repeated outputs represent independent people.

A worked pricing example shows why scope matters. Ahrefs lists a $100 package with 7,000 checks and $0.015 per additional check. Under those published rules, a 12,000-check workload would cost $175 for that check package and overage, calculated as $100 + 5,000 × $0.015. Any required base subscription or separately purchased index access is additional. (help.ahrefs.com)

Are “AI search volumes” actual user-query counts?

Not necessarily. Ahrefs’ documentation dated 3 September 2026 says its AI-adjusted volume uses the Google volume of a parent keyword multiplied by an estimated platform-usage ratio. It also records a methodology change on 31 August 2026, warning that earlier fetched numbers may not match current values. (help.ahrefs.com)

Peec’s current pricing documentation describes prompt volume as a 1–5 relative-demand score for the topic behind a prompt. That is a different measurement product from a count of user searches. (peec.ai)

Procurement rule: require the provenance, unit, geography and methodology version for any claimed volume. Keep estimated demand separate from monitored answers, citations and recommendations.