How are the prompts designed, and can I customise them?
Each platform runs a fixed 50 prompts per cycle: 44 unbranded discovery and comparison prompts plus 6 branded fact checks. The free check previews 5 suggested buyer prompts, self-serve plans fix their 50 at project set-up and freeze them, and project or managed engagements get a hand-designed prompt map. Once frozen, the set does not change during the term.
Version 1.1 · Last updatedEvery platform runs a fixed 50 prompts per cycle: 44 discovery and comparison prompts that never name the brand, plus 6 fact-check prompts that do. The free check previews 5 suggested buyer prompts; self-serve plans fix their 50 prompts at project set-up and freeze them; project and managed engagements get a prompt map designed by hand. Customisation happens at set-up. After the freeze the set does not change for the rest of the term, which is what makes one cycle comparable with the next.
How the 50 prompts are structured
| Group | Count | Purpose | In the mention-rate denominator? |
|---|---|---|---|
| Unbranded discovery / comparison | 44 | Simulate a buyer who does not know you yet and see whether the AI brings you up or recommends you on its own | Yes (observations that produced a normal answer) |
| Branded fact checks | 6 | Name the brand and test whether the AI describes it accurately | No |
Three routes: preview, self-serve set-up, hand-designed
- Free check at /en/geocheck: previews 5 suggested buyer prompts based on your website URL without running them on any AI platform, so you can see how buyers might phrase things.
- Self-serve plans: the 50 prompts are settled during project set-up. Once set-up is complete the set is frozen and the same set is asked on every selected platform, cycle after cycle.
- Project and managed engagements: a prompt map designed by hand, with prompt count and cycle agreed in the contract; within a project the cycle and the prompt set stay fixed all the same.
Why the set is frozen
The value of monitoring is in the trend, not in a single snapshot. Only when the same prompt set is asked again on the same platform every 72 hours can a change in mention rate be attributed to the AI platform, its sources or your own remediation rather than to a changed question. After the freeze there is no ad-hoc rewording, no burst sampling and no re-asking outside the cycle. The prompt version is one of the fields stored with every observation, so the wording used in any cycle can be traced back.
Why unbranded prompts are the core
Real exposure is whether the AI mentions you when the buyer does not know you exist. If all 50 prompts named the brand, the AI would almost always recognise it and the numbers would look good while meaning nothing. That is why 44 prompts are discovery and comparison questions written from the buyer's point of view with no brand name in them, and the 6 fact checks answer only one question, whether what the AI says about you is correct, without entering the mention-rate denominator.
- What if I realise after the freeze that a prompt is badly written?
- Self-serve plans do not offer mid-term edits. If the set genuinely needs redoing, contact us via /en/contact; completed cycles are never rewritten. Project contracts can define a change process and version numbering for the prompt set.
- Can I have more or fewer prompts?
- Self-serve plans are fixed at 50 prompts, with no per-prompt surcharge and no per-prompt discount. Project and managed engagements can agree a different count in the contract.
- Is each prompt asked on more than one platform?
- Yes. Every selected platform runs the same 50 prompts, each once per platform by default; with three platforms on the cross-platform plan one cycle is 150 planned observations.
Last updated . The capabilities page and the pricing page are authoritative where they differ.
Help centre
Back to the index, or email contact@miaowageo.com.