Skip to main content

How it works

One prompt gives you an answer. Rough Diamonds gives you a measurement.

Asking AI about your company can produce a confident answer. That answer can change the next time you ask.

One answer is not a measurement

AI answers can vary. A company can look strong in one prompt and disappear in another.

That is why Rough Diamonds does not base a result on one question or one answer.

A measurement compares multiple signals. A prompt gives you one reply.

We measure two different things

Before AI can recommend a company, it needs a picture of what that company does. Those are not the same test.

AI Understanding

Rough Diamonds can ask about the company directly, because the goal is to test what AI understands about that company.

AI Recommendation

Rough Diamonds tests recommendation through situations where someone is looking for a relevant product or service. It does not simply ask whether AI would recommend the company by name.

Being understood does not guarantee being recommended.

Understanding is tested directly

Rough Diamonds tests whether the tested AI has a clear picture of the company. That includes identity, category, offer, audience and important facts — and whether that picture stays stable across answers.

  • Company identity
  • Category
  • Offer
  • Audience
  • Important facts
  • Stability of that picture

Semantic profile testing — Identity, offer and audience are tested directly — not guessed from one recommendation reply.

Recommendation is tested through buyer situations

Asking “Would you recommend this company?” is still a brand question. That is not how most buying conversations start.

A buyer usually asks AI for help choosing. Rough Diamonds tests whether the company naturally comes forward in those situations — closer to the moment a potential customer actually uses AI.

Buyer-situation testing — Recommendation is tested in concrete need contexts, closer to a real buying question.

We look for patterns

Multiple tests reduce dependence on one unusually good or bad response. Stability and fact consistency matter because a high score should reflect a clear picture, not contradictory answers. The complete pattern matters more than an individual response.

  1. One measurement, multiple signals

    The score is less dependent on one lucky or unlucky answer.

  2. Semantic profile testing

    Identity, offer and audience are tested directly — not guessed from one recommendation reply.

  3. Buyer-situation testing

    Recommendation is tested in concrete need contexts, closer to a real buying question.

  4. Stability check

    Multiple answers are compared, so one unusually good or bad response has less weight.

  5. Fact consistency

    Important facts are checked across outputs. A high score should not hide contradictions.

  6. Pattern, not one answer

    The complete pattern matters more than any single response.

The result is two scores

AI Understanding

How clearly does the tested AI appear to understand the company?

AI Recommendation

How strongly does the tested AI recommend the company when a buyer asks for providers?

Rough Diamonds reports AI Understanding and AI Recommendation separately. There is no combined overall score.

This is observed behaviour in the tested AI. It is not a prediction of sales, and it is not a claim about every AI system.

What about my company?

A Rough Diamonds Report applies this measurement to your company. It helps show where AI already has a clear picture, where that picture is weak or inconsistent, where your company is recommended, where others come forward instead, and what you can clarify.

  • Where AI already has a clear picture
  • Where that picture is weak or inconsistent
  • Where the company is recommended
  • Where others may come forward instead
  • What can be improved