OrbTalk to Orb

Prompt panel

How we measure AEO

Visibility without accuracy is a liability. Orb scores answer-engine performance on three axes: found, cited, and correct—using a fixed prompt panel you can repeat over time.

Found — present in the answer or sourcesCited — named or linked as attributionCorrect — description matches reality

Found

Does the answer (or its visible sources) include your brand, product, or a clearly attributable claim? Absence here is a retrieval or entity problem first.

Cited

Are you explicitly named or linked as a source—not merely paraphrased into the prose? Citation is the AEO win condition for many B2B categories.

Correct

When you appear, is the description accurate? Wrong citations are tracked as failures even if “visibility” looks high.

Prompt panel methodology

  1. Define the panel. A stable list of prompts that real buyers ask (category, comparison, “best for”, local, and objection prompts). Freeze wording for a measurement cycle.
  2. Choose surfaces. The engines and modes your audience actually uses. Document versions / dates; models change.
  3. Run & record. Capture full answers and visible citations. Score found / cited / correct per prompt with a short rubric—not vibes.
  4. Aggregate carefully. Report rates and qualitative misses. Do not invent market share or “AI traffic” numbers you cannot observe.
  5. Feed the loop. Misses become audit and content work. Re-run the same panel after changes.

What we refuse to fake

No fabricated percentage lifts, no made-up “share of AI answers,” no anonymized logos implying customers we do not have. If a number is not measured, it does not appear on this site.

Orb self-experiment (Hong Kong)

Orb is running its own AEO experiment on this brand and category. We are applying the same audit → citability → mentions → prompt panel loop described on How we work.

Results: coming. When we have repeatable panel outcomes worth publishing, they will land here—with methodology and caveats. Until then, treat any third-party “AEO leaderboard” claims with skepticism unless they show their prompts.

Want a baseline panel?

Bring your category prompts. We’ll score found / cited / correct and show you the gaps—without inventing benchmarks.

Talk to Orb