AI answers vary. Our measurement makes that variability visible.
RankMeasure does not treat a single AI answer as a ranking. We use documented buyer prompts, sampled answer runs, intent weighting and repeatable scoring to measure recommendation patterns across ChatGPT, Claude, Gemini and Perplexity.
Multiple answer runs
Important prompts are sampled repeatedly because generated answers can vary from run to run.
Buyer intent matters
Commercially important prompts carry more weight than broad informational questions.
Confidence is visible
We report consistency rather than hiding uncertainty behind one precise-looking score.
From buyer question to recommendation score.
The same documented sequence is applied to each benchmark so results can be interpreted and repeated over time.
Prompt pack
20–25 buyer questions mapped to category, service, location and commercial intent.
AI engines
Prompts are run across ChatGPT, Claude, Gemini and Perplexity.
Sampling
High-intent prompts receive repeated answer runs where required.
Classification
Business presence, rank, mentions, citations and competitors are captured.
Scoring
Intent weighting and engine results are combined, with confidence shown separately.
Prompt packs
A Founding Benchmark uses approximately 20–25 prompts. Prompts cover a range of intent levels and may include service, location, comparison and high-value purchase questions.
Intent weights
Each prompt can carry a commercial-intent weight. Higher-intent questions contribute more strongly to the overall recommendation score.
Money prompts
Money Prompt Share is calculated only from prompts with an intent weight of 0.7 or higher. This separates commercially important visibility from broad informational visibility.
Recommendation presence scale
The internal scoring model distinguishes a leading recommendation from a passing mention or an absence.
Aggregation
Prompt and engine contributions are combined using the benchmark's documented weights. For repeated samples, RankMeasure uses the median result rather than treating one unusually strong or weak generation as representative.
AI Recommendation Score (ARS)
The resulting benchmark score is normalized to a 0–100 scale. The score is intended to describe observed recommendation presence under the documented sample conditions — not a guaranteed future ranking.
Confidence thresholds
Confidence reflects how consistently the sampled runs produce the same directional recommendation result.
Why this matters
Two businesses can have the same recommendation score but very different stability. Confidence prevents an unstable pattern from being presented as if it were deterministic.
Free Snapshot: Initial Signal
The free experience uses approximately 6–8 prompts across four AI engines. Because the sampling depth is different, its score is labeled Initial Signal, not the full ARS.
Founding Benchmark: full scoring
The £149 benchmark expands the prompt pack, applies intent weighting, repeats higher-intent prompts, calculates AI Recommendation Score, Money Prompt Share and Confidence, and includes the prioritized action plan.
Scan metadata
What RankMeasure does not claim.
Not a deterministic ranking
AI systems generate variable answers. A benchmark describes sampled recommendation behavior under documented conditions.
No guaranteed outcome
Neither RankMeasure nor any third party can guarantee that an AI assistant will recommend a specific business in a future answer.
Not traditional SEO position tracking
The product measures recommendation visibility in generated answers, not a fixed position in a conventional search results page.
Want to see the methodology in a real report layout?
Explore the sample benchmark or start with a free recommendation snapshot.