One AI answer is not measurement
AI can return different brands, rankings and recommendations when the same buyer question is repeated. Trustidex uses defined buyer-need test sets and repeated measurement to identify the pattern.
Follow the measurement- 1
Define the test set
Each assessment defines the same factual inputs before measurement begins.
- Buyer needs
- Decision questions
- Competitors
- Market
- Language
- Platforms being measured
- 2
Repeat the measurement
Trustidex repeats buyer-need tests across the selected AI platforms, markets and languages. A single response is treated as an observation. The repeated pattern is the measurement.
- 3
Classify the outcome
Each result is assigned a commercial outcome so appearing in an answer is separated from winning the recommendation.
- Recommended
- Shortlisted
- Mentioned
- Absent
- Competitor recommended instead
- 4
Measure consistency
Recommendation Probability™ shows how consistently AI recommends the company for a defined buyer need.
- Buyer need
- Competitor
- Platform
- Market
- Language
- Measurement date
- 5
Diagnose competitor displacement
Trustidex examines why a competitor was preferred and prioritises the evidence gaps. The company or its agency implements the changes.
- Missing evidence for a buyer requirement
- Weak independent validation
- Stronger trusted-source support for a competitor
- Mentioned but excluded from the final recommendation
- Material differences by platform, market or language
- Available evidence not recognised for the buyer need
- 6
Estimate revenue at risk
Revenue at risk is an estimate, not a confirmed loss. Trustidex uses documented commercial inputs and disclosed assumptions, presents a range, and does not assume that one AI answer equals one lost deal.
- 7
Version and validate
Each measurement creates a dated record of the defined test set and its results. After changes are implemented, the same tests show whether Recommendation Probability™ improved, declined or remained unchanged.
The measurement record can compound
Over time, versioned records can build longitudinal knowledge about which evidence changes are associated with improved recommendation outcomes.
What the assessment shows
- Recommendation Probability™
How consistently AI recommends you for a defined buyer need
- Competitor displacement
Where and when another company is recommended instead
- Revenue-at-risk range
What may be commercially exposed
- Priority evidence gaps
What should be addressed first
- Validation result
Whether the recommendation outcome changed after implementation and re-testing
We show ranges, disclose assumptions and never present an estimate as a confirmed loss.
Establish the baseline first
Then decide whether recurring measurement is warranted.