How we measure · The instrument
ARES-C
An instrument, not an opinion.
ARES-C is the instrument behind everything Pitot measures: a fixed, versioned framework that scores how AI systems perceive a brand. The instrument does not move. The brands, the models and the moment are what it measures.
The five pillars — as one shape
Click a dimension to explore
↓
01 · Authority
Authority
How heavily the machine weights you when it answers.
A citation in an independent publication outweighs a hundred self-descriptions.
What we measure
The density and independence of sources that reference your organisation across the domains the machine treats as authoritative.
- Citation breadth across independent publications
- Source authority — peer-reviewed, archival, institutional
- The shape of the conversational graph around your name, across topics and sectors
- Cross-model consensus on which references the machine returns first
Why it matters
The machine assigns weight not to what you say about yourself but to what others have said about you in contexts you did not control. Organisations with thin Authority are frequently accurate in the machine's mouth — but never recommended.
How to repair
Authority is the slowest dimension to build because it sits outside your perimeter. Repair begins with mapping the sources the machine already trusts in your category, identifying which of them carry none of your signal, and establishing a sustained contribution discipline that corrects the asymmetry over years — not quarters.
02 · Reputation
Reputation
Not whether you appear. How the machine describes you when you do.
Recommended versus mentioned. Leader versus option.
What we measure
The sentiment, warmth and confidence with which AI describes you, read across systems and across time.
- Tonal register across systems when the machine is asked about you
- Confidence calibration — whether the machine hedges or asserts
- Competitive framing — returned as the reference, or as one of several
- The trajectory of sentiment across the last thirty-six months
Why it matters
A brand can be present in every relevant answer and still lose every relevant decision, because the machine describes it in the wrong tonal register. The difference between a leading firm in the category and one option in the category is not semantic. It is commercial.
How to repair
Faster than Authority, harder than it looks. It requires synchronising the tone across every surface the machine reads — press, site, partnerships, interviews, published work — so that the register the machine eventually adopts is the one the brand itself would choose, if given the choice.
03 · Expertise Visibility
Expertise
Whether your real depth leaves a machine-readable trail.
Invisible expertise is, operationally, no expertise at all.
What we measure
The gap between the expertise you actually possess and the expertise the machine can see.
- The distance between practised expertise and published expertise
- Structured topical representation — whether your real specialisation is visible in the retrieval graph
- Depth of representation per named capability
- The quiet-expert problem — senior figures whose work is consequential but machine-invisible
Why it matters
Many of the most accomplished organisations and individuals are read by the machine as thinner than they are — because their work lives in rooms, projects, client engagements and peer relationships, none of which are machine-readable. In an AI-mediated world, the expertise that cannot be surfaced cannot be chosen.
How to repair
The most rewarding dimension to repair, because the material already exists. The intervention is not to generate expertise but to catalogue the expertise that was never catalogued — publicly, independently, and at a depth the machine will weight.
04 · Signal Consistency
Signal Consistency
Not a score of level but of stability — whether you are the same brand across every model, every paraphrase, and over time.
AI aggregates patterns. Inconsistent signals are diluted into noise.
What we measure
The structural coherence of your public record — not its volume.
- Consistency of positioning across the full history of public expression
- Coherence across channels: site, press, product messaging, partnerships, interviews
- Drift across management transitions
- The ratio of signal to noise, measured against itself over time
Why it matters
In the language of the machine, inconsistency becomes invisibility. The brand that has said one true thing about itself for a decade outranks the brand that has shouted a dozen different things loudly.
How to repair
The hardest dimension, because it is cumulative — the record is the record. The work is not to fabricate retroactive coherence but to establish a new sustained coherence going forward, clean enough and maintained for long enough that the machine eventually weights the new pattern more heavily than the old.
05 · Context
Context
The frame through which the other four are scored.
Wrong frame, wrong reading.
What we measure
Whether the machine has placed you in the right category, with the right peers, against the right competitive set.
- Category assignment across models
- Peer-set composition — who you are being compared to
- Frame accuracy — read as premium, volume, niche, or misclassified entirely
- Misalignment between your intended frame and the machine's resolved frame
Why it matters
A luxury manufacturer read as a volume producer, a premium service firm read as a commodity provider — these are context errors, and they quietly rewrite every other reading the machine produces.
How to repair
Often the fastest to diagnose and the most strategic to correct, because a small upstream shift in how you are framed can restore correct interpretation across hundreds of downstream readings. The repair begins with locating where the frame broke — usually a category assumption made years ago, on a surface nobody has been monitoring since.
— · The build
Built like an instrument
A measurement that wants to be believed has to behave like one.
Most of the score is read by extraction, not judgement.
Anchored to behaviour
ARES-C is anchored to observable AI behaviour — is the brand named, recommended, described correctly — not to taste. Most of the score is read by extraction, not judgement: counted, reproducible, hard to argue with.
Where judgement is unavoidable
A fixed judge is used, and its agreement with trained human raters is published. Model versions are pinned and re-checked against control items every cycle, so a brand's movement is never confused with a model's drift.
Always with an uncertainty band
Every published number carries a stated uncertainty — 72 ± 4, never a false 72.3.
The integrity layer
A public index that names competitors is only worth anything if it cannot be bought.
- A brand cannot buy a better Index score. It can only commission an Audit that diagnoses one. The wall between the public benchmark and paid work is structural, not promised.
- Every cycle is pre-registered. The method and the brand universe are locked before the run, so nothing can be cherry-picked after.
- Reliability is published, not asserted.
- Brands may challenge a disputed score, on the record.
A published method, published reliability, pre-registration and a visible conflict-of-interest wall are, together, the part no competitor can copy.
Current methodology · ARES-C 2.0
The delivery
The instrument is delivered as an Audit.
Where the Index reads a whole sector at a glance, the Audit reads one brand to the floor — scored across the five pillars, configured to your market, your customer, your purchase context.