Measure · The Audit · Operational

The Audit

What AI tells your buyers — and where it's wrong.

A Pitot Audit is a bespoke, in-depth diagnosis of how AI perceives a single brand against its named competitors. Where the Index reads a whole sector at a glance, the Audit reads one brand to the floor — configured to your market, your customer, your purchase context.

Commission an Audit → See the process ↓
The Audit · multi-pass diagnosis · triple-checked for stability
W1 W2 W3 W4 W5 W6 Phase 01 Declared Position Week 1 · 2 Phase 02 Machine Reading Week 2 · 5 Phase 03 Strategic Interpretation Week 4 · 6 The Reading delivered
  1. Phase 01 · Week 1–2 Declared Position Client questionnaire · stakeholder interviews
  2. Phase 02 · Week 2–5 Machine Reading ARES-C audit · 14 systems · named cohort
  3. Phase 03 · Week 4–6 Strategic Interpretation Report · briefing · working session
  4. Terminal The Reading Delivered · Week 6
Click a phase to read ↓

Declared Position.

What you say you are — before the machine has spoken.

The starting brief is the self you would like to be read as. It is not yet a reading.

What we do
A structured capture of the organisation's declared position: category claim, priority audiences, strategic narrative, the distinctions leadership intends to carry.
  • Leadership questionnaire and internal narrative audit
  • Structured interviews with up to four senior stakeholders
  • Review of brand architecture, published material, recent campaigns
  • Baseline capture of the intended positioning, committed to in writing
Why it matters
Without a declared position, the reading in Phase 02 is floating — a measurement with no referent. The audit's central move is the distance between these two surfaces. The declared self is the axis against which the machine's reading is scored.
Deliverables
Baseline brief, stakeholder alignment map, initial gap hypothesis — the document the audit measures against. Issued at end of Week 2.

Machine Reading.

What the instrument actually sees.

Ten frontier systems. One prompt library. Five dimensions. One named cohort.

What we do
The full ARES-C audit, executed as a controlled instrument run — not an impression exercise, not a sample.
  • Ten frontier AI systems, queried on fresh sessions — no conversation history contaminates the reading
  • The full prompt library, structured across the five ARES-C dimensions
  • Benchmarked against a named competitive cohort agreed with leadership
  • Three concurrent scoring passes per response, majority rule on dispute, median on ties
  • A full evidence ledger — every reading is reproducible, the raw record survives the run
Why it matters
This is the machine-truth layer — the portrait the frontier systems have already assembled, irrespective of what any PR cycle or brand deck says. Every strategic recommendation downstream inherits from here. Get this layer wrong and the rest of the work advises on the wrong organisation.
Deliverables
ARES-C scorecard across the five dimensions with coherence index, cohort comparison, risk register, raw evidence ledger. Issued at end of Week 5.

Strategic Interpretation.

The point at which the instrument yields to judgement.

The audit's value is the interpretation, not the data.

What we do
Senior-led synthesis of the machine reading against the declared position. Not a longer report — a shorter one.
  • The dimensions that demand intervention, and which can safely wait
  • The sequencing of the intervention — commercial logic, not an instinct list
  • Reframe questions leadership must answer before execution begins
  • Where the nearest, highest-value corrections sit — the first moves that pay back fastest
  • In-person briefing and a working session with leadership close the engagement
Why it matters
Organisations that act on the machine reading alone — without the synthesis — tend to optimise the wrong dimension first. The human layer is not ornament. It is the step at which a reading becomes a decision.
Deliverables
Written report, in-person briefing, working session with leadership, priority action plan. The Reading, delivered. End of Week 6.
The finding most firms never see

AI does not only overlook brands. It invents things about them.

Prices, policies, provenance, ownership — stated with total confidence to a buyer who has no reason to doubt. A brand described wrongly by a system millions consult is exposed to a risk it cannot see and is not measuring. Surfacing that misattribution is a first-class outcome of every Audit, not a footnote. Every recommendation is tied to a mechanism — change this, and here is the modelled effect on what AI says — not opinion. Findings are triple-checked for stability, and disagreement between models is reported as a result in its own right.

Up