Agentic Footprint Audit 01.

I measure and orient your presence before AI agents.

When a buyer asks ChatGPT, Claude, Gemini, Perplexity or Mistral to recommend a house like yours, what comes out? Your words? Your codes? Your real peers? Or substitutes that flatten you? The Agentic Footprint Audit tests 5 AIs across 6 intent categories, 100 calibrated questions in French and US English, 500 scored responses, and delivers a 12-chapter dossier in 1 to 2 weeks after scoping.

What the audit covers

One proprietary method. A factual measure of your agentic footprint.

The Agentic Footprint Audit measures the agentic conversation in 1 to 2 weeks after scoping: 5 AIs, 100 calibrated questions across 6 intent categories, 500 scored responses, read by a senior hand.

The format

Agentic Footprint Audit

Measures how the five dominant AIs restitute your house to real buyers, in French and US English. Designed for an executive committee pitch on a short deadline, with a signed reading.

Contractual lead time 1 to 2 weeks after scoping
AI scope 5 AIs: ChatGPT, Claude, Gemini, Perplexity, Mistral Le Chat
Volume measured 100 questions calibrated in 3 layers (50 FR + 50 US English), 500 responses scored
Weighted criteria Presence, Position, Accuracy, Narrative Preservation, Actionability, Sentiment
Asymmetries revealed Linguistic market FR vs EN · Agentic sovereignty Mistral vs US AIs · Competitive distance (peer-set vs tier-confusion)
Deliverable 12 chapters + 2 appendices, executive dossier format (~50,000 words core, ~120,000 words appendices), interactive HTML private + PDF
Action plan 5 baseline actions across the board + 10 to 12 bespoke actions calibrated on the detected asymmetries
Delivery 45-min video conference with Mickaël Tsakiris, optional 2-h executive workshop
Choose if

You want a factual measure of your footprint on the five dominant AIs, on a short deadline, in French and US English, to decide an editorial or agentic trajectory at the executive committee.

The deliverable, in four extracts

Twelve chapters. Six dimensions. Five argued hypotheses. A prioritised action plan.

Glimpses from two real audits — an independent Parisian fashion house and a premium ready-to-wear house — anonymised. Data and findings unchanged.

Section 01 Audit · Parisian fashion house

Executive summary

Global presence 92% House mentioned in responses
Spontaneous Discovery 52% Queries without brand name
Inaccuracies detected 8 Priority corrections
Co-cited competitors 8 Brands captured by LLMs

Radical asymmetry between dimensions where the house is named by the prospect (above 80) and those where it must be found (33 to 53). The agentic acquisition pipeline is the territory to reclaim.

Section 03 Agentic cartography · 6D

Composite score + six independent sub-scores

44 /100 Grade C
  • Discovery
    34
  • Comparison
    53
  • Knowledge
    42
  • Purchase
    45
  • Ethics
    57
  • Loyalty
    46

The shape of the score tells more than its value. A house at 44/100 with Discovery at 34 and Ethics at 57 is not the same house as another flat at 44/100.

Section 05 Signals detected

Three findings structure the action territories

  1. 01

    Failing agentic recruitment.

    When the prospect searches by distinctive signals (without naming the house), Perplexity only finds it 33% of the time. ChatGPT 60%, Claude 66%, Mistral 71%, Gemini 80%. The blind spot is precisely where the non-existent conversion plays out.

  2. 02

    Convergent factual hallucinations.

    The five AIs collectively hallucinate the house's founding date. Several converge on a common wrong date, the others diverge — the correct date appears in no response. A clear signal of a public reference badly indexed on a basic fact.

  3. 03

    Operational lead is invisible.

    E-commerce replatforming delivered, cloud ERP migration completed, Asia-market app with a loyalty card: projects publicly documented by the partners. None of the five AIs returns them spontaneously. An asymmetry between the real operational lead and the documentary density that can be restituted.

Section 09 Prioritised action plan

Quick wins, structural projects, narrative doctrine

  1. Step 1

    Editorial quick wins

    • « House history » page enriched with pivotal dates, JSON-LD Organization + foundingDate
    • Structured FAQ (positioning, signature, independence, governance)
    • llms.txt published, priorities described explicitly
  2. Step 2

    Structural projects

    • Rebuild of flagship product pages (Schema Product + Brand + reviews)
    • Editorial statement signed by leadership on the operational lead (replatforming, ERP, Asia)
    • External sourcing: thematic third-party articles, Wikipedia seeding plan
  3. Step 3

    Agentic narrative doctrine

    • Agentic brand platform: narrative invariants, what agents are allowed to say
    • Re-test on the initial queries, measuring the delta across the six sub-scores
    • Recurring editorial brief sent to priority third-party media

The two referenced audits were delivered in April and May 2026. Brand names, competitors cited by name, precise dates and technology partner names are anonymised or aggregated. Data and findings remain those of the real audits.

The six audited dimensions

Six independent lenses. Six precise business questions.

Each dimension matches a real query type from a prospect or a client. Reading performance dimension by dimension reveals where the house shines and where it falls off — an average score masks that asymmetry.

DISCOVERY COMPARISON KNOWLEDGE PURCHASE ETHICS LOYALTY unintentional recruitment positioning vs peers history, codes, doctrine acquisition modalities commitments, perception standing, programme, after-sales
Six axes, six client questions, six readings. The radar is the outline — the report delivers the silhouette for your house.
  1. 01

    Discovery

    « Are you cited when searching your category? »

    25%
  2. 02

    Comparison

    « What does the agent say when comparing you to a competitor? »

    15%
  3. 03

    Knowledge

    « What the agent knows — and what it invents. »

    20%
  4. 04

    Purchase

    « Does the agent put you on the transactional path? »

    15%
  5. 05

    Ethics

    « Do your CSR commitments arrive intact? »

    15%
  6. 06

    Loyalty

    « What does the agent recommend once the first sale is closed? »

    10%
Best of both worlds

The power of the algorithm. The knowledge of the sector. Combined.

An automated GEO audit delivers a dashboard. A senior consultant delivers an opinion. In luxury, neither is enough on its own. NarrativeGuard, my proprietary method, harvests what no human eye could cross-reference at this scale; and it's me, Mickaël Tsakiris, who reads what no algorithm could interpret. The method is the product.

The proprietary method

NarrativeGuard

My proprietary LUXE ÆTERNAI method: a workflow that orchestrates a pool of agents. They put the 100 calibrated questions to the five AIs, cross-reference the answers across several passes, and score each against a grid of six weighted criteria (Presence, Position, Accuracy, Narrative preservation, Actionability, Sentiment). The value isn't in a tool — it's in the method: what I choose to measure, how I calibrate it, and the reading I draw from it.

  • Puts the structured queries to the five AIs across multiple passes
  • Scores every answer against six proprietary weighted criteria
  • Cross-references the answers, detects convergent and divergent patterns
  • Archives the raw, timestamped dataset, reusable for your subsequent audits
Sector knowledge

Mickaël Tsakiris

Twenty years of digital in luxury (LVMH, Kering, Razorfish, WIDE, Australie GAD). Seven years in Gen AI. I'm the one who decides what the figures mean for your house — not another. Five argued hypotheses, a prioritised plan: what no algorithm can formulate alone.

  • Reads the gaps between what you build and what the agents retain
  • Formulates five argued, sourced, ranked hypotheses
  • Sizes each workstream in effort × impact, proposes an owner type
  • Delivers to the Exec Board, defends the arbitrations, supports the follow-up
The flow

Three steps. Your inputs feed the audit, not the other way.

  1. 01 Scoping

    Your inputs define the scope.

    Workshop with your team (Executive Committee, CMO, CX Director, e-commerce Director). You validate the peer-set (real competitors, tier-confusion brands, aspirational references), the signature house codes and the calibrated questions. For the Agentic Footprint Audit, composition follows the proprietary 3-layer method: 50% strictly generic common base (cross-house comparability), 25% peer-set specific to your house, 25% core business validated with your teams. Without this step, the audit measures the wrong ground.

    • FormatVideo workshop
    • DeliverableSigned scoping note
  2. 02 Collection

    NarrativeGuard takes over the harvest.

    Five AIs (ChatGPT, Claude, Gemini, Perplexity, Mistral), queried across several time windows to neutralise generation randomness. Proprietary scoring weighted by prompt category. Raw responses archived in appendix, rigorous attribution (AI, language, prompt identifier), complete documented dataset. Mechanical phase, no client interaction.

    • Flash scope5 AIs, 100 questions × 2 languages
    • DeliverableFull raw dataset
  3. 03 Analysis & delivery

    Expertise turns data into decisions.

    Critical reading of the dataset, formulation of the five argued hypotheses, prioritised action plan. Delivery to the executive committee: Q&A format, defence of the arbitrations, follow-up included.

    • DeliveryExec Board presentation + follow-up
    • Deliverable12-chapter dossier + action plan
After the audit

Three possible next steps. None imposed.

→ Path 1

You bring it in-house.

The calibration call is included. You keep the dataset, the report, the plan. Your teams execute. Many houses run the audit with no immediate follow-up, and come back later to measure the impact.

→ Path 2

You hand me the doctrine.

AI Doctrine & Roadmap offer. We translate the audit's hypotheses into an agentic brand platform — what narrative agents are allowed to carry, what client experience we delegate to them, what data we open.

→ Path 3

You take a retainer.

Monthly Retainer offer. Regular strategic presence to steer execution of the plan, arbitrate the workstreams, recalibrate quarterly. A single point of contact, engaged over time.