What we measure, how we measure it, and why it matters.

Most AI representation services tell you whether AI mentions your brand. We tell you whether AI is accurate about your brand — and we prove it before we report it.

This page explains the methodology behind ARADP at a high level — written for decision-makers who want to understand what they are buying without needing a technical background.

The problem with existing AI monitoring

There are two common approaches to AI brand monitoring available today:

Approach 1 — Mention monitoring

Tools that tell you how often your brand appears across AI platforms. They measure frequency and sentiment but do not confirm accuracy. A brand can be mentioned frequently and inaccurately simultaneously — and mention monitoring will show a positive result.

Approach 2 — Answer Engine Optimisation (AEO)

Services that optimise your content to increase the likelihood of AI mentioning your brand. AEO addresses visibility — whether AI finds you. It does not address accuracy — whether AI describes you correctly when it does find you.

ARADP addresses the third layer that neither approach covers: ground-truth-confirmed accuracy. We do not ask whether AI mentions you. We ask whether what AI says about you is correct — and we verify the answer against your actual published information before reporting it.

The ARADP methodology — three principles

Principle 1 — Ground truth first

Before running a single AI query, we document what your business actually publishes. This becomes the ground truth against which every AI response is tested.

This is the critical difference between ARADP and observation-based approaches. We do not note that AI said something unusual. We confirm that AI said something wrong — wrong against a specific, documented, dated primary source.

Principle 2 — Classification before reporting

Every finding is classified before it appears in a report. We classify against two dimensions: what type of distortion it is, and what can be done about it.

Distortion type is classified against our confirmed taxonomy of many patterns across many diagnostic classes. Remediation is classified using our proprietary CIA framework.

Principle 3 — Revenue quantification

Every confirmed finding is modelled against five revenue loss pathways. Revenue at risk is quantified individually for each finding — not estimated globally — and modelled across three scenarios.

The five pathways are proprietary and provided in full to clients as part of the Step 2 Full Diagnostic engagement.

The distortion taxonomy

ARADP is built on a proprietary taxonomy of many confirmed distortion patterns — specific, named, classified ways that AI misrepresents brands. Each pattern has been identified from a real client diagnostic, confirmed against real published information, and validated across multiple clients and platforms.

Patterns are organised into diagnostic classes covering many specific failure modes.

The full taxonomy is proprietary and client-confidential. It is provided in full to clients as part of the Step 2 Full Diagnostic engagement. The free 60-second scan maps your business against the full taxonomy and identifies the highest-probability patterns before a single query is run.

AI platform behaviour changes continuously as training data is updated. A distortion confirmed today may resolve in six months — or new distortions may emerge. This is why Step 4 monitoring exists and why quarterly re-runs are a core part of the programme.