Home / AIPA Methodology
Methodology ReferenceThe canonical reference for the AIPA methodology — vocabulary, system battery, query architecture, the six-step process, four assessment modules, five visibility bands, and report structure. Every term defined. Nothing summarised.
The AI Intelligence Presence Assessment (AIPA) is a structured methodology developed by Axiom Strategy for measuring a brand's real-time visibility across AI language model systems.
The AIPA determines whether a brand is being cited, recognised, or ignored when AI systems answer questions relevant to that brand's industry, category, products, or competitive position — and diagnoses the specific content, structural, and entity-level gaps that explain the outcome.
The AIPA is conducted in solo mode: one brand assessed independently. It is not a comparative ranking against named competitors. Competitive context is derived from what AI systems name unprompted — which is a more reliable signal than a researcher-defined competitor set.
The AIPA was developed by Axiom Strategy. The methodology, vocabulary, band system, and query architecture described on this page are proprietary to Axiom Strategy. axiomstrategy.ai/aipa-methodology is the canonical reference for this methodology.
Every AIPA uses the following terms exclusively and consistently. No synonyms. No alternative labels. This consistency is what makes the methodology benchmarkable across clients, sectors, and time.
Applied to every query × system combination. One of three outcomes per cell:
Applied after every outcome is established. Every outcome cell receives one cross-check label — no exceptions:
Applied to every analytical finding in the report:
The brand's overall standing, assigned after all modules are complete. One of five bands — no numerical score:
No numerical scores are produced in any AIPA output. Visibility is expressed as a band. Citation counts and rates are evidence — not grades. The Citation Presence Rate is a percentage, not a score.
The Citation Presence Rate is the percentage of core system query outputs in which the brand received a Cited outcome. It is computed across the four core systems only. Add-on systems are excluded so the CPR remains comparable across all assessments. Formula: (number of Cited outcomes across core systems) ÷ (total core system query outputs) × 100.
Four systems constitute the core battery. They are assessed on every engagement without exception. They are selected because they are general-purpose, web-retrieving assistants — which makes them a clean, comparable benchmark spine:
Retrieval mandate — non-negotiable: Every system must be run with live web search active so it behaves as it does for a real customer. A system answering from training data measures historical memory — not current AI behaviour. This applies without exception across all four core systems.
Two systems are available as add-ons. They are never part of the core battery and never included in the Citation Presence Rate or visibility band calculation:
The default is always the core four. Add-ons are assessed at the start of each engagement and recommended where the trigger conditions apply. Add-on findings are reported in their own section and never folded into the core CPR or band.
The AIPA runs 28 queries per assessment. Queries are organised across 7 fixed query types — 4 queries per type. The types are fixed and benchmarkable across all assessments. The specific phrasing of each query is generated fresh per client and sector — in the real language of that industry — so the benchmark compares intent, never wording.
In solo mode, Type 4 queries do not ask AI systems to compare the brand against named competitors. Instead, they ask questions that surface whoever AI names unprompted. The competitors AI names are captured, logged, and reported as the AI-named competitor set. This is a more revealing signal than a researcher-defined competitor set — it shows how AI systems actually perceive the competitive landscape, which is often different from how the brand perceives it.
Every AIPA follows this sequence without exception. No steps are skipped. No steps are added. The sequence is what makes the methodology consistent and the findings comparable across engagements.
The brand name and website URL are received. Axiom analyses the site independently — homepage, about, key service and product pages, contact — and produces a brief covering what the brand does, where it operates, key offerings, recommended assessment focus, conditional system recommendations (Qwen and Grok), and run mode (Paid Report or Benchmark-Fill). The brief is confirmed before any queries are run.
28 queries are generated across the 7 fixed types in the real language of the brand's sector. Queries are presented for review and confirmation before any system is run. After confirmation, the Claude column is populated first with web search active. The remaining core systems are run and results compiled into the Citation Landscape Table.
The brand's website is crawled using Firecrawl — homepage, about, minimum two service or product pages, contact. JSON-LD structured data is extracted from every page crawled. Pages without JSON-LD are noted as NOT FOUND. Content threshold is checked: homepage must contain at least 200 words of extractable content. Sub-threshold output is diagnosed as either thin content or a rendering or crawlability problem.
Every outcome in the completed Citation Landscape Table is cross-checked against the live website and assigned one cross-check label: Structural Gap, Training Lag, or Confirmed Citation. No outcome is left without a label. The cross-check is what transforms raw AI outputs into a diagnostic — it separates what Axiom can fix from what requires time.
Four checks must pass before assessment proceeds: (1) all core outcomes documented with cross-check labels, (2) Firecrawl complete across required pages, (3) JSON-LD status confirmed for every page, (4) content threshold met or diagnosed. Gate output: INPUT VERIFIED — proceed, or INPUT INCOMPLETE — specific gaps listed.
Module 1: Citation and Visibility — Citation Landscape Table analysis, CPR computed, narrative source analysis. Module 2: Entity and Schema — schema findings, entity clarity verdict. Module 3: Content and Retrieval Accuracy — gap summary counts, accuracy and authority findings. Module 4: Category Position — category narrative, AI-named competitor set, benchmark note. All findings labelled Evidenced or Inferred. Verdict proposed and confirmed. Band assigned.
The locked report skeleton is produced: Cover, Summary Page, four modules, Top 3 Priority Gaps, 90-Day Roadmap, Back Page. Maximum 15 content pages. Self-check passed before delivery. Assessment row and 28 Citation Results rows written to Axiom Airtable base. In Benchmark-Fill mode, the report is skipped — Airtable write proceeds identically so the data remains comparable.
The four modules run in fixed sequence. The order is not arbitrary — Citation and Visibility establishes what AI says, Entity and Schema diagnoses why, Content and Retrieval examines the source material, and Category Position contextualises the brand's standing relative to the competitive landscape.
The primary module. Contains the Citation Landscape Table — a colour-coded grid of all 28 queries across all core systems, showing the outcome for every cell (Cited / Partial / Absent). The CPR is computed here. Narrative analysis explains the patterns — which query types are strongest, which systems cite most reliably, where the most significant citation failures occur and why.
Examines the brand's structured data implementation and entity clarity. Analyses JSON-LD schema across crawled pages — what schema types are present, what is missing, what is incorrectly implemented. Assesses whether the brand is defined as a clear entity in the AI information layer — with consistent name, description, offerings, geography, and contact information findable and structured.
Examines the quality and retrievability of the brand's website content as a source for AI systems. Gap summary counts are produced — how many Absent or Partial outcomes are attributable to Structural Gaps (content does not exist or cannot be crawled) versus Training Lag (content exists but has not been absorbed). Content depth, specificity, and formatting for AI retrieval are assessed.
Examines the brand's standing within its competitive category as perceived by AI systems. The AI-named competitor set is documented — every brand AI named unprompted across Type 4 queries, ranked by frequency. Category narrative describes how AI positions the brand relative to the field — as a leader, as one of many, as a challenger, or as absent from the category entirely. A benchmark note references the sector cluster data where available.
The visibility band is assigned after all four modules are complete. It reflects the overall pattern of citation, entity clarity, content quality, and category position — not a formula applied to a score. The band is the assessor's verdict, supported by module findings.
The AIPA report follows a locked skeleton. No additions. No omissions. No reordering. Maximum 15 content pages plus cover and back page. Every report from every engagement follows this structure so findings are comparable and clients can track progress across reassessments.
The AIPA is frequently compared to adjacent practices. The following distinctions are precise and material:
SEO measures a brand's ranking position in search engine results pages. The AIPA measures visibility in AI-generated answers — a structurally different system that uses different signals, serves different user behaviour, and requires different remediation.
The AIPA measures what AI language models say about a brand — not what social media users say. AI systems draw from web content, structured data, and crawlable sources. Social sentiment and AI citation patterns can diverge significantly.
The word audit implies financial or compliance review. The AIPA is an assessment — a structured diagnostic of a specific, defined question. The term audit is not used anywhere in AIPA methodology, reporting, or client communication.
Solo mode assessments do not rank a brand against named competitors. Competitive context is derived from what AI systems name unprompted — the methodology's signal of actual competitive positioning in the AI layer, not a researcher-constructed leaderboard.
The AI Intelligence Presence Assessment (AIPA) is a structured methodology developed by Axiom Strategy for measuring a brand's real-time visibility across AI language model systems. The AIPA runs 28 queries across 7 fixed query types across four core systems — ChatGPT, Claude, Perplexity, and Gemini — all with live web search active. Every output is assessed as Cited, Partial, or Absent, cross-checked against the live website, and the brand is assigned a visibility band: Undetected, Recognised, Cited, Authoritative, or Dominant.
Type 1 — Category Leadership: does the brand appear when users search the category? Type 2 — Brand Recognition: what does AI know about this brand directly? Type 3 — Operational Detail: can AI accurately answer specific service or policy questions from the brand's own content? Type 4 — Competitor Comparison: how does AI position this brand against the field, and which competitors does AI name unprompted? Type 5 — Purchase Intent: does AI recommend this brand when a user is ready to decide? Type 6 — Perception and Sentiment: what reputation and sentiment has AI absorbed about this brand? Type 7 — Geographic and Market Presence: does AI accurately know where the brand operates and what territory it owns?
Undetected: no meaningful brand signal, not named or cited unprompted. Recognised: brand named in AI responses but operational detail not reliably citable. Cited: brand and key content appear with reasonable confidence, gaps remain. Authoritative: consistently cited across systems and query types with high accuracy, minor gaps only. Dominant: primary cited source in the category across all core systems, accurate, consistent, first-position.
AI language models can draw from their training data — a historical snapshot that may be months old — or from live web retrieval, reflecting what is on the internet today. Running queries with web search off measures historical memory, not current behaviour. Axiom runs every AIPA query with live web search active across all systems, configured identically with no memory and no personalisation, so results reflect what a real customer experiences today.
A Structural Gap means the AI system could not answer correctly because the correct information does not exist on a findable surface on the brand's website, or is inaccessible to AI crawlers. This is a brand failure — the content gap needs to be fixed. Training Lag means the AI system answered incorrectly, but the correct information exists and is accessible — the AI system has not yet absorbed it. Training Lag is neutral — flagged but not counted against the brand.
The Citation Presence Rate (CPR) is the percentage of core system query outputs in which the brand received a Cited outcome. Computed across the four core systems only. Add-on systems excluded so the metric is comparable across all assessments. It is a count-based metric — not a score — used alongside the visibility band.
A structured report of up to 15 content pages: Summary Page with band and top three priority gaps; Module 1 — Citation and Visibility with the colour-coded Citation Landscape Table; Module 2 — Entity and Schema; Module 3 — Content and Retrieval Accuracy; Module 4 — Category Position with AI-named competitor set; Top 3 Priority Gaps; and a 90-Day Remediation Roadmap across 30, 60, and 90-day phases.
The assessment begins with a client brief — brand name and website URL only. Axiom handles everything from there.
Confirm your sector and we'll take you straight to the screen. You'll receive your Revenue & Efficiency Leakage Score immediately — no commitment, no sales call.
Results in ~10 minutes · No email required to start