This methodology page documents how an AI visibility audit is designed and executed. It exists so buyers can evaluate the rigour of any audit they receive and so teams can replicate the approach internally.
Principles
The audit is built on three principles: measurement before opinion (baselines come from runs, not impressions), fixed conditions (prompts and platforms are held constant so changes are attributable), and transparency (every finding traces to evidence the client can re-run).
Step 1 — Prompt portfolio design
The audit defines 20-50 prompts spanning the client’s commercial intent: category questions, comparison queries, solution questions and brand-specific prompts. Each prompt records its intent and expected answer structure. The portfolio is the fixed instrument for the entire engagement — changing it between runs invalidates comparisons.
Step 2 — Baseline capture
Each prompt runs against the agreed engines (typically ChatGPT, Gemini, Perplexity, Google AI Overviews and any market-relevant regional platforms), with model versions recorded. For every answer the audit records: brand presence (cited, mentioned, absent), citation type, source pages cited, and competitor presence. Results are stored as raw runs, not summaries.
Step 3 — Technical diagnosis
In parallel, the audit checks the technical estate: AI crawler access and behaviour, robots.txt rules, rendering, sitemap hygiene, structured data coverage, and entity consistency across properties. Each finding is graded by impact and effort, mirroring the audit checklist.
Step 4 — Content and authority diagnosis
Priority pages are evaluated for answer-first structure, extractability and intent alignment. Authority signals (authorship, original data, link profile quality, third-party accuracy) are assessed without fabricating standards — where no benchmark exists, the audit says so.
Step 5 — Gap analysis and prioritisation
Findings are mapped to workstreams and prioritised by expected impact on retrievability and citation rate. The output is a scored report with a sequenced roadmap: technical fixes first, then content, then authority — because later workstreams depend on earlier foundations.
Step 6 — Reporting and handover
The audit delivers the raw run data, the scored checklist, the roadmap, and the baseline dashboard configuration so the client can continue measurement. This handover is what makes the audit a starting point rather than a report that gathers dust.
Limitations
An audit measures a point in time under fixed conditions. Model updates between runs are recorded, not hidden. No audit can promise citations — it can only establish where you stand and what to do next. The audit service applies this methodology; GEO reporting continues the measurement afterwards.