What an AI visibility audit tells you
Before spending on answer engine, generative engine or AI Overviews work, it is worth knowing three things: whether AI systems mention your brand at all when buyers ask about your category, whether what they say is accurate, and whether any of that currently sends you traffic. Most businesses have no idea, and most tools that claim to tell them present a sample as a score. The AI visibility audit, one of the four services under my AI search optimization hub and part of my SEO services, answers those questions with a repeatable method: a fixed prompt set, sampled across the main systems from your target location, recorded, and re-run on a schedule so change can be seen. It reports what is measurable as fact, what is sampled as evidence with the method stated, and does not report what cannot be measured.
Who this is for
- Marketing leads asked by management “are we showing up in ChatGPT” who need an evidence-based answer.
- Brands that suspect an assistant is describing them wrongly or recommending a competitor.
- Businesses in Australia, the UK and the USA watching informational clicks fall and wanting to know which queries now carry an AI answer and who is cited.
- Companies deciding whether AI search work is worth a budget line this year.
What the audit measures
Measured directly
- AI referral traffic: sessions, engaged sessions and conversions from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar sources in GA4, with a channel group configured so they are visible every month afterwards.
- AI crawler activity: which AI crawlers (GPTBot, PerplexityBot, ClaudeBot, Google-Extended, Applebot-Extended and others) visit the site, how often and which pages, from server logs where the host provides them, and what robots.txt currently allows.
- Foundation checks: rankings for the underlying pages, structured data validity, and entity consistency (name, address, description, founders, products, prices) across the site, Google profile, LinkedIn and directories.
Sampled with the method stated
- Prompt set: thirty to a hundred prompts built from your keyword map and buyer questions: category recommendations (“best X in Y”), comparisons, “is X good”, how-to and cost questions, and brand-name prompts.
- Presence and accuracy: each prompt run in ChatGPT, Perplexity, Gemini and Claude and checked for Google AI Overviews and AI Mode from your target country, recording whether you are named, what is said, which sources are cited, and which competitors appear.
- Share of voice: how often you versus each competitor appears across the set, reported as a trend between samples rather than a precise figure, because answers vary by session, location and model version.
- Source map: the third-party pages cited most often for your category, which is the input for the generative engine optimization work.
Not reported
AI Overview impressions or clicks as a separate line (Google does not provide them), how many people saw your brand in an answer without clicking, a stable “AI ranking”, or attribution of a sale to an answer read a week earlier. If a report shows you those, it is guessing.
How the audit runs
- Scope. Target markets, competitors and the prompt set agreed with you, drawn from your keyword map and the questions your sales team hears.
- Direct measurements. GA4 channel group configured, logs analysed, robots.txt and structured data reviewed, entity facts compared across profiles.
- Sampling. The prompt set run across the systems from the target location, with answers and citations recorded. This is done by hand and with tooling where it is reliable, and the method is documented so it can be repeated exactly.
- Analysis. Presence, accuracy, competitors, cited sources and the gaps between what you publish and what the systems draw on.
- Report and recommendations. What to fix first, which of the other AI search services would move the results, and what to expect.
- Tracking (optional). The same prompt set re-run monthly or quarterly, with a trend report alongside referrals and rankings.
A baseline audit takes five to ten working days depending on the prompt set and the number of markets.
What you receive
- The baseline report: referral data, crawler activity, foundation findings, prompt-set results with presence, accuracy, competitors and cited sources.
- The documented method, so anyone can repeat the sample.
- Prioritized recommendations, with a clear statement of what is measurable afterwards.
- With tracking: a monthly or quarterly trend report and the GA4 setup for AI referrals.
The audit is priced as a project, quoted on request and tracking is included at the Growth and Enterprise tiers on the SEO packages and pricing page. The work it usually leads to is described on the answer engine optimization and AI Overviews and AI Mode optimization pages, and the technical fixes it surfaces go through my technical SEO service.
Next step
Request a free AI visibility check: I will run ten prompts for your category across the main systems and reply with where you appear, where competitors appear and whether a full audit is worth it. Everything else is under SEO services.