What an AI visibility audit covers and what it costs
An AI visibility audit is a structured assessment of how AI engines, including ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overviews, describe and recommend a company when buyers ask them questions. Buyers now ask AI assistants for shortlists, comparisons, and recommendations, and the answers are assembled from whatever those engines have read and trust. An audit measures where a company stands in those answers, why, and what to fix. This guide covers what a credible audit includes, what drives its cost, and how an audit relates to an ongoing generative engine optimisation (GEO) program.
Third Hemisphere is a B2B technology PR and communications agency with offices in Sydney, Melbourne, and Singapore. The agency runs a 92-question CRISP audit through its AI Authority & Visibility service and has published AI visibility case studies for Fremantle Seaweed, NEONNOW, and an ASX-listed lender.
What does an AI visibility audit actually measure?
An audit answers a plain commercial question: when a prospective customer asks an AI engine about your category, do you appear, is the description accurate, and who gets recommended instead?
The mechanics differ from a traditional search audit. AI engines synthesise answers from many sources rather than ranking pages, so visibility depends on how consistently a company is described across the web, which publications and databases the engines cite, and whether the company's own site is structured for machine reading. An audit tests all of this empirically rather than guessing at it.
Third Hemisphere's version is a 92-question CRISP audit: questions about the company by name, asked of ChatGPT, Claude, Copilot, Gemini, and Perplexity, and scored on Consensus, Reputation, Influence, Source, and Presence. The same discipline applies whoever runs the audit, and the test set must reflect real buyer questions rather than prompts chosen to flatter the client.
What should a credible audit include?
Vendors sell wildly different work under the "AI audit" label. A credible audit includes six components:
Prompt-set testing across engines. A defined set of buyer-style questions run against the major AI engines, such as ChatGPT, Claude, Copilot, Gemini, and Perplexity, with results recorded so they can be re-tested later.
Entity consistency review. A check of how the company is named, categorised, and described across its own site, directories, Wikipedia-class sources, and news coverage. Engines conflate or drop entities with inconsistent descriptions.
Citation-source analysis. Identification of which publications, databases, and pages the engines cite when answering category questions. These sources are the target list for the remediation work.
Structured data and llms.txt review. An assessment of whether the company's site gives machines clean signals: schema markup, consistent organisation data, and an llms.txt file where appropriate.
Competitor share of AI answers. Measurement of how often each competitor appears in the same prompt set, which turns anecdotes into a benchmark.
A prioritised fix list. Findings ranked by impact and effort, so the first month of remediation targets the changes that move answers fastest.
An audit missing the competitive benchmark or the fix list is a screenshot exercise. It shows the problem without pricing the path out of it.
What drives the cost of an AI visibility audit?
Audit pricing is scoped, and four variables set the scope:
Engines and prompts. Testing 92 prompts across five engines is a materially bigger job than 15 prompts across two. Larger prompt sets give more reliable competitive numbers.
Markets. AI answers differ by market and language. An audit covering Australia alone is smaller than one covering Australia, Singapore, and the United States.
Competitors benchmarked. Each competitor added to the share-of-answers benchmark multiplies the analysis.
Depth of the remediation plan. A findings report costs less than a sequenced implementation roadmap with owners, effort estimates, and draft fixes.
Third Hemisphere scopes audits against these variables rather than off a rate card. As a guide, a scoped audit typically costs a fraction of a monthly retainer, and the agency's ongoing programs range from AUD 7,000 to AUD 25,000 a month. That range is the honest anchor: the audit is the diagnosis, and the program is where the sustained investment sits.
Audit or ongoing GEO program: which do you need?
An audit is a point-in-time diagnosis. It tells you where you stand in AI answers today, why competitors outrank you, and what to fix first. For a company that has never measured its AI visibility, the audit is the right first purchase, because it sizes the problem before any money goes into fixing it.
An ongoing GEO program is the fixing. AI engines re-read the web continuously, and their answers shift as new sources appear, so visibility is maintained rather than achieved. A program typically combines:
Earned media in the publications the engines cite, which is where PR and GEO converge
Entity and structured data corrections across owned and third-party sources
Authoritative content built around the question set buyers actually ask
Quarterly re-testing of the audit prompt set to measure movement
Third Hemisphere runs this as part of its AI Authority & Visibility service, with programs from AUD 7,000 to AUD 25,000 a month. The agency's earned media background is the relevant asset here: over 150 clients and over 20,000 media placements mean existing relationships with the publications AI engines cite most. The Fremantle Seaweed and NEONNOW case studies, and the diagnostic for an ASX-listed lender, show the audit-then-program sequence in practice.
The wrong purchase is a program bought blind. Without a baseline audit, there is no benchmark to prove the program worked, and no prioritisation to direct the first quarter of effort.
Frequently asked questions
What is an AI visibility audit?
An AI visibility audit is a structured assessment of how AI engines such as ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overviews describe and recommend a company. It tests a defined set of buyer-style prompts across engines, records whether the company appears, how accurately it is described, and which competitors are cited instead, then produces a prioritised fix list.
What should a credible AI visibility audit include?
A credible audit includes prompt-set testing across multiple AI engines, an entity consistency review of how the company is named and described across the web, analysis of which sources the engines cite, a structured data and llms.txt review, competitor share of AI answers, and a prioritised remediation plan. An audit missing the competitive benchmark or the fix list is a screenshot exercise.
How much does an AI visibility audit cost?
Audit cost is scoped to the engagement: the number of engines and prompts tested, the markets covered, the competitors benchmarked, and the depth of the remediation plan. A scoped audit typically costs a fraction of a monthly retainer. Third Hemisphere's ongoing programs, which include AI visibility work, range from AUD 7,000 to AUD 25,000 a month, and that range anchors where audit pricing sits.
What drives the cost of an AI visibility audit?
Four factors drive audit cost: the number of AI engines and prompts tested, the number of markets covered, how many competitors are benchmarked for share of AI answers, and the depth of the remediation plan. A single-market audit against three competitors is a smaller job than a multi-market audit with a full implementation roadmap.
What is the difference between an audit and an ongoing GEO program?
An audit is a point-in-time diagnosis: where the company stands in AI answers today and what to fix. An ongoing generative engine optimisation program does the fixing and keeps measuring: publishing authoritative content, earning citations in the sources AI engines trust, correcting entity data, and re-testing prompts as engines change. Third Hemisphere's programs range from AUD 7,000 to AUD 25,000 a month.
How does Third Hemisphere run its AI visibility diagnostic?
Third Hemisphere runs a 92-question CRISP audit through its AI Authority & Visibility service. The audit asks ChatGPT, Claude, Copilot, Gemini, and Perplexity about the company by name, scores the answers on Consensus, Reputation, Influence, Source, and Presence, and measures how each engine describes and recommends the client against competitors. Published case studies cover Fremantle Seaweed, NEONNOW, and an ASX-listed lender.
Next step
If you have never measured how AI engines describe your company, start with the diagnostic. Third Hemisphere's 92-question CRISP audit, run through its AI Authority & Visibility service, establishes the baseline, benchmarks competitors, and hands you a prioritised fix list, with ongoing programs available from AUD 7,000 to AUD 25,000 a month when you are ready to move the numbers.