How to Run an AI Visibility Audit (Free Template)

Most brands have no idea whether AI answer engines recommend them, ignore them, or describe them wrong. An AI visibility audit fixes that: it is a structured test that runs a fixed set of buyer prompts across ChatGPT, Gemini, Perplexity, and Google AI Overviews, then scores how often your brand appears, how much answer space it owns versus rivals, and whether the facts are right. This guide gives you the exact six-step method — plus a free scorecard template — so you can run your first AI visibility audit this week and repeat it every month.

What is an AI visibility audit?

An AI visibility audit is a repeatable measurement of how a brand shows up inside AI-generated answers. You define a set of real buyer questions, run them across the major AI engines, and record three things: whether you were mentioned, how much of the answer you owned relative to competitors (your share of voice), and whether the AI described you accurately. The output is a scorecard you can benchmark and re-run.

Think of it as the AI-era equivalent of a keyword-ranking report — except the "SERP" is a paragraph of prose that recommends two or three vendors and forgets the rest.

Why AI visibility is worth auditing now

The reason is simple: buyers have moved, and most brands have not measured the move. The data on how far it has gone is striking:

• 73% of B2B buyers now use AI tools like ChatGPT and Perplexity during purchase research, and Forrester puts the figure at 94% for the most recent purchase cycle.
• 69% of buyers chose a different vendor than they originally planned based on AI chatbot guidance, and roughly a third bought from a vendor they had never heard of before the AI surfaced it.
• Yet only 14% of brands have any AI visibility strategy at all.

That gap is the opportunity. When an answer engine names three vendors for "best [your category] tool," being one of the three is worth more than a page-one Google ranking — and being absent is invisible in a way ranking #11 never was. An audit tells you which side of that line you are on.

You cannot optimize what you never measured. An AI visibility audit turns "are we in the answer?" from a guess into a number.

How to run an AI visibility audit in six steps

Step 1: Build a representative prompt set

Your audit is only as good as its questions. Aim for 30–50 prompts — a minimum of 15 produces directional data, but 30–50 gives you something stable. Spread them across three intent types:

• Category prompts — "best [category] software," "top tools for [job to be done]." These test whether you enter the consideration set at all.
• Comparison prompts — "[you] vs [competitor]," "alternatives to [competitor]." These test how you fare head-to-head.
• Brand prompts — "what is [your brand]," "is [your brand] any good." These test accuracy and sentiment.

A quick way to size the set: [number of topic clusters] × [12–15 questions per cluster]. Write prompts the way a real buyer types them, not the way a marketer writes headlines. If you need help finding the exact phrasing buyers use, our guide to AI keyword research walks through mining the prompts your buyers actually ask.

Step 2: Fix your competitor set (the denominator)

Share of voice is a fraction, and the denominator decides everything. Count only two rivals and your share looks inflated; count every tangential brand and it collapses. Choose a fixed list of 3–7 real competitors that reflects your actual market, then hold it constant across every audit. Changing the competitor set between runs makes your trend line meaningless.

Step 3: Run every prompt across every engine — and repeat

Run all prompts through each engine you care about: ChatGPT, Gemini, Perplexity, and Google AI Overviews at a minimum. Two rules matter here:

1. Use fresh sessions. Memory and personalization contaminate results. Log out or use a clean context each time.
2. Repeat each prompt 3–5 times. AI answers are non-deterministic — the same question yields different vendors on different runs. A single snapshot misrepresents your true standing; averaging across repeats stabilizes it.

Record the raw answer text for every run. That transcript is your evidence base for the next step.

Step 4: Score three metrics

For each engine, roll your runs up into three numbers:

Presence rate tells you if you are in the room. Share of voice tells you how loud your voice is once you are. Accuracy tells you whether the room is hearing the truth. A brand can score 80% presence and still lose deals if the AI keeps citing a discontinued price or a two-year-old feature gap.

Step 5: Diagnose the gaps

Now read the transcripts for why. Patterns cluster into three fixable buckets:

• Absence — you are never named. Usually a citation and authority problem: the sources the AI trusts do not mention you. Note that 84% of AI citations come from earned media, not your own site, so this is often a PR and review-presence gap.
• Under-indexing — you appear, but rarely and late. Usually a content-coverage problem: you lack the structured, extractable pages AI engines quote from.
• Inaccuracy — you appear, but the AI is wrong. Usually a freshness problem: outdated third-party pages or thin owned content the model is defaulting to.

Each bucket points to a different fix. Our CITE framework for getting cited by ChatGPT covers the absence and under-indexing fixes in depth, and the marketer's guide to generative engine optimization covers the structural work that makes your content extractable.

Step 6: Turn the audit into a monitor

A one-time audit is a photograph; you need a film. AI answers drift weekly as models retrain and sources change, so re-run the same prompt set, same engines, same competitor list on a monthly cadence (weekly if your category is fast-moving). Chart presence rate and share of voice over time. The trend line — not the snapshot — is what tells you whether your GEO work is compounding. Pair it with downstream traffic data using our guide to tracking AI search traffic in GA4 to connect visibility to conversions.

Tips and reminders

• Standardize ruthlessly. Same prompts, same engines, same competitor set, same cadence — every run. Consistency is what makes the numbers comparable.
• Weight by position. A brand named first in an answer earns more trust than one buried in a caveat. If you can, score first-mention separately.
• Watch sentiment, not just presence. Being mentioned as "the expensive option" is a visibility problem disguised as a win.
• Automate the boring parts. Running 40 prompts × 4 engines × 5 repeats by hand is 800 queries a month. This is precisely the kind of continuous, structured measurement that autonomous marketing agents are built to run on a schedule — so the audit becomes a living dashboard instead of a quarterly scramble.

The AI visibility scorecard (copy this)

Rebuild this as a spreadsheet, one row per engine, and fill it each cycle:

Ten fields, five minutes to read, and a number you can defend in a board meeting. That is the whole point: an AI visibility audit replaces "I think we show up in ChatGPT" with a measurement you can improve on purpose.

Frequently asked questions

What is an AI visibility audit?

An AI visibility audit is a structured test that runs a fixed set of buyer prompts across AI answer engines like ChatGPT, Gemini, and Perplexity, then scores how often your brand is mentioned, how much answer space you own versus competitors, and whether the AI describes you accurately. It produces a repeatable scorecard you can benchmark and track over time.

How many prompts do I need for a reliable audit?

A minimum of 15 prompts gives directional signal, but 30–50 prompts produces stable, defensible data. Spread them across category, comparison, and brand-specific intents, and repeat each prompt three to five times because AI answers are non-deterministic and shift between runs.

How is AI share of voice calculated?

AI share of voice is your brand mentions divided by all brand mentions — yours plus a fixed competitor set — across your prompt set, expressed as a percentage. The competitor list is the denominator, so choose 3–7 real rivals and hold that list constant across every audit to keep your trend line valid.

How often should I run an AI visibility audit?

Monthly is the practical default because AI answers drift as models retrain and sources update; weekly suits fast-moving categories. A single snapshot misleads — it is the trend in presence rate and share of voice over successive audits that tells you whether your optimization is actually working.

Can AI visibility audits be automated?

Yes. The core work — running dozens of prompts across several engines, repeating each for stability, and scoring the transcripts — is repetitive and rule-based, which makes it ideal for automation. Autonomous marketing agents can run the same audit on a schedule and surface the trend, turning a periodic project into a continuous monitor.

Sources

• 73% of B2B Buyers Use AI Tools in Purchase Research — PR Newswire
• Half of B2B Software Buyers Now Start Research With AI Chatbots — G2 / Demand Gen Report
• AI Search Visibility Statistics 2026: 8,400-Prompt Brand Tracker — Visionary Marketing
• Measuring Share of Voice Inside AI Answer Engines — Single Grain
• How to Measure AI Share of Voice: Methods, Tools and Benchmarks — Cognizo
• How to measure AI share of voice — Scrunch