AEO Audit: What It Is, How To Run One, And How The Models Behind The Answers Score

An AEO audit is a structured check of whether AI answer engines like ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews name your brand, cite your pages and describe you accurately.

AEO stands for answer engine optimisation, and the audit is the measurement step you run before you change anything.

If you'd like an expert to walk through your AI visibility with you, book a free SEO strategy session and get a custom visibility plan.

The problem it solves is simple, because buyers now ask AI tools who to hire or what to buy, and a brand that isn't named never gets the enquiry.

In this post I'll explain what goes into an AEO audit, how to run one step by step and how to turn the results into fixes.

Near the end, I'll show you how the AI models behind those answers score on this leaderboard, and what that does and doesn't tell you.

What is an AEO audit?

An AEO audit answers one question, which is whether the AI tools your buyers use mention you when it matters.

A classic SEO audit measures rankings, links and technical health, while an AEO audit measures mentions, citations and accuracy.

An AEO audit isn't a piece of software, and it doesn't come with its own AI model.

It's a method you run with a spreadsheet, the AI tools themselves and a few free checks.

How to run an AEO audit, step by step

These are the steps I'd follow on any brand, in this order.

1. Write the prompt panel

Write 20 to 50 questions a real buyer would type into an AI tool before choosing a company like yours.

Mix best-of, comparison, problem, brand, alternatives and local questions so the whole buying journey is covered.

2. Run the prompts

Ask every prompt in ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and Google AI Mode, using a fresh chat each time.

Run each one at least twice, because AI answers change from one chat to the next.

Google says AI Overviews and AI Mode can use different models and techniques, so log them as two separate tools.

3. Log the answers

For each answer, write down whether you were named, which competitors were named, which sources were cited and whether your facts were right.

Add the date to every row, because the dated log is what lets you prove change later.

4. Check crawler access

Open your robots.txt file and confirm the crawlers you want aren't blocked.

I read OpenAI's crawler page, Perplexity's crawler page and Google's AI features guide on 17 October 2026.

Google says a page must be indexed and eligible to show a snippet to appear as a supporting link in AI Overviews or AI Mode, with no additional technical requirements.

5. Check entity facts

Write a one-page fact file and compare it with your About page, social profiles, Google Business Profile and directory listings.

Google's advice includes keeping Merchant Center and Business Profile information up to date.

6. Check third-party coverage

List the domains that appear most often in the cited sources and check whether your brand is on each one.

My strategy session page says AI platforms don't recommend brands simply because those brands publish more on their own websites.

It says they look for trusted publications, expert articles, independent reviews, comparisons and credible third-party references.

7. Grade your key pages

Read the first sentence of your ten most important pages and check that each one answers the main question directly.

My own rule is to keep that first answer under fifty words and add a short FAQ section to every important page.

8. Check schema basics

Make sure your structured data is valid and matches the visible text on the page.

Google says you don't need new machine-readable files, AI text files or special schema.org markup to appear in AI Overviews or AI Mode.

📞 Want my team to go through these steps with you? Book a free SEO strategy session and you'll leave with a clear AI visibility roadmap and a competitor gap analysis that shows why competing brands appear more often.

How to score an AEO audit and choose the fixes

Score each area from 0 to 2, where 0 means broken, 1 means partly done and 2 means solid.

I rank the fixes by impact first and effort second.

Crawler and indexing blocks come first, because nothing else works until the AI can read your pages.

Wrong facts come next, because they are cheap to fix and they hurt every answer.

Answer-first rewrites and FAQ sections come third, and third-party coverage is the slow, high-value work that runs alongside everything.

AEO audit vs SEO audit: what's different?

The two audits overlap, because AI answers still lean on pages that search engines can crawl and index.

The difference is what each one treats as success.

I run both, and the AEO audit usually tells me which classic SEO fixes matter most for the AI side.

Who should run an AEO audit?

Any business that wins customers through research and recommendations should run one.

Local service businesses should run one because AI tools now answer "best near me" style questions.

Software companies should run one because comparison and alternatives questions decide which tools buyers try.

Consultants and coaches should run one because bylines, interviews and reviews are the kind of third-party references my strategy session page says AI platforms look for.

Agencies should run one for every new client, because the dated log becomes the baseline every later report is judged against.

How to measure progress after an AEO audit

Re-run the same prompt panel every month and compare it with your dated baseline.

Google counts AI Overviews and AI Mode traffic inside the "Web" search type of the Search Console Performance report.

Bing Webmaster Tools added an AI Performance report in public preview on 10 February 2026, and Microsoft says it shows how often your pages are cited in Copilot and AI summaries in Bing.

The number that matters most is still enquiries and sales.

What an AEO audit can't do

Nobody can guarantee what a model says, and that includes me and my agency.

An audit shows you the footprint the models read and how they describe you today, and it can't control tomorrow's answer.

Be wary of anyone who guarantees a mention, or who sells an AI text file or special schema as the secret fix.

📞 Want a straight answer about your own brand? Grab a free SEO strategy session with my team and we'll map the bylines, reviews, comparisons and research that fit your market.

Which AI models sit behind the answers you're auditing?

Now for the GoldieBench part, and I need to be clear about what it measures.

GoldieBench scores models on one-shot builds such as landing pages, apps, games and visuals, on a 0 to 10 scale.

It doesn't measure which brands an AI recommends, and it doesn't measure SEO or AEO results.

An AEO audit doesn't ship its own model, but every answer you log was written by a model from one of the big makers.

This is how those makers' models score on their live pages today.

Claude Opus 5 is the highest of these at 8.27, with GPT-5.6 Sol close behind at 8.16.

Gemini 3.6 Flash averages 7.08, and it's the only Gemini model on the board, so it's the closest view I have of the Gemini family.

Perplexity and Microsoft Copilot have no model on GoldieBench, so I can't give you a score for either.

The board also can't tell you which exact model a consumer app is running at any moment, because the makers change that themselves.

So please don't read these scores as a ranking of which AI is easiest to get mentioned in.

What the model scores mean for your AEO audit

The scores matter in a narrower way than people expect.

They don't change how you run the audit, because you test the tools your buyers use whatever their scores are.

They matter more if you use a model to help run the audit, for example to sort hundreds of logged answers or to draft answer-first rewrites.

A model that follows a detailed brief well first time makes fewer mistakes for a person to catch.

Even so, the system around the model matters more than the model, because a clear fact file and a human review beat a top model with no checks.

I use AI agents to run the repetitive parts of my own audits, and a person still approves every change that goes live.

Related reading on GoldieBench

My King of AEO post looks at answer engine optimisation from the model side.

My AI SEO optimization agency post covers what to ask an agency that runs this work for you.

In the same batch, I also wrote about the free AEO audit and the free GEO audit.

If you'd like my team to look at your brand, you can book a free SEO strategy session and leave with a plan for traffic, mentions and leads.

Also On Our Network

• Agent Operatorsthe 10-step operator method for an AEO audit
• AI Income Deskthe five lead leaks an AEO audit finds for a business
• AI Tool Verdicttwelve AEO audit checks ranked by impact
• agentos.guidean AEO audit quick start for your Agent OS
• aiprofitboardroom.commy guide to getting found in Perplexity

Write your prompts, log the dated answers, fix the blocks first and judge the system around the model, and you'll have run a proper AEO audit.