What you get in a free AEO Armory audit, section by section.

A free AEO Armory audit asks ChatGPT and Google AI Mode 8 questions your customers would ask, reads up to 15 pages of your site, and turns both into one report with a prioritized fix list. It takes a few minutes to run and costs nothing: a free account gets 3 reports a day, no card. AEO Armory is our tool, so this walkthrough covers what each section shows, how to read it, and what the audit does not do.

Before the report: what you enter

You sign up at AEO Armory with your email and a one-time code (no password), then add your business name, website, industry, role and company size. Each new report asks for five things:

• Website and business name.
• What you sell, in your customers' words: "roof repair", not "residential exterior solutions".
• Where, or blank if you sell nationally or B2B. The questions change to match.
• Up to two competitors, optional. These are the businesses the report compares you against.

The scan then reads your site, writes the questions, sends them to both AI engines and collects the answers. It usually finishes in a few minutes. If an engine is slow, the scan stops waiting after about 8 minutes and scores the answers it has.

1. The AI visibility score

The headline number, out of 100, sums up how often and how high the AI answers named you across the questions. Being named first in an answer earns full marks, being named lower earns less, and having your site cited without being named earns a little. Next to it you get three more numbers: how many answers named you (for example 4 of 16), your share of voice against the competitors you entered, and a separate score for ChatGPT and for Google AI Mode. The full method is in how the AI visibility score is calculated.

2. The question table

This is the most useful part of the report. Each row is one of the 8 questions, with a column per engine showing whether you were named and at what position, which of your competitors were named, and whether your site was cited as a source. A question that mentions your business by name is shown but not scored, since being named there proves nothing.

Read it row by row. Questions where you are named lower down are usually the quickest wins; questions where only a competitor is named tell you where to look next.

3. You vs. competitors

A short table lists you and each competitor you entered, with how many answers named each and how often each site was cited. If a competitor is named in 9 answers and you are named in 3, that gap is the story of the report. From here you can open the Competitor Gap Analysis tool to compare your site with theirs page by page.

4. The sources AI cites

The answers lean on third-party pages: directories, review sites, publications, Reddit and Quora threads. The audit collects the domains the answers cited, leaves out your site, your competitors' sites and the sites of businesses the answer itself recommended, and turns the most-cited ones into off-site tasks. Each tells you which engine cited the page, for which of your questions, and whether to get listed, earn reviews or join the conversation. If no Reddit or Quora thread was cited, you get a task to build a presence there anyway.

5. The 9 readiness checks

Site readiness is a separate score: how easy your site makes it for AI to read, trust and quote you. It averages nine checks, each scored 0 to 100 with a one-line finding:

• AI crawler access (blocked AI bots, sitemap, llms.txt)
• Readable without JavaScript
• Answer-ready formatting
• Questions answered (FAQ)
• Facts and sources
• Structured data
• Trust and sharing signals
• Response speed
• Plain-English readability

A page-by-page table underneath shows words, schema types, question headings, facts, lists and response time for every page the scan read. Our AI citation readiness checklist explains each check.

6. The fix list

Every finding becomes a task, tagged technical, content or off-site, with a difficulty, an estimate in minutes and a priority score that weighs impact, the size of the gap and confidence against effort. Site-wide fixes get a small bonus, because one change helps every page. Each task shows:

• Why it matters and the evidence from your site or the AI answers
• The fix, with steps for common setups such as WordPress with Rank Math or Yoast
• Paste-ready text where it applies: robots.txt lines, schema, an llms.txt, a list of questions to answer, with a copy button

Tasks that are best practice rather than proven, such as adding an llms.txt, are labeled that way and ranked low.

7. The prompt plan

The report lists 15 to 30 questions you should be the answer to, each with an intent (ready to buy, cost, comparing, how-to, local, trust), what to create (a new article, a page update or an FAQ answer) and whether AI names you for it today. It totals the work in hours and months. Turning the prompt plan into content covers how to use it.

8. The PDF and your dashboard

Every report is saved in your dashboard, and a print button produces a clean PDF to send to your developer, marketer or agency. From the dashboard you can run the same report again with the same inputs.

9. The fix checker

When you finish a task, press "I've done this: check it". For fixes it can test, the checker fetches your live site and confirms the change: robots.txt rules, a sitemap, an llms.txt, HTTPS, Organization or LocalBusiness schema on the homepage, and page-level items such as an answer near the top, question headings, author and date, title and description, a single H1, noindex, alt text and text readable without JavaScript. Off-site and content tasks it cannot test are simply marked done; their effect shows up in your next report.

The honest limits

• Two engines. ChatGPT and Google AI Mode, not Perplexity, Gemini or Copilot.
• Eight questions. A sample of what buyers ask, not every question in your market.
• Answers vary run to run. The same question can name you today and not tomorrow, so one report is a snapshot. Compare several.
• US, English. Answers are collected as a US user in English.
• No tracking history yet. There are no daily tracking or trend charts; you compare reports yourself.

Within those limits, it tells you where you stand and what to do first, free.

Frequently asked questions

How long does an AEO Armory audit take?

Usually a few minutes. The scan stops waiting for slow AI answers after about 8 minutes and scores what it has.

Is the AEO Armory audit really free?

Yes. A free account, made with your email and a one-time code, gets 3 full reports a day with no card.

Which AI engines does AEO Armory test?

ChatGPT and Google AI Mode, with 8 customer questions each, collected as a US user in English.

Does AEO Armory change my website?

No. It only reads your public pages. You or your developer make the changes, and the fix checker confirms them.