What Should an AEO Audit Actually Deliver?

"AEO audit" now means almost anything. We have reviewed deliverables that were a re-badged technical crawl, a list of prompts someone typed into ChatGPT one afternoon, and a slide deck of assertions with no evidence attached at all. Buyers in Vancouver are being asked to pay real money without a shared definition of what they are buying.

So here is ours. Eight outputs, each connected to the next. If a proposal is missing several of them, you are buying something narrower than the name suggests.

1. The buyer questions you should be found for

Answer engines respond to questions, not keyword strings. The audit starts by writing down the actual questions your buyers ask at each stage — problem-aware, solution-aware, vendor-comparison — and fixing that list as the prompt set everything else is measured against. If the audit skips this, the rest of it has nothing to be measured against.

2. Technical retrieval

Can the systems reach and read your content at all?

• AI crawler access in robots.txt, checked per user agent rather than assumed.
• Server-rendered content, so the answer is not locked behind JS.
• Status codes, redirect chains and canonical consistency.
• Page speed and response times for crawler timeouts.
• llms.txt presence and accuracy where you use one.

This layer explains most "we rank well but never get cited" cases we diagnose.

3. Entity facts

Answer engines need to know who you are before they will vouch for you. The audit checks that your business name, address, founding details, services and leadership are stated consistently across your site, your Business Profile, and the third-party sources that describe you — and that your Organization schema's sameAs links point to profiles that actually exist and agree with each other. Contradictions here are quietly expensive.

4. Evidence quality

This is the output most audits skip. Answer engines preferentially quote content that states something checkable: a figure with a source, a date, a named method, a specific process. Pages of adjectives get skimmed and dropped. The audit should mark, page by page, whether the content offers anything an engine could safely repeat with your name attached.

5. The source landscape

For each question in your prompt set, who is currently being cited? Directories, competitors, publications, forums, or nobody. That map tells you whether you are competing against a well-defended incumbent answer or an open field — and those need completely different plans.

6. Structured data

Validated, not merely present. Organization, LocalBusiness, Service, Article and Breadcrumb markup should be consistent, error-free, and matched by the visible content on the page. FAQ markup belongs only where real, visible questions and answers exist on the page.

7. A measured visibility baseline

Without a baseline, no later claim of improvement can be checked. A baseline means: the fixed prompt set from output one, run against named engines, on recorded dates, with the answers saved. Alongside it, Google's dedicated generative AI reporting — which Google's June announcement now states rolled out worldwide as of August 31, 2026 — gives an impressions-based view of Google's own surfaces.

Two cautions worth putting in writing. Those reports are impressions only. And impressions are a measure of your site's visibility, never of how many people searched a topic.

8. An accountable roadmap

The deliverable is a prioritised list where each item has an owner, an effort estimate, the expected effect, and a date the baseline gets re-measured. Anything that cannot carry those four fields is a recommendation, not a plan.

Claims to question

• "We guarantee ChatGPT citations." Nobody controls that.
• "Our AI visibility score is 62." Ask what the scale is, what was measured, on which engines, and on what date.
• "We track AI traffic." Ask precisely which surfaces and how the sessions are identified.
• Keyword volume, difficulty or CPC presented as AEO data. Those are paid-search and SEO tool metrics; they do not describe answer engine behaviour.

What our own visibility data shows

For transparency, here is where we sit. Between August 4 and 31, 2026, Search Console recorded aeo services at 179 impressions and 16.27 average position, aeo audit services at 26 impressions and 11.19, and aeo audit pricing for small sites at 24 impressions and 6.38. These are ThinkProfits visibility observations from our own property. They are not search volume, and they should not be read as market demand for anyone else.

If you want the work as well as the diagnosis, our AEO services cover the retrieval and entity layers, our SEO audit services handle the technical foundation, and our SEO services and content writing services carry the evidence-quality work.