From ranking to being the answer
Search engine optimization is about position: where your page sits in a list of links. Answer engine optimization is about inclusion: whether an assistant that writes a single answer names your company in it and points to your pages as sources. The buyer who asks “which clinic in Tallinn does implants for English speakers?” and receives three names never sees a list of ten links. If you are not one of the three, the rest of your marketing did not get a turn.
The two disciplines share foundations. The engines retrieve from the open web; Google states that AI features in Search rely on the same crawling and indexing as ordinary results, and OpenAI, Perplexity and Anthropic each operate their own crawlers with documented user agents. A page that cannot be crawled or rendered is invisible to both kinds of engine. Beyond that, the practices diverge: link building and keyword placement matter less; being the clearest, most consistent source for a specific question matters more.
What an AEO audit checks
1. What the engines say today
The audit begins with evidence, not with your website. A set of buyer questions (awareness, comparison, commercial, local, brand) is run on each engine, more than once, from your market, and every answer is kept verbatim with its sources. From those answers the audit records whether you were named, whether your site was cited, which competitors were named, and what the engines claimed about you.
2. Accuracy
Engines state wrong things with confidence: a wrong price, a wrong city, a service you no longer offer. An AEO audit compares each claim about you against the facts you supply and grades the mismatches by consequence. A wrong opening hour is a nuisance; a wrong regulated claim about a medical service is a liability. These corrections come first in any fix list because they are both the most damaging and usually the cheapest to fix.
3. Crawl access
Each engine company publishes which user agents it uses and what for. OpenAI documents separate agents for model training, for search indexing and for user-initiated browsing. Perplexity documents its own crawler. Anthropic explains how site owners can allow or block its crawlers. Google’s crawler overview separates Googlebot, which feeds Search and AI Overviews, from the Google-Extended token that controls whether content may be used for Google’s other AI products. A site that blocks every crawler with “AI” in the name also blocks the agents that feed live answers, which is a common and usually unintended way to disappear. The audit reads your robots.txt and reports, per crawler, what you allow.
4. Citability of your key pages
An engine can only cite a page it can read and quote. The audit fetches your homepage and your most important pages as a crawler would and checks: is the main text present without JavaScript, does the page open by answering the question it exists for, are facts stated plainly rather than implied, is there structured data describing something real, is there a visible date. Pages that are all hero image and slogan score poorly here, however well they convert human visitors.
5. Competitor share
Who is named when you are not, and on which question types. This is not market share; it is a count within the sample. It tells you which competitors the engines treat as the default answer for your category, and the sources they are cited from, which is often the fastest route to being included yourself.
What is well supported, and what is not
Honest AEO advice distinguishes between things that are documented or directly observable and things that are plausible but unproven.
| Lever | Status |
|---|---|
| Allowing the search and browsing crawlers in robots.txt | Documented by the engine companies; directly observable in which pages get cited. |
| Server-rendered text on key pages | Observable: crawlers that fetch raw HTML cannot cite text that only appears after scripts run. |
| Consistent facts across your site and third-party listings | Observable in the accuracy flags; the engines repeat what the most-cited sources say. |
| Answer-first page structure, clear headings, FAQ sections | Reasonable and consistent with how retrieval works; the size of the effect is not established. |
| Structured data (JSON-LD) | Useful where it describes real things; no engine has published that it changes inclusion in AI answers. |
| llms.txt | A published proposal; no major engine has documented using it. Cheap to add, low expected impact. |
An audit should label each recommendation accordingly. When ours lists a fix, the impact is stated as a hypothesis, not a promise.
How to run one
If you want to try it yourself before buying anything: write ten questions a buyer might ask, in their words, not yours. Ask each on ChatGPT, Perplexity, Google and Claude in fresh, logged-out sessions, twice. Paste every answer into a document with the date. Mark your name and your competitors’ names. Read your robots.txt. Disable JavaScript in your browser and open your three most important pages. You will learn most of what an audit tells you, slowly and without the comparability.
The audit version of the same exercise is what we sell: more questions, four engines, repeat runs from your market, the technical checks done by code, the accuracy comparison against your facts, and a person reading it before it ships. The methodology describes each step, and the sample report shows the result.