AEO & GEO Readiness Audit

How is the readiness score calculated?

Being quoted by an answer engine is three separate problems wearing one coat, and the audit scores them separately because sites usually fail at only one of them. The first is retrieval: whether a crawler that does not run JavaScript can fetch your HTML, is permitted to do so by your robots policy, and can tell which URL is the real one. The second is extraction: whether, having parsed the page, there is a short self-contained passage worth lifting rather than a wall of qualified marketing prose. The third is attribution: whether the passage comes with a named human, a publisher, a date and matching structured data, so an engine has something concrete to credit.

Each of the twelve checks carries a weight reflecting how often it is the actual blocker in real audits rather than how often it appears in checklists. Server rendering and the presence of a direct opening answer carry the heaviest weights, because a page failing either of those cannot be rescued by anything else on the list. A missing llms.txt or thin source citation carries a light weight — worth fixing, never the reason you are absent from an answer.

Your answers are converted to a percentage of the weight available, both overall and within each pillar, which is why the pillar bars often disagree sharply with each other. A well-built marketing site commonly scores in the eighties for retrieval and the thirties for extraction: perfectly crawlable, with nothing on it shaped like an answer. That gap is the finding, and it is the reason the tool reports a weakest pillar rather than a single number alone.

The remediation list is not the checklist order. Every partial or failing answer contributes a residual weight equal to the score still unclaimed, and the list is sorted by that residual. Fixing the top item is always the largest single available gain, which matters when the realistic budget is one afternoon rather than one quarter.

How do you use the AEO & GEO Readiness Audit?

1. 1.Paste the URL of the page you actually want cited — a money page or a reference guide, not your homepage — and run the scan. The tool requests the page once, unauthenticated, and also reads /robots.txt and /llms.txt on the same host.
2. 2.Read the per-check evidence lines. Each one reports what was actually found in the HTML — word count, canonical value, schema types, heading counts — so you can confirm the verdict yourself with view-source.
3. 3.If the page is behind a login, blocks bots, or cannot be fetched, switch to the self-assessment checklist tab and answer the twelve questions manually. A page a scanner cannot reach is a page an AI crawler cannot reach either.
4. 4.Read the pillar bars before the headline number, then open the “How to fix what scored low” panel and work it from the top. Each entry explains why the check matters, gives numbered steps, copy-paste markup and a way to verify the fix. Re-scan after each change.

What can this tool not tell you?

• It audits one URL at a time, not a whole site. It also cannot see pages that require a login, and sites that block automated requests will fail the fetch entirely.
• It cannot tell you whether you are currently being cited. Only prompting the engines directly, or reading referral logs for ChatGPT and Perplexity user agents, answers that.
• It scores structure, not authority. A technically flawless page on a domain nobody references will still lose to a weaker page from an established source.
• The weights reflect how these systems behave now. Retrieval behaviour changes with each model generation, and any tool claiming permanent rules about it is guessing.

Why most pages fail on extraction, not on crawling?

The uncomfortable pattern across audits is that the technical layer is usually fine. Modern frameworks server-render by default, canonicals are generated automatically, sitemaps are produced by a plugin nobody has touched in two years and still works. Teams arrive expecting a plumbing problem and score in the eighties on retrieval. Then the extraction pillar comes back at thirty and the conversation changes, because the fix is not a ticket for an engineer — it is rewriting the first two hundred words of every important page, which nobody owns.

The reason is that a decade of conversion-focused copywriting taught the opposite instinct. Good landing page copy opens with tension, withholds the specifics, and moves the reader down the page toward an action. A generative engine reads that opening and finds nothing it can repeat as an answer, so it keeps looking, finds a competitor who states the fact plainly in sentence one, and quotes them instead. The page that converts best from a click can be the page least likely to earn the click in the first place.

The resolution is not to abandon persuasive writing but to front-load one honest, self-contained paragraph before it. Readers who already know they are in the right place skim past it; readers who do not get their answer immediately and trust you more for it; and the engine gets a clean block with a beginning and an end. In practice this single change moves more pages from paraphrased-without-credit to quoted-by-name than every schema improvement combined.

Attribution is the quieter failure. Plenty of otherwise strong pages carry no byline at all, or a company name where a person should be, and engines weigh resolvable human authorship heavily when deciding whose version of a contested claim to repeat. Adding a real author with a bio page describing actual expertise is a half-day of work that no competitor can copy quickly, which makes it one of the few durable advantages left in this discipline.

What do worked examples look like?

What do people ask most about this tool?

Which related tools should you try next?