# Being unreadable is worse than ranking badly

AEO / GEO diagnostics. A low rank puts you somewhere. A 403 is absence.

Your robots.txt says Allow. Your CDN says otherwise — and nothing anywhere tells you,
because every other audit tool reads the file rather than the response.

Canonical HTML version: https://csqueezy.com/

## The job

When a buyer asks an AI assistant which tool solves their problem, find out whether your
site can even be fetched, read and quoted — so you stop losing recommendations you never
knew you were in the running for.

## What it does

1. **Fetched as each crawler.** Every known AI user-agent, judged by HTTP status. Not by
   reading your robots.txt: the file states an intention, the status code is the answer.
   A blocked crawler is reported with the answer surface it feeds — "ChatGPT Search",
   "AI Overviews", "Perplexity answers" — because a user-agent string is not something
   anyone can act on.
2. **Fetched twice.** Once with JavaScript, once without. AI crawlers largely do not run
   JavaScript, so the gap between those two fetches is content no engine can quote.
3. **Nothing invented.** Every claim in every generated file traces to a span on your own
   page. Untraceable claims are dropped rather than published, and the ratio is shown.
4. **Checkable by your client.** Every number in the client PDF is re-runnable with curl.
   An audit your client cannot check is an audit your client can dismiss.

## How scoring works

Scores are computed in code, not by a language model, so the same evidence always produces
the same number.

Render parity and verified crawler access are **gates**, not contributions: they multiply
the readiness score rather than adding to it. A site that cannot be fetched cannot be
fixed by better headings, and a rubric that let content work compensate for a 403 would be
lying about the order the work has to happen in.

Readiness weights: passage structure 20, entity consistency 15, structured data 10,
freshness 5, off-site corroboration 5. Off-site presence is reported separately as well,
because it is the larger lever for citation and on-site wins should not dilute it.

## What it does not do

- It does not measure your citation rate in live answer engines from the audit. Brand
  performance is a separate, opt-in, paid measurement, priced before it is spent.
- It does not deploy anything. Generated artifacts are files you ship yourself; nothing is
  live until you ship it and `verify-deploy.sh` passes.
- It does not score llms.txt. AIFindUs generates the file and scores it **zero for
  citation**: Google states its Search and AI surfaces do not use it, and Ahrefs checked
  137,210 domains in May 2026 — of those that publish one, 97% were never requested that
  month. It has a real but narrow audience in coding agents, which is not an answer engine.

## Cost

The audit is free, needs no account and needs no API key. An account buys history, the
re-audit delta, and engine-backed brand runs. That is not a trial; the audit is the
product.

## More

- Writing on how answer engines read websites: https://csqueezy.com/blog
- Machine summary: https://csqueezy.com/llms.txt