# How Answerworthy works: measure, score, fix

> Answerworthy runs your buyers’ real questions through AI engines, scans your site the way AI crawlers read it, and turns both into a ranked fix list.

Answerworthy works the way an honest analyst would: it asks the engines your buyers ask, reads
your site the way AI crawlers read it, stores the raw evidence for every claim, and only then
tells you what to change.

## The four layers

| Layer | The question it answers | How Answerworthy does it |
| --- | --- | --- |
| Citation tracking | Are you in the answer? | Panels of real buyer questions run on each engine’s native API; citations and mentions are recorded per engine, per question, over time. |
| Crawler analytics | Is AI even visiting? | Verified AI-bot traffic on your domain, harvested from your CDN logs: which bots come, what they read, what they get blocked from. |
| On-site readiness | Can AI read what it finds? | 91 deterministic checks covering crawl access, structured data, entity signals, answer liftability and site foundation, fetched as GPTBot, ClaudeBot and PerplexityBot. |
| Actions & proof | What do you fix first? | Every failed check and every lost answer becomes a ranked recommendation. When a fix ships, the next scan and the next panel run prove it. |

## What makes the measurement trustworthy

- **Native engine APIs**: panels run against each AI engine directly, never through a reseller proxy.
- **Raw responses stored**: every engine answer is archived verbatim; every dashboard number traces to one.
- **Observed vs. inferred**: unverifiable findings are labelled as such at the database level, never blended.

## Start here

- Run the free readiness scan: https://app.aeo-test.ai
- Learn the fundamentals: https://answerworthy.ai/learn/what-is-aeo