> ## Documentation Index
> Fetch the complete documentation index at: https://docs.anymorph.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Site Audit report
> What the AI-readability diagnosis covers, what you get back, and what we deliberately leave out.
This page explains what an Anymorph Site Audit actually delivers: which pages we test, how we test them, what a finding looks like, and how you confirm a fix. For the screen-by-screen guide, see [Site Audit](/dashboard/site-audit).
## Why this matters
When ChatGPT, Perplexity or Google's AI features answer a question, they **open your pages at that moment and read them**. They do not read the way a person does.
* Text that appears only after JavaScript runs in the browser is invisible to them.
* Text inside images (ingredient tables, spec sheets, price tables, shipping notices in your detail images) is invisible to them.
* If a firewall, security service or robots.txt rule blocks AI search bots, they cannot open the page at all.
* If the top of the page is all menus and navigation, they decide the page has nothing to say and move on.
A page that looks fine to a person can be empty to an AI. A page an AI cannot read never gets cited. In Anymorph's own data, when an AI cites a brand's website, more than 90% of the time it is a page that already existed. Before you create new content, the pages you already have need to be readable.
## What we deliver
### 1. We read your site the way AI does
We pick your main pages and **fetch them exactly as the ChatGPT, Perplexity and Google search bots do**, then put what the bot received next to what a person sees in a browser. This is a real fetch, not an estimate.
Every page gets one verdict:
| Verdict | Meaning |
| ---------------------- | ----------------------------------------------------------------------- |
| **Readable** | AI can read the key facts as they are |
| **Partially readable** | Some key facts (price, specs, capabilities) are hidden from AI |
| **Unreadable** | AI cannot open the page, or opens it and finds it empty |
| **Unreachable** | The search bots were refused entry |
| **Not measured** | We could not measure it. This is never reported as a fault in your site |
### 2. We tell you why
There is no single score. Each finding names the exact stage where a page fails.
* **Fetch — access** — a firewall, security service or robots.txt rule blocks AI search bots. Or the page is so large or slow that the bot gives up.
* **Index — indexability** — a noindex or nosnippet setting excludes the page from search.
* **Extract — body text** — key facts exist only after JavaScript runs, only inside detail images, or only in machine-only data blocks the bots discard.
* **Snippet** — the first lines after the title are menus, so the AI never opens the page.
We also show which facts are missing, in plain words. For example: "On this product page the price, size and material only exist inside the detail images, so AI cannot read them."
### 3. We give you a fix you can paste
Not "improve your content". Every finding comes with all of the following filled in:
* **Where** — which page and which spot on it, with instructions that match your platform (Cafe24, Godo, Shopify, WordPress, custom builds)
* **Current** and **Proposed** — the exact replacement, ready to paste: an HTML block, one robots.txt line, a setting value, or a rewritten heading and opening paragraph
* Who does it — **Developer**, **Content**, **SEO** or **Infra**
* How long it takes — **Minutes**, **Hours** or **Days**
For facts trapped in detail images, we write the HTML block that lists those facts as a table. Keep the images; add the block.
### 4. You confirm the fix with one button
After the change is live, press **Verify** on the finding. We re-fetch only the affected pages the way AI does and tell you on the spot whether it is resolved. Resolved findings move to **Verified fixed**. If the problem returns later, the finding reopens on its own.
### 5. We re-run it every month
Sites change. The audit runs again automatically once a month and picks up anything new. You can also run it yourself any time with **Re-run audit**.
### 6. You get a report
**Export** downloads a PDF you can hand to leadership or your developers as-is. It contains the summary, the per-page verdicts, the causes, the proposed changes and how to confirm each one.
## What you see on the screen
* **Can AI read this site?** — the one-line answer, for example "1 of 23 sampled pages is only partially readable".
* **Access gate** — a grid of pages against the ChatGPT, Perplexity and Google bots, showing which ones get in.
* **Policy check** — things that are facts, not faults, such as opting out of AI training bots.
* **Findings** — grouped in the order Fetch → Index → Extract, each with the number of pages affected.
* **Reader view** — for any page, **Text the bot received** on the left and **Text the browser shows** plus **Text inside images** on the right, with every fact marked readable or hidden.
* **Details** — the evidence, the **Proposed changes** with a copy button, the **Verify** button and the status.
## What the audit does not do
* **It does not give a score.** A number like "AI readiness 78" leads to no action. You get per-page verdicts and evidence instead.
* **It does not recommend things that have no effect.** An llms.txt file, generic structured markup, or padding word counts have been checked against real citation data and left out.
* **It does not treat blocking AI training bots as a problem.** Blocking GPTBot or ClaudeBot is a company policy choice. We only flag it when the search and answer bots are blocked.
* **It does not create new content.** The audit fixes pages and settings you already have. New pages are the job of [On-site](/dashboard/on-site).
## What you need to prepare
* Nothing to run it. A site address is enough. No access, no installation.
* Someone to apply the fixes: a developer or a content owner. Most fixes take minutes to hours.
* If you are on Cafe24, Godo or Shopify, the finding tells you where in the admin to paste the change.
## FAQ
That is the point. AI does not look at the screen; it reads the raw document the server sends. When the screen and the raw document differ, AI only sees the raw document.
No. Keep the images and add a short block that states the key facts from them as text. We write that block for you.
Being readable is a precondition, not a guarantee. A page AI cannot read is never cited. Track the outcome after your fixes in [Visibility](/dashboard/visibility) and [Citation](/dashboard/citation).
No. Training bots are a policy choice and appear under **Policy check** only. The audit flags a block only when the search and answer bots are refused.
## Related
The screen-by-screen guide.
Check the outcome after your fixes land.
See which sources AI actually pulls from.
Create new pages built to be cited.