> For the complete documentation index, see [llms.txt](https://alhena.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://alhena.gitbook.io/docs/features/ai-visibility/agent-readiness.md). # Agent Readiness Check whether AI shopping agents can get into your store, read your pages and understand your products, and see which fixes raise your score the most. The **Agent readiness** tab checks your store the way an AI shopping agent sees it. Alhena reads a sample of your pages and the files agents look for, scores your store out of 100, and lists the fixes that would raise the score, biggest first. The other AI Visibility tabs show how often AI engines recommend you. Agent readiness covers the step before that: whether an agent can get into your store, read your pages and understand your products.

{% hint style="info" %} Agent readiness sits under **AI visibility** in the dashboard sidebar, on desktop. If you don't see it, ask your Alhena contact to switch it on. {% endhint %} ## Run a check 1. Open **AI visibility → Agent readiness**. 2. The first time, enter your store's web address and click **Check my store**. After that, click **Run check** whenever you want fresh results. 3. Give it a few minutes. You can leave the page while the check runs, and your last results stay on screen until the new ones are ready. Each store gets **4 checks a month**. The line under **Run check** shows how many are left and the date they reset, the first of each month. A check that fails, or that your store blocks, doesn't use one up. Alhena also checks your store each time your scheduled AI Visibility analysis runs, and those checks don't use any either. Under the store name, the header shows when the store was last checked, how many pages Alhena read out of the pages in your sitemap, and how many checks ran. ## How Alhena reads your store Alhena doesn't crawl every page. Pages of one type are built from one template, so a sample of each type finds what's wrong with all of them. 1. It reads your sitemap and sorts the pages by type: product, collection, content and root pages. 2. It reads a sample of every type and every product format the way AI agents do: from Alhena's servers, without running JavaScript. 3. It checks the files agents look for across your whole store: `robots.txt`, your sitemap, `llms.txt` and the agent files under `/.well-known/`. An issue found on every page read of one type is reported for **All pages** of that type. Issues that vary from page to page show as **Most pages** or **Some pages**, so you know to fix those page by page. Click the info icon next to **Pages read** to see the page counts behind your check.

## The score Your store gets one score out of 100 across 16 checks, in three buckets: | Bucket | Points | What it covers | | ------------------------- | ------ | ---------------------------------------------------------------------------------------- | | **Product facts** | 50 | Whether your product pages carry what an agent needs to quote and compare a product. | | **Access & crawlability** | 30 | Whether agents are allowed in and can read your pages. | | **Agent protocols** | 20 | The files and tools made for AI agents, like product search, cart, checkout and sign-in. | Product facts carry the most weight because they are what an agent tells a shopper, and they are the part a store controls. Store platforms handle `robots.txt`, readable pages and agent files by default, so those alone can't earn a high score. The score puts your store at one of four levels: | Level | Score | | ------------------- | -------- | | **Perfect score** | 100 | | **Most in place** | 80 to 99 | | **Basics in place** | 50 to 79 | | **Needs work** | 0 to 49 | Checks Alhena couldn't run, and checks you [exclude](#exclude-a-check-that-doesnt-apply), are left out of the score. They don't count as zero. To see the points behind every check in the dashboard, open the **⋯** menu next to **Run check** and choose **How the score works**. {% hint style="info" %} The score covers whether agents can reach and read your store. It doesn't cover whether each fact is correct, or whether AI tools recommend you today. For that, see [Overview](/docs/features/ai-visibility/overview.md) and [Prompts](/docs/features/ai-visibility/prompts.md). {% endhint %} ### The 16 checks **Product facts (50 points)** | Check | Points | What it looks for | | ------------------------------- | ------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Product facts in your pages** | 50 | Ten facts in each product page's structured data: name, image, price, stock status, brand, description, shipping, returns, GTIN or MPN, and ratings or reviews. | **Access & crawlability (30 points)** | Check | Points | What it looks for | | ------------------------------------- | ------ | ------------------------------------------------------------------------------------------------------------------------- | | **Pages readable without JavaScript** | 13 | Whether your content is in the page itself, or only appears after JavaScript runs. | | **AI crawler rules** | 9 | Whether your `robots.txt` lets AI crawlers reach your pages. | | **Indexability** | 4 | The signals that let engines include and understand a page: noindex, canonical, title, meta description and main heading. | | **Sitemap** | 3 | Whether your store publishes a sitemap, which agents use to find your pages. | | **Page weight** | 1 | How large each page's HTML is. | **Agent protocols (20 points)** | Check | Points | What it looks for | | --------------------------------- | ------ | ---------------------------------------------------------------------------------------------------------------------------------------------- | | **Store capabilities file** | 5 | The Universal Commerce Protocol (UCP) file that tells agents what they can do in your store, like search products, build a cart and check out. | | **Store tools for AI assistants** | 4 | Whether an AI assistant can find and call your store's tools, like product search. | | **Personal Agent Protocol file** | 3 | Whether personal agents can find your store's own agent and tools through a `poppy.json` file. | | **Sign-in discovery** | 2 | Whether agents can find how to sign a shopper in to your store, which they need for orders and returns. | | **In-page tools** | 1 | Whether your pages offer tools to an agent browsing them. | | **llms.txt** | 1 | Whether you publish an `llms.txt` file, a short guide to your site for AI tools. | | **Checkout file** | 1 | The Agentic Commerce Protocol (ACP) file that lists how agents can check out. | | **Agent Skills index** | 1 | Whether your store lists skill files that teach agents how to use it. | | **Markdown pages** | 1 | Whether an agent can ask for a product page as Markdown, which is cleaner for it to read than HTML. | | **Content Signals** | 1 | Whether your `robots.txt` says how AI tools may use your pages: for search, for answers or for training. | A check tagged **Early** is a standard that is still new, often a draft that few stores and agents use yet. Early checks carry few points. ## Read your results Four views sit under the header: **Overview**, **Fixes**, **Issues by page type** and **Checks**. ### Overview The Overview is the summary of your last check: * **Readiness score**: your score, your level and the score where the next level starts. * **Access & crawlability, Product facts, Agent protocols**: the points earned in each bucket, the number of fixes, and the points still to gain. * **Top fixes**: the five fixes worth the most points. Click one to open it. * **Fix plan**: the fixes you plan to do next and the score they would get you to. * **Issues by page type**: the main issues on your product, collection, content and root pages. * **Checks**: how many checks pass, pass in part or are missing. * **Needs attention**: shown only when something got in the way of the check, such as bot protection that turned Alhena away from some pages. Your score also shows on the AI Visibility [Overview](/docs/features/ai-visibility/overview.md) tab, with a **View report** link back to this tab. ### Fixes Every fix in one table, ordered by the points it's worth.

| Column | Meaning | | ------------ | --------------------------------------------------------------------------------------------- | | **Plan** | Tick a fix to add it to your [fix plan](#fix-plan). | | **Fix** | What to change, and what's wrong today. | | **Where** | The pages it applies to, or **Store-wide** for a file or setting that covers the whole store. | | **Fixed in** | Where the change is made, such as Shopify admin, your shipping settings or your review app. | | **Points** | What the fix adds to your score. | Use the chips above the table to filter by bucket. Click a fix to open it:

* **What's wrong** and **How to fix**, with example markup where it helps. * **Where**: the page types affected and how many pages, with example pages you can open or copy. * **Copy fix prompt**: copies a ready-made prompt with the problem, the fix and example pages. Paste it into an AI assistant, or send it to your developer. * **About this check**: the check behind the fix. ### Fix plan The fix plan is your short list of what to do next. It starts with your top three fixes. To change it, click **Edit plan** and tick or untick fixes in the **Plan** column of the Fixes view. The plan shows your score today, the score the plan would get you to, and who usually makes each change, such as catalog and merchandising or a developer. Click **Copy brief** to copy a message for your team or agency. It lists each fix in the plan, who does it, what to do and an example page: ``` I'm making my online store, alhenademo.com, ready for AI shopping agents. Please make these 3 changes: 1. Add GTINs to product data. Product pages have no GTIN or MPN. Who: Catalog and merchandising · Fixed in: Shopify admin What to do: Add gtin or mpn to your Product JSON-LD. On Shopify, fill in each variant's barcode. If a look or bundle is sold as one item, give it an MPN (your style code). Example page: https://alhenademo.com/products/solenne-meridian-nightstand-white-oak ... ``` Your plan is saved in the browser you build it in, so each teammate keeps their own. ### Issues by page type Pick a page type on the left to see its issues, how far each one reaches (**All pages**, **Most pages** or **Some pages**), the points for fixing it, and example pages.

Points are for the whole fix, across every page type it touches. Store-wide issues, like a missing agent file, aren't tied to a page type. You'll find those in the Fixes view. ### Checks All 16 checks with their result and the points earned. Checks that pass are grouped at the bottom. **Product facts in your pages** also lists which of the ten facts are on all, some or none of your product pages.

Click a check to open **About this check**: your result, why the check matters, how Alhena tests it, how it's scored, and references for the standard behind it.

## Exclude a check that doesn't apply Some checks may not fit your store. A brand that doesn't collect reviews, for example, can't add ratings. To leave a check out of your score: 1. Open its fix and click **Exclude this check**. 2. Pick the reason. The form shows what your score becomes. 3. Click **Exclude**.

Alhena keeps running an excluded check, and you can bring it back at any time: click **Excluded checks** in the header and choose **Include again**. ## Download a report Open the **⋯** menu next to **Run check** and choose **Download report** to get your results as a PDF you can share. ## Check another store The store menu at the top left of the tab shows your store and your other profiles. **Check another store…** in the same menu is a Pro feature. It creates a new profile for that store, with its own Agent readiness, its own fixes and its own 4 checks a month. Your own store's score doesn't change. ## When a check can't finish * **Your store's bot protection turns the check away.** AI agents that read from their own servers may be refused the same way, so Alhena reports it. When only some pages were refused, it shows under **Needs attention**. When Alhena couldn't read enough to score your store, you see what it could still check in place of a score. In both cases the fix is to check that AI search and shopping agents are allowed in your CDN or bot settings. A blocked check doesn't use one of your monthly checks. * **The check fails or takes too long.** Run it again. A failed check doesn't use one of your monthly checks.