---
title: Schema
description: Check one page or a whole site for structured data, keep every result in the project history, and close the loop from gaps to deployed fixes.
---

Use **Schema** to see which Schema.org structured data your pages expose, what is missing, and whether a fix actually reached the live site. Every check is saved in the project, so you can hand results to a developer, come back later, and check again to confirm the gaps are gone.

The work follows one loop: **check** a page or a whole site, **generate** the missing markup, **deploy** it on the site, and **check again**. The guide band at the top of the page walks through these steps and marks the ones you have done. Close it with **✕** when you no longer need it; it stays hidden for that project in your browser.

## Check a page

Open **Schema** from the sidebar, keep **Single page** selected, enter the full address of a public page and select **Check schema**. If you leave the field as it is, PromptEye checks your project's domain. An address without `https://` is treated as HTTPS. The page must be publicly reachable: pages behind a login, drafts and local previews cannot be checked.

When the check finishes, its result opens on the right and the page appears in **Analysis history**. If the page cannot be fetched, you see an error and nothing is added to the history, so you can fix the address and try again. Some sites block automated requests; PromptEye then retries by opening the page in a browser, which takes longer. If the site blocks that as well, the check fails.

## Check a whole site

Switch to **Whole site**, enter the domain and select **Collect pages**. PromptEye reads the site's XML sitemap (`/sitemap.xml` or `/sitemap_index.xml`) and lists the pages up to two levels below the address you entered, shallowest first with the homepage at the top. Untick pages you do not need, such as terms or privacy pages, or use **Add more addresses** to paste your own, one per line.

- **Up to 50 pages per run.** If the sitemap lists more, you see the 50 shallowest pages and a note about the limit. Run the audit again for the rest.
- **Duplicates.** Addresses that point to the same page (for example with and without `www`) are moved to a separate **Duplicates** list and left unticked.
- **No sitemap.** If the site has no sitemap, the list stays empty; paste the addresses you want instead.

Select **Check N pages** to queue the checks and return to Schema. The pages appear in the history as **Checking…** and fill in one by one; the checks run on PromptEye's side, so you can leave the page while they finish. A page that cannot be fetched shows **Could not be checked**; select **Remove** to clear that row. A page that is already waiting in the queue is not queued twice.

Only people who can edit the project can run checks, collect pages or generate markup. People with view-only access see the same history and results without those buttons.

## Read the results

The checker reads JSON-LD and, when it cannot parse any JSON-LD block, Microdata markup on the page. JSON-LD can be written as a single object, as a list of objects or inside `@graph`; every object found counts as one schema. Microdata is read with its nesting: items inside another item, such as the list items of a BreadcrumbList, an Offer, a Brand or a PostalAddress, belong to their parent and are not counted as separate schemas. A block of markup with no `@type` makes the page **Missing required fields**, and the page's details show a notice explaining which problem to fix. It does not guess structured data from visible text. When the page source has no structured data, PromptEye also opens the page in a browser and waits a few seconds, so markup added by JavaScript, for example through Google Tag Manager, is found as well; such a check takes longer. Some AI crawlers read only the page source and never see markup added by JavaScript, so keep important markup in the page source.

Each detected schema shows its required and recommended properties as present or missing. A page gets one of three results:

- **Valid** — every detected schema has its required properties.
- **Missing required fields** — at least one schema lacks a required property. Fix these first.
- **No schema** — the checker found no structured data it recognizes. The details then suggest a starting point: Organization and WebSite for a homepage, BreadcrumbList for any other page.

The **Recommendations** column counts what is left: **Critical** is the number of missing required properties, **Recommended** the number of missing recommended ones. The checker requires the properties Google asks for, plus a few extra ones that help AI assistants understand the page. For some types that means more than the page-level fields:

- **Product** needs an image, and its offer needs a price and a currency. If the offer is a list, every offer needs both. A price range (an aggregate offer with a low price) counts as a price, and a price of 0 counts too.
- **Recipe** needs an image.
- **Job posting** needs a job location. Remote jobs marked as `TELECOMMUTE` are not asked for one.
- **Software application** needs a price in its offer (0 is fine, a currency is not required) and either a rating or a review.

Related types, such as a mobile app or a product group, are checked with the rules of their parent type. A page that passed before can now show **Missing required fields**; saved results keep their old status until you check the page again.

A valid result means the checker found the fields it evaluates; it does not guarantee rich results, indexing or AI citations.

## Use the analysis history

**Analysis history** lists every checked page, one row per page with its newest result. It is shared with everyone who can open the project, and earlier checks are kept, not overwritten.

PromptEye treats addresses that differ only in `www`, `http`/`https`, a trailing slash, a `?query` or a `#fragment` as the same page, so `http://www.example.com/blog/` and `https://example.com/blog` share one row and one history. Pages are grouped by site; subdomains such as `shop.example.com` form their own group. Filter by result or type, search by address, collapse a site, or clear **Group by site** for a flat list.

Select a site's row to open its summary: how many pages are valid, have gaps or have no schema, which types appear on how many pages, what is missing across the site, and the list of its pages.

## Close the loop with Recommendations

For every site with work left, **Recommendations** shows one row with the loop's four steps and the action for the step you are on:

1. **Checked** — the site has checked pages.
2. **Generate** — select **Generate** to build markup for every page that needs it (see [Schema Generator](../schema-generator/)).
3. **Deploy** — **Download** the ZIP package for your developer, and after the files are on the site select **Mark as deployed**. PromptEye cannot see your deployment, so this step is your confirmation.
4. **Check again** — re-checks every checked page of that site. Keep Schema open until it finishes; the result opens in the site summary.

The loop closes when every page of the site was checked after you marked the deployment and all of them are valid. Generating a newer package after marking the deployment moves the site back to **Deploy**. **Priority** grows with the share of the site's pages that still have issues.

## Generated packages

When you finish the generator, the package is saved in the project. A page's details show **Generated schema package** with the date, a **Download package (.zip)** button, **Generate again** (which reopens the generator with the same data) and up to five earlier versions to download. A site package from the whole-site generator appears in the site summary.

While a package is newer than the page's last check, the page shows **waiting to be deployed** and its history row offers **Check again ›**, so you can confirm the deployment worked.

## Compare before and after a fix

A page's details show its newest result and a **Check history** list with every earlier check of that page, newest first. Select an older check to show it in place of the newest one and switch back and forth. **Check again** re-checks the exact address used in the newest check; the new result becomes current and the earlier one stays in the history.

## If you cannot open Schema

The sidebar disables project workspace items when you do not have access to an active project workspace. A missing sitemap or missing Schema.org code does not hide **Schema**. If the item is disabled, select a project you can access or ask an owner for access.

## Common questions

### Are results saved if I leave the page?

Yes. Every successful check is saved in the project history. Whole-site checks keep running after you leave; only **Check again** for a whole site from Recommendations needs the page to stay open.

### Why does Collect pages find nothing?

PromptEye found no sitemap at `/sitemap.xml` or `/sitemap_index.xml`, or the sitemap has no pages under the address you entered. Paste the addresses with **Add more addresses** instead.

### Why does it find zero schemas?

The page did not expose structured data the checker recognizes. Confirm the address is the intended public page and that its JSON-LD or Microdata is deployed there, then check again. Malformed JSON-LD blocks are ignored.

### I checked again but the result did not change.

The checker reads the page as it is served now. If the new markup is not in the response yet — the change is not deployed or a cache still serves the old page — the result stays the same. Confirm the markup is on the exact public URL and check again.

### Does marking a site as deployed change anything on my website?

No. It only records that you deployed the latest package, so Recommendations can move to **Check again**. Nothing in PromptEye changes your site.

### Can I delete a check?

No. Checks stay in the project history. A failed whole-site row can be removed with **Remove**.

### Can I check a page on another domain?

Yes. Any public page can be checked; it is saved in the current project's history under its own site group.
