# Main entity solved

Schema is how machines read your content. Every page is about one thing. Lumar names it as a schema.org type, on every page of your crawl, and tells you where your markup says something else.

## One page, one main entity

A product page mentions many things: the shop, a breadcrumb, the price, reviews, a block of FAQs. Search engines and AI assistants read schema to work out which of them the page is about. That one thing is the main entity. Lumar names its most specific schema.org type, here a Product.

## Every type, two rounds

- About 740 schema.org types can be what a page is about. Types that fit any page, like WebPage, are set aside first.
- They are too many for one question, so round one splits them into three questions, asked together. Every type is weighed; nothing is ruled out early.
- Each question keeps its five strongest types. If a question cannot be answered, it is split and asked again; anything still unanswered is reported, never dropped.
- Round two weighs the fifteen finalists together, side by side.
- The answer comes with a confidence score, so close calls show their runner-up.

## Read in code, not guessed

Lumar reads your JSON-LD and microdata directly. No model looks at your markup. It finds the part of your markup that describes this page, and reads what it says the page is about. An FAQ block on the page names its own main entity, but it describes the block, not this page, so it never counts as the page's answer.

## Wrong is a different kind of thing

- Markup says Article, the page is a BlogPosting: same branch of schema.org, just less specific, so a match.
- Markup says Article, the page is an Event: different branches, a real mismatch.
- Markup names nothing: reported as missing, never counted as wrong. On lumar.io, comparing exact types flagged 15 of 17 blog posts; comparing branches leaves the 2 real ones.

## What it found on lumar.io

| Page | Main entity | Confidence | Declared | Verdict |
| --- | --- | --- | --- | --- |
| lumar.io/about/ | Corporation | 0.99 | none | No schema |
| lumar.io/platform/analyze/ | SoftwareApplication | 0.99 | none | No schema |
| lumar.io/podcast-human-first-digital/ | RadioSeries | 0.90 | none | No schema |
| lumar.io/blog/best-practice/seo-for-news-publisher-websites/ | BlogPosting | 1.00 | Article | Schema agrees |
| lumar.io/webinars-events/upcoming-webinar-beyond-the-browser/ | Event | 0.64 | none | No schema |
| lumar.io/ebooks/2025-seo-trends/ | Book | 0.60 | Article | Wrong type |

## Built for every page of every crawl

About 740 types weighed for every page, in two rounds, by one model built to choose, not a chatbot, in about 5 seconds a page. Fifteen finalists are weighed together in round two.

Request a demo: https://www.lumar.io/request-demo/

Page: https://agentic-readiness.lumar.io/main-entity