What is microdata in SEO?

If you open the source code of a product page and see attributes like itemprop="price" scattered through the HTML, that is microdata: structured data written onto the same tags the customer reads, instead of in a separate block. We are not talking here about statistical microdata, the files with the individual records of a survey, which share the name and a good share of the searches. They matter to you for two reasons: Google still reads them as structured data, even though it recommends another format,⁠^{[1]Google Search Central. Introduction to structured data markup in Google Search} and if your store already has them, they coexist with any other markup you add.

How it works

The HTML standard defines microdata as a mechanism for adding to a document nested groups of name-value pairs, in parallel to the content that already exists.⁠^{[2]WHATWG. HTML Standard: Microdata} There is no separate block: the data is in the tag that shows it. Five attributes do it:⁠^{[2]WHATWG. HTML Standard: Microdata}

The value of each property comes from the tag according to its kind. On a meta it is its content attribute; on a link, its href; on a time, its datetime; on a data, its value; on the rest, the text the tag contains, which is what the visitor sees.⁠^{[2]WHATWG. HTML Standard: Microdata} That last rule is what defines the format: in most cases, what the machine reads and what the person reads are literally the same thing.

The words written in itemtype and itemprop are those of Schema.org. Its getting started guide, which teaches the vocabulary precisely with microdata, sums it up in three steps: itemscope marks the part of the page that talks about one thing, itemtype says what the thing is (for example https://schema.org/Movie) and itemprop labels each of its properties. An item can carry others inside it, like a movie's director described in turn as a person.⁠^{[3]Schema.org. Getting started with schema.org using Microdata}

Google supports three structured data formats. It describes microdata as an open HTML specification for nesting structured data within the content, with attributes on the tags, and RDFa as an HTML5 extension that works in a similar way. It recommends the third, JSON-LD, which goes in a script tag and is not mixed in with the visible text.⁠^{[1]Google Search Central. Introduction to structured data markup in Google Search} Recommending one is not discarding the others: well-made microdata markup works just as well for qualifying for rich results, and the Rich Results Test reads JSON-LD, RDFa and microdata alike.⁠^{[4]Search Console Help. Rich Results Test} There is no need to rewrite it just because of the format.

In a store

A shirt's offer marked up with microdata. Fictional example: the store, the product, the image and the price are made up.

<div itemscope itemtype="https://schema.org/Product">
  <h1 itemprop="name">Calella linen shirt, blue</h1>
  <img itemprop="image" src="/img/calella-blue.jpg" alt="Calella linen shirt in blue" />
  <div itemprop="offers" itemscope itemtype="https://schema.org/Offer">
    <span>$29.90</span>
    <meta itemprop="price" content="29.90" />
    <meta itemprop="priceCurrency" content="USD" />
    <link itemprop="availability" href="https://schema.org/InStock" />In stock
  </div>
</div>

Three things the example shows. The first: the name is not written twice, because the text of the h1 is the value of name. The second: the product carries a second item inside it, the offer, which is an Offer with its own itemscope; that is how types nest in this format. The third is the price. The customer sees "$29.90", but Schema.org asks for the price with a decimal point and no symbol, and the currency separately, with its ISO 4217 code,⁠^{[5]Schema.org. Offer} so the value for machines goes in meta tags, and availability, which in the vocabulary is a URL, in a link.⁠^{[2]WHATWG. HTML Standard: Microdata}

That has a cost. As soon as a value does not come from the visible text but from a meta, it can stop matching what the page shows, just like a forgotten JSON-LD block. And because the attributes live inside the template, switching themes or rewriting a template can wipe out the markup without anyone deciding it: the new HTML does not have to carry the same attributes.

How to check it

1. Open the source code of a product page and look for itemscope. If it shows up, the page carries microdata; the itemtype next to it tells you what type each item is.
2. Also look for application/ld+json. If the page carries both formats, check that they describe the same product with the same name, the same price and the same availability.
3. Run the URL through the Rich Results Test. It reads all three formats, tells you which rich results it detects and with what errors, and also accepts a pasted code snippet.⁠^{[4]Search Console Help. Rich Results Test}
4. Review every meta and every link with itemprop. They are the values that do not come from the visible text, and each one has to correspond to something the customer sees on the product page.⁠^{[6]Google Search Central. General structured data guidelines}
5. Repeat the review after switching themes or templates, because the markup travels with the HTML.

Related concept: JSON-LD

Microdata and JSON-LD write the same vocabulary; what changes is where the data lives. In microdata it is inside the visible tags, so the page's text is the value; in JSON-LD it is in a separate script block that is not mixed in with the text the user sees.⁠^{[1]Google Search Central. Introduction to structured data markup in Google Search} The first makes it harder for the two to say different things, because they are the same text, and easier to lose the markup when touching the template. The second lets you change the markup without touching the layout, which is why Google recommends it,⁠^{[1]Google Search Central. Introduction to structured data markup in Google Search} with the opposite risk: that the block falls behind when the page changes.

To keep going, the structured data reading path puts the rest of the terms in order, and the guide to auditing a product page's structured data walks through a product page property by property, in whatever format it has.