What does schema markup do for an AI engine?
Schema markup tells an AI engine what kind of page it is reading, who stands behind it, when it was updated and which parts answer which question. Schema markup is structured data, usually JSON-LD, that states those facts in a fixed vocabulary from schema.org. Google's documentation puts it this way: "Structured data is a standardized format for providing information about a page and classifying the page content." Schema is common already: data from GeoTest's scans of 5,278 homepages between 4 and 27 October 2026 shows 79.0% carried some JSON-LD, 55.2% marked up an Organization or LocalBusiness, and 27.2% a BreadcrumbList. Schema cannot make weak content quotable, but it stops strong content from being misread: an article without a date, a business without a name and address, or a Q&A an engine does not recognise as one.
Which schema types should you add first?
The schema types to add first are the six below, because they describe who you are and what each page is. JSON-LD is the script format Google recommends for this markup, placed in the page's HTML.
| Type | Where | What it tells the engine |
|---|---|---|
| Organization | Sitewide | Who you are, your logo, contact details and profiles |
| WebSite | Homepage | The site's name and URL |
| Article or BlogPosting | Articles and guides | Headline, author, published and modified dates |
| BreadcrumbList | Every inner page | Where the page sits in the site |
| FAQPage | Pages with visible Q&A (only 7.9% of the 5,278 homepages) | Each question and its answer |
| Product, LocalBusiness, Service | Where they apply | Prices, locations, what you sell |
Does schema have to match the visible page?
Schema has to match the visible page, and that rule matters more than which types you add. Markup that says something the page does not, such as FAQ answers that are not visible, a rating with no reviews, or a dateModified that disagrees with the visible date, is worse than none. GeoTest checks that the JSON-LD parses, that its primary type matches the page, and that its properties agree with the visible text. Google's own guidance on AI features states plainly that markup is not a shortcut: "There's also no special schema.org structured data that you need to add." And FAQ rich results are gone: Google wrote in 2023, in a Search Central post, that they "will only be shown for well-known, authoritative government and health websites", and as of 7 May 2026 it stopped showing them in Search for every site, through a deprecation notice on its FAQPage documentation.
Does FAQPage schema still help now that Google has dropped FAQ rich results?
FAQPage schema still helps now that Google has dropped FAQ rich results, because the rich result was one Google display feature, not what the markup tells an engine about the page. Google narrowed FAQ rich results to government and health sites in 2023 and stopped showing them for every site on 7 May 2026, as Search Engine Land reported; FAQPage is still a valid schema.org type that engines parse. According to the GeoTest glossary, GeoTest credits FAQPage markup with 2 points "only when at least half its questions are visible on the page; markup with no visible Q&A scores 0." In GeoTest's scans of 5,278 homepages, only 7.9% carried it.
How do you check that schema markup is valid?
Checking that schema markup is valid takes two tests: the JSON-LD must parse, and what it says must match the page. According to the GeoTest glossary, "GeoTest checks that the main schema type fits the page, that every JSON-LD block parses, and that names, ratings, prices and dates agree with the visible text." A free Quick Scan runs those checks on any page in about 15 seconds, and the schema.org vocabulary lists the properties each type accepts.
Which date fields in Article schema matter for AI engines?
The date fields in Article schema that matter for AI engines are datePublished and dateModified, because they tell an engine how current the page is. GeoTest gives the full 5 recency points to an article updated within 3 months, 3 points within 6 months and 1 within 12 months, and 2 more points when the schema dateModified matches a date shown on the page within 7 days. A dateModified with no visible date to confirm it earns 1 of those 2.
How many business websites get schema markup right?
Most business websites carry some schema markup, but fewer get the type right. In the GeoTest Benchmark Report 2026, 34.8% of the 4,999 sites lacked structured data of a type that fits the homepage, and the sample earned 68.3% of the Structured Data points available to it. A page that fixes its type, its BreadcrumbList and its meta tags can usually close that gap in an afternoon.
How much is schema worth in a GEO Score?
Schema is worth 12 of the rubric's 100 points in a GEO Score, under Structured Data: a primary type that fits the page, JSON-LD that parses and agrees with the page, a BreadcrumbList, supporting types, and the meta description, Open Graph and Twitter card. Because it is an afternoon of work, it is usually the fastest category to fix. The larger gains are in the text: the 2023 GEO study found that adding sources, statistics and quotations raised visibility in AI answers, with a 40% lift at the top end. Structured Data is the rubric category that scores this markup, and it is the third-heaviest after Citation Worthiness and Answer Structure. FAQPage schema: the fastest GEO win walks through the most useful type in detail.
