Schema markup in plain terms: what to add and what to skip

Schema markup is a set of machine-readable labels you add to a page so that software knows what it's looking at. A person can see that £49.99 is a price and that five gold stars are a rating. A crawler is inferring until you tell it outright. In my plain-terms guide to technical SEO I summed structured data up in one line as labels for machines, and this is the longer version, including the part most guides skip, which is that Google has spent three years reducing what those labels earn you.

I add structured data to every site I build, and I'd still call it one of the most oversold corners of SEO. Both halves of that sentence are true at once. The labels are cheap and useful, the visible rewards have been shrinking since 2023, and anyone pitching schema as a rankings lever is a few years out of date.

What the markup actually does

It does two jobs. The first is eligibility for rich results, the extra furniture Google can attach to a listing, review stars, prices and stock levels, breadcrumb trails, event dates, recipe cards. The second is quieter. It helps machines understand your pages with confidence rather than inference, which matters a little more each year as more of the things reading your site are not people.

Two limits are worth stating plainly. Markup is not a ranking factor, and Google has been consistent about that for years. A page with perfect schema and thin content still loses to a good page with none. And eligibility is not a guarantee. You can implement everything correctly and Google may still choose not to show the feature.

In practice the markup is a small block of JSON-LD, Google's recommended format, sitting invisibly in the page's code. Visitors never see it, and it changes nothing about the design. It's purely a message to the machines.

Three years of shrinking rewards

The reason I keep saying the rewards have shrunk is that Google has been steadily retiring rich results.

In August 2023 it deprecated HowTo results and restricted FAQ dropdowns to well-known government and health sites. In late 2024 the sitelinks search box went. In June 2025 seven little-used types were retired, claim review, estimated salary, special announcement and vehicle listings among them. And on 7 May 2026 FAQ rich results stopped appearing for everyone, closing the government-and-health exception and ending the feature for good.

The lesson isn't that schema is dying. It's that any single rich result is rented visibility, and Google can repossess it without much notice. The features come and go; the understanding the markup provides is the durable half of the bargain. That's why the sensible strategy is to mark up what your pages actually are, rather than chasing whichever feature currently looks rewarding.

What's still worth adding

Google's list of supported features is long, but for most businesses the shortlist is short. As of mid-2026:

Markup	Who it's for	What it gets you
Organization	Every business	Brand identity, logo and company details in knowledge panels
LocalBusiness	Businesses serving a physical area	Hours, contact and review details tied to your premises
BreadcrumbList	Everyone	Tidy page paths in results instead of raw URLs
Product (with offers and ratings)	Ecommerce	Prices, stock and stars in results and merchant listings
Article	Blogs and publishers	Better headline and image treatment, modest but free
Event, JobPosting, Recipe, VideoObject	Only if that's what the page is	Dedicated result features that remain well supported

Product markup is the biggest visible win left. Stars and prices on a results page change click behaviour in a way almost nothing else on this list does, and for stores the merchant listing formats lean on the same markup. I covered where it fits in the wider template in product page SEO.

LocalBusiness is the local equivalent, and it works best when it agrees exactly with your Google Business Profile, same name, same address, same phone number, so the two sources confirm each other. I've written separately about which Business Profile work actually moves rankings.

Organization is the one nearly everyone skips and shouldn't. It goes in the template once, tells Google who the site belongs to, and quietly feeds knowledge panels and brand results. Ten minutes, done forever.

What to skip

New FAQ or HowTo markup, if the goal is a Google rich result. Those results are gone. Existing markup does no harm, Google has said leftover structured data causes no problems, so there's no rush to strip it out, but don't invest new effort in it for Google's benefit. You'll notice this post still ends with an FAQ section. It's there because readers scan for questions, and the markup behind it stays because removing it would gain nothing.

The retired types. Anything on the June 2025 deprecation list earns nothing now. If a plugin added those types for you, they're clutter rather than markup.

Review stars on your own testimonials. Marking up reviews of your own business, hosted on your own site, is what Google calls self-serving, and it stopped showing stars for those back in 2019. The rule hasn't changed, and neither has the number of sites quietly trying it anyway.

Everything the plugin offers, just in case. Markup should describe what the page actually is. Spraying types across pages they don't match achieves nothing at best, and markup that misleads, invented ratings being the classic, can earn a manual action. If you wouldn't defend the label out loud, don't add it.

Where AI search fits

FAQ schema's final deprecation drew extra attention because it had been widely recommended as a way to feed AI systems. It's worth being honest about what's actually known here. Microsoft has said publicly that schema helps Bing's language models understand content. Google has said nothing tying its AI features to markup, and the testing I've seen doesn't show markup alone winning citations in AI answers. My working position is that structured data is cheap insurance for machine readers of every kind. Add the types your pages justify anyway, then let clear writing and real authority do the heavy lifting, because nobody's markup is rescuing a page an AI doesn't consider worth quoting.

How I implement it

Structured data belongs in the template, not pasted into individual pages. Hand-added markup drifts, prices change, products sell out, and stale labels are worse than none. On the sites I build, the markup generates from the page content itself, so the two can't disagree. It's a standard part of the technical SEO work I do on every project, alongside the crawling and indexing fundamentals in my technical SEO checklist.

Checking your own site takes five minutes: run a key page through Google's Rich Results Test, then keep an eye on the enhancement reports in Search Console, which flag markup that breaks. That's the whole maintenance burden. This is one of the lower-effort corners of SEO, which is exactly why it attracts overselling.

Schema markup FAQs

Is schema markup a ranking factor?

No. Google has been consistent that structured data doesn't boost rankings. What it does is make pages eligible for rich result features and help machines understand content accurately. Where a feature like review stars still appears, it can lift clickthrough, which is a real benefit, just an indirect one.

Should I remove FAQ schema now the rich results are gone?

There's no need. Google has said leftover structured data causes no problems, and FAQPage remains valid schema.org vocabulary that machines can still read. Remove it if you're tidying the codebase anyway, but don't make a project of it. The mistake would be adding new FAQ markup expecting a Google feature that no longer exists.

Should I use JSON-LD or microdata?

JSON-LD. It's Google's recommended format, it sits in one self-contained block rather than being woven through your HTML, and it's far easier to maintain and debug. Microdata still validates, but I haven't had a reason to choose it in years.

Does schema markup help with AI search?

Partly, and the evidence is thinner than the advice industry suggests. Microsoft has confirmed schema helps Bing's language models interpret content, while Google has made no equivalent commitment for its AI features. Treat markup as cheap insurance that helps machines read you accurately, not as a lever that wins AI citations on its own.