Schema markup in 2026 still moves rankings
A funny thing happens when you bring up Schema.org structured data at an SEO meetup in 2026. The senior people nod. The newer people look slightly embarrassed for you. Schema is one of those topics that has been around long enough to feel dated, and so the assumption is that everyone has done it and that the marginal value has been wrung out.
Both of those assumptions are wrong.
Most sites I audit still have either no structured data, broken structured data, or structured data that is technically valid but useless. And the value has, if anything, gone up. Not because Google quietly upweighted it (the official line has always been that schema is a hint, not a ranking factor in the narrow sense), but because the AI engines genuinely depend on it. Schema is now the cheapest way to be readable to Google, Bing, ChatGPT browse, Perplexity, and Claude in a single edit.
The state of structured data in 2026
Schema.org is fifteen years old. The vocabulary is huge. Most sites use a tiny fraction of it: Article, Organization, BreadcrumbList, FAQPage, maybe Product if there is a shop. That is a sensible starting point. The mistake is treating it as a one-and-done.
What has changed in the last two years is that the consuming surfaces have multiplied. In 2018, schema mostly served Google’s rich results. The cost-benefit was concrete: mark up your FAQ and you might get an FAQ rich result in the SERP. Today the same markup is read by half a dozen different systems, each of which uses it slightly differently. ChatGPT browse will use the Organization markup to decide whether to identify you by your trading name or your legal name. Perplexity will use the Article markup, including the author reference, to decide whether to attribute a quotation. Google AI Overviews will use the AggregateRating and Review markup to decide whether to include you in a comparison answer at all.
In practice the effort is the same as it was, and the payoff is broader.
What I actually mark up
The list I recommend to every consulting client is shorter than they expect. In rough order of impact:
Organization, with cross-referenced Person founder. This is the foundation. The Organization gets a stable @id URL such as https://yoursite.com/#organization. The founder is referenced by @id to a Person entity defined on the About page. That link, machine-readable, is what allows engines to know that the company and the human behind it are the same brand. It sounds trivial. The number of sites where it is missing or broken is high.
Article markup on every published article, with author by @id. Same logic. Every article references the single Person entity by stable @id. The result is a knowledge graph where every piece of content can be traced back to its author. Engines reward that.
Service markup for each core service, with provider by @id. On the services pages. The provider field points back to the Organization. This is the bit that lets engines say “this company offers X” with confidence.
BreadcrumbList on every page that has a breadcrumb. Cheap, well-defined, helps engines understand site structure.
FAQPage for genuine FAQ content only. Do not stuff fake FAQs onto product pages to game rich results. Google has been catching this for years, and the AI engines treat over-formatted pages with suspicion.
That is most of it. The rest of the Schema.org vocabulary is genuinely vast, and there are valid uses for HowTo, Event, JobPosting, LocalBusiness, and dozens more depending on what you actually sell. The list above is the foundation that 80% of consulting clients need 80% of the value from.
Common mistakes I see in audits
Invalid JSON. Surprisingly common. The schema validator in Google Search Console catches the obvious ones, but plenty of sites generate their schema with templating that breaks when an editor leaves a quote unclosed in a description field. Test with the actual Rich Results Test tool, not with assumptions.
Schema that contradicts the visible page. If your Article markup says the publication date was 12 March but the visible byline says 22 March, the engines distrust both. Worse, sometimes the schema dates are months ahead of the visible page because of a CMS bug nobody noticed. Run a periodic sample audit and check that the JSON-LD and the rendered HTML agree.
Author by string, not by @id. Most CMS plugins emit "author": "Jane Smith" as a plain string. That is technically valid, and it conveys almost nothing. Engines cannot link Jane Smith on this article to Jane Smith on the next one. Use the Person entity with a stable @id, defined once on the author’s profile page and referenced everywhere else.
Stale or absent canonical URLs in schema. The url field on Article should match the canonical URL of the rendered page. Sites that serve under multiple domains, or have a staging subdomain accidentally indexed, get this wrong constantly.
Markup buried in a single page. The Organization markup should be on the homepage at minimum, ideally on every page. The Person markup should be on the About page. If you only emit Organization on one obscure landing page, engines may never see it.
The cheapest test you can run today
Open your site’s homepage. View source. Find the JSON-LD script. Copy the contents. Paste into the Rich Results Test at search.google.com/test/rich-results. Read what it says. If it is empty, if it errors, if it returns nothing identifiable, you have just found the single highest-leverage technical improvement on your roadmap.
Then do the same for an article page and for your About page. If those three pages emit clean, cross-referenced Organization, Article, and Person schema, you are ahead of more than half the sites I audit.
Final note
Structured data is unfashionable because it is invisible. There is no satisfying screenshot to share at the end of a project. The wins compound slowly and quietly: a slightly higher click-through rate from a better-formatted SERP, a slightly higher chance of being cited correctly in an AI answer, a slightly better internal mental model that the engines hold of your brand.
Quietly compounding wins, on a foundation that costs almost nothing to maintain, are the kind of work I keep coming back to.