Schema Markup That Actually Moves the Needle

01What schema markup is — and what it actually does

Schema markup is structured data: a standardised vocabulary (from Schema.org) that you add to your pages, usually as JSON-LD, to describe what your content means in terms machines can read without guessing. Instead of leaving a search engine to infer that "4.8" is a rating, that "£1,200" is a price, or that "Platinum" is an organisation, you state it explicitly. That is the whole job: removing ambiguity.

Here's the part most guides get wrong. Schema markup is not a ranking factor. Google has said this plainly and repeatedly. Adding structured data will not, on its own, push you up the results. What it does is make your pages eligible for rich results — the enhanced listings with star ratings, prices, event dates and breadcrumbs — and help search and AI systems understand the entities and facts on your page. Those are different things, and conflating them is why so much schema effort gets wasted on markup that can never produce a visible result.

In 2026 there's a second job that matters just as much. Answer engines (the boxes that give a direct answer — Google's AI Overviews, voice assistants) and generative engines (ChatGPT, Gemini, Perplexity and friends) increasingly use structured data to verify claims, establish that two things are related, and assess whether a source is credible enough to cite. Schema doesn't write the answer for them, but it gives them clean, attributable facts to lean on. That's the AEO and GEO payoff, and it's why this is worth doing properly rather than not at all.

02Why most schema markup is dead weight

Walk into most websites' code and you'll find schema sprayed everywhere — half-implemented, missing required properties, or describing things the page doesn't actually show. None of it does anything. Google is explicit on two points that kill the majority of badly-done markup. First: structured data must describe content that is visible to the user. "Don't add structured data about information that is not visible to the user, even if the information is accurate." Second: accuracy beats completeness. In Google's words, it's "more important to supply fewer but complete and accurate recommended properties rather than trying to provide every possible recommended property with less complete, badly-formed, or inaccurate data."

The other reason markup is dead weight is eligibility. A schema type only produces a rich result if (a) Google supports a rich result for it, and (b) you've supplied every required property. Partial implementation gives you zero lift — there is no half-credit. Add Product schema without a price or availability and you get nothing. So the goal isn't "more schema." It's the right types, implemented completely, matched to what's genuinely on the page.

This matters more in 2026 because Google has been actively trimming the rich-result surface. HowTo rich results were retired back in 2023. FAQ rich results were dropped on 7 May 2026. Several review and merchant-related features have been tightened. If your schema strategy was built around features that no longer render, you're maintaining code for results that will never appear. Knowing which types still pay out is the whole game.

03The schema types that actually move the needle in 2026

Here are the structured data types worth your time, roughly in order of how broadly useful they are. Implement the ones you're genuinely eligible for, completely, and ignore the rest.

1. Organization — the most underrated type in the AI era. It defines your business as an entity: name, logo, URL, contact points and (crucially) sameAs links to your verified social and Wikidata/Wikipedia profiles. This is what feeds the Knowledge Graph and what generative engines lean on when deciding whether you're a real, citable entity. If you do nothing else, do this on your homepage.
2. Product — for ecommerce and any priced offer. Drives the price, availability and review-star treatment in results. Required properties (name, offers, price, availability) must be present and must match the page. Still one of the highest-value rich results there is.
3. Review and AggregateRating — the star ratings. Powerful for trust and click-through, but tightly policed: it must reflect genuine reviews shown on the page, for the right entity, and self-serving reviews of your own business are restricted. Use it honestly or not at all.
4. Article (and NewsArticle / BlogPosting) — signals content type, headline, author and publish date. It won't always show a visible rich result, but it clarifies authorship and freshness, which supports E-E-A-T signals and helps AI engines attribute claims to a named author and date.
5. LocalBusiness — for anyone with a physical or service-area presence. Pins down name, address, opening hours and geographic relevance. Directly improves how AI systems judge local relevance and is foundational for local pack and map visibility.
6. Breadcrumb — modest but reliable. Produces the breadcrumb trail in the listing and helps engines understand your site hierarchy. Low effort, consistently rendered.
7. Event — for anything time-and-place bound: webinars, launches, workshops. Surfaces dates and locations directly in results and is well-supported.
8. JobPosting — if you hire, this feeds Google's jobs experience with role, salary and location. High-intent and still strongly supported.
9. Recipe, Video, Course, SoftwareApplication, Dataset — niche but valuable if they fit your content. Each unlocks a specific rich treatment; use the matching one only when the content genuinely is a recipe, video, course and so on.

Notice what's not on the list as a rich-result play: FAQPage and HowTo. They had their moment; that moment has passed for visible results. More on FAQ below, because there's nuance.

04FAQ schema after the May 2026 deprecation

On 7 May 2026, Google stopped showing FAQ rich results in Search. The FAQ search-appearance filter, the rich-result report and Rich Results Test support are being removed in June 2026, with Search Console API support ending in August 2026. This completes a deprecation that began in 2023, when Google first restricted FAQ rich results to government and health sites. Importantly, this is a search-appearance change, not an algorithmic one — rankings aren't being adjusted because of it. Pages that relied on big FAQ dropdowns for click-through may simply see those listings shrink back to normal.

So should you rip out your FAQ schema? No need to rush. FAQPage is still a valid Schema.org type and won't harm your page. But be clear-eyed about why it's there. It will no longer earn you SERP real estate. Its only remaining value is as a small machine-readable signal that may help AI systems parse a clean question-and-answer pair — and even that is unproven, not something Google has confirmed. The real lesson is bigger: don't build schema strategy around features Google can withdraw. Keep genuine, concise FAQ content on the page because it's useful to humans and easy for answer engines to lift. Drop the assumption that the markup itself buys you anything visible.

05Schema for AEO and GEO: getting cited, not just decorated

There's an honest debate worth surfacing. Some practitioners argue technical schema barely matters for AI citations, because large language models read natural-language content directly and Google has said LLMs don't strictly require structured data. That's partly true and a useful corrective to schema hype. But it's not the whole picture. Structured data still does something content alone can't: it states facts unambiguously and links entities together, which is exactly what an AI system needs when it's deciding whether a claim is trustworthy enough to repeat with your name attached.

The practical takeaway for AEO and GEO is to treat schema as a reinforcement layer, not the foundation. The foundation is clear, factual, well-structured writing — direct answers near the top, specific claims, named entities, comparisons and lists. Schema then confirms in machine terms what the prose already says: that you are this organisation, that this is the author, that this product costs that, that this rating is real. The types that pull double duty here are Organization (entity identity), Article (authorship and date), Product and Review (verifiable facts), and LocalBusiness (geographic context). Get those right and you've covered both the visible-rich-result job and the AI-verification job with the same effort.

One more GEO-specific move: invest in sameAs and consistent entity data across your site and the wider web. When your Organization schema, your Wikidata entry, your LinkedIn and your About page all agree on the same facts, you become a stable, recognisable entity — and stable entities are the ones AI engines cite with confidence. Inconsistency is what makes a model hedge or pick a competitor instead.

06How to implement schema without wasting effort

A tight, no-fluff workflow that avoids the common traps:

• Use JSON-LD. Google explicitly recommends it as the easiest format to implement and maintain at scale and the least error-prone. It lives in a single script tag, separate from your HTML.
• Pick types you're genuinely eligible for. Match the schema to what the page actually is and shows. No invisible-data markup, ever.
• Fill every required property, completely. Partial markup earns nothing. Add recommended properties too where they're accurate — they increase the chance of enhanced display.
• Validate before and after launch. Run every type through Google's Rich Results Test (for rich-result eligibility) and the Schema.org Markup Validator (for general validity).
• Monitor in Search Console. Watch the Enhancements and rich-result reports for live errors and impressions. Re-audit after any template, theme or CMS change — that's when markup silently breaks.
• Keep it current. Google withdraws and changes rich-result support regularly. Review your schema types against the current Search Gallery at least twice a year, and don't cling to deprecated features.

Do this and your structured data stops being decorative code and starts being an asset: more rich results where they still render, cleaner entity recognition, and a better chance of being the source an AI answer engine names. That's schema markup that actually moves the needle — a small, precise set of types done properly, not a sprawl of markup done for its own sake.