Yes, AI engines like ChatGPT, Perplexity, Gemini and Google's AI Overviews need to understand a page quickly and confidently before they'll cite it. Schema gives them that certainty, especially on pages where the visible content is harder to parse and adds context, lightweight indexing and relevance to your content.
Do I need all 10 types below? No — and that's really the point of this article. The right schema for a page depends on what that page is for. A service page, a job posting and a blog article all need different markup.
What should I fix before schema? Your basic technical inventory. If Google Search Console is full of 404s, unresolved redirects and canonical conflicts, no amount of schema will make AI engines trust the page underneath it.
Most conversations about GEO (Generative Engine Optimisation) jump straight to content — fact-density, quotable statements, answering the right questions. All correct. But underneath the content sits something less glamorous and, frankly, more decisive: whether your site's code tells AI engines what they're actually looking at.
That's schema markup's job. And it's one of the few genuinely technical levers you have real control over in GEO.
Schema markup is structured data — almost always written as JSON-LD today — that sits in a page's code and describes its content using a shared vocabulary from schema.org. It doesn't change how a page looks. It changes how confidently a machine can describe what's on it.
Here's why that distinction matters more in 2026 than it did five years ago. Traditional search crawlers were reasonably good at rendering a page and inferring meaning from headings, layout and text. AI crawlers and retrieval systems work differently — many rely on speed, on partial rendering, or on content that's been fetched without executing every script. If your page relies on JavaScript-rendered content, dynamic personalisation, PDFs, embedded tools, or components that load after initial paint, an AI engine may only ever see a fraction of what a human visitor sees.
Schema closes that gap. A well-implemented Organization, Service or Article schema block is a plain, structured summary of the page that exists independently of how (or whether) the visual content renders. It's the difference between an AI engine guessing at your business from scattered text, and reading a clean, unambiguous statement of who you are, what you do, and where you do it.
Worth being precise here too: Google's own current guidance on AI Overviews and AI Mode says there's no special schema required to be featured in them — but it also says structured data should always match what's visibly on the page. Schema isn't a shortcut around having good content. It's a translation layer that makes good content easier to trust.
Where schema tends to fall down isn't the writing of it — it's the maintenance of it. Markup gets added once, at launch, hard-coded into a template, and then drifts silently out of sync as the visible content changes. A price changes. A service is renamed. Nobody touches the JSON-LD. That mismatch between what's marked up and what's actually on the page is one of the most common structured data problems we come across in the audits we run as part of client maintenance and support work — and it's also one of the more damaging ones, because inconsistent schema undermines the very trust it's supposed to build.
The fix isn't cleverer code — it's better ownership. At NXT, we build schema markup directly into Umbraco content blocks rather than hard-coding it into templates. That means when a content editor updates a service description, an FAQ answer, or an address, the structured data updates with it automatically, without a developer needing to touch the underlying page. Schema becomes part of how a page is edited, not a separate technical layer bolted on afterwards.
That's also why schema needs to sit inside routine site maintenance, not a one-off launch task. New page templates get built. Old ones get retired. Content changes constantly. An audit that treats schema as "done" after go-live is an audit that's already out of date.
If an AI engine can't read your organisation's name, location and services with confidence straight from the code, it's not going to cite you with confidence. Getting the schema foundations right needs to be part of ongoing site maintenance, not a design-phase afterthought.
Peter Hinchliffe Strategy Director
It's tempting to treat schema as the exciting part of a GEO project and the technical inventory as background admin. In practice, the inventory work has to come first.
If Google Search Console is showing a long tail of unresolved warnings — 404s that were never redirected, legacy temporary redirects that were never made permanent, "Google chose different canonical than user" conflicts caused by www/non-www or trailing-slash inconsistencies, pages sitting in "Crawled – currently not indexed" — then AI crawlers are inheriting the same confusion Google is. A page can carry flawless Organisation and Service schema and still be functionally invisible if the crawler can't reliably settle on which URL is the canonical, trustworthy version of it.
Practically, that means before (or alongside) any schema rollout:
Clear the legacy 404s and convert any remaining temporary (302) redirects to permanent (301) ones
Resolve canonical conflicts — usually a www/non-www or trailing-slash inconsistency once you dig into it
Review pages marked "Crawled – currently not indexed," which is often a sign of thin or templated content rather than a technical fault
Re-check this inventory periodically, not just once — canonical issues in particular have a habit of quietly reappearing after site changes
A clean inventory is what makes schema worth doing. Skip it, and you're building a very precise signpost pointing at a locked door.
You'll notice this article opened with a short "quick answers" block before a word of narrative content. That's deliberate, and it's becoming close to a requirement rather than a stylistic choice.
AI engines and time-poor readers want the same thing: a direct, fact-dense answer near the top of the page, before the supporting explanation. A short GEO summary — three or four sentences or bullet points answering the most likely questions about the page — gives AI systems something clean and quotable to extract, and gives human readers who arrived via an AI answer the confirmation they're in the right place.
It's not a replacement for good structured data. It's the human-readable counterpart to it — the same clarity schema provides in code, expressed in plain sentences a person (or a language model) can lift directly.
These are the ten we'd put in front of most clients first, roughly in the order they tend to matter. Treat this as a strong starting point, not a mandate — more on that below.
The baseline. Name, logo, official URL, sameAs links to verified social profiles, and contact details. This is the entity record an AI engine builds everything else around — get it wrong or leave it incomplete, and every other schema type on the site inherits the ambiguity.
For any business with a physical presence or a regional service area, Name, Address and Phone (NAP) data belongs in structured data, not just the footer. A common gap we see: full NAP details displayed on the page but never mirrored into the Organization or LocalBusiness schema — and areaServed set to something generic like "Worldwide" on a site that's actually built around strong regional positioning. If local trust matters to your business, your schema needs to say so as clearly as your homepage does.
Establishes the site and the individual page as distinct entities, and is the schema type that enables features like a sitelinks search box. Foundational, low-effort, easy to get wrong through neglect rather than difficulty.
Breadcrumb schema describes a page's position in the site hierarchy — Home > Digital Services > Umbraco CMS, for example. It's a small piece of markup that does a disproportionate amount of work for both traditional crawlers and AI retrieval, because it tells a machine exactly how a page relates to the rest of the site without needing to infer it from navigation menus.
Every distinct service line deserves its own Service schema — specific to what that service actually does, not a templated paragraph duplicated with minor wording changes across five other service pages. Duplicated, boilerplate service descriptions are one of the more common issues we find during structured data audits, and they're exactly the kind of thin, generic content AI engines are least likely to cite.
For blog and insights content: headline, author, publish date, and — critically for GEO — a modified date that's actually kept current. AI systems weight recency and authorship as trust signals; an Article schema with a three-year-old date and no author attribution undermines a genuinely current piece of content.
Worth a specific note here, because the picture changed materially in 2026. Google added a deprecation notice to its FAQ structured data documentation on 7 May 2026, confirming that FAQ rich results — the expandable Q&A dropdown that used to appear under search listings — no longer display in Google Search, with reporting support in Search Console and the Rich Results Test being withdrawn through June and August 2026. In practice this closes a door that had already been mostly shut since August 2023, when eligibility was restricted to authoritative government and health sites.
What hasn't changed: FAQPage remains a valid schema.org type, Google's own documentation confirms unused structured data doesn't cause problems, and clean question-and-answer content continues to be parsed by Bing, Perplexity, voice assistants, and the various retrieval systems AI engines draw on. The lesson isn't "stop using FAQs" — it's stop treating FAQPage schema as a SERP decoration and start treating direct, visible Q&A content as a genuine comprehension aid. That's also consistent with the wider shift we're seeing across client sites: moving pages away from marketing-style copy and toward direct, informational answers to the questions people actually arrive with — which is exactly the approach behind the quick-answers block at the top of this page.
Genuine, verifiable third-party reviews — Clutch, Google Business Profile, Trustpilot — marked up as AggregateRating. This is one of the more powerful trust signals available to a services business, provided the source is independently verifiable rather than self-collected testimonials dressed up as structured data.
If you're hiring, JobPosting schema is close to essential — it feeds Google for Jobs and gives AI engines a structured, dateable record of an open role, rather than a paragraph they have to parse from a careers page.
Not markup in the strict sense, but it belongs on this list because it does a closely related job. The Open Graph protocol lets any web page become a properly-formed object when it's shared — on Facebook, LinkedIn, X, or pulled into a preview card anywhere else. Title, description, image and type (og:title, og:description, og:image, og:type) are the baseline four; get them wrong — a typo, a broken image path, a mismatched title — and it's often the most visible structured data error on the entire site, because it shows up the moment someone shares a link. It's a quick audit item and a disproportionately embarrassing one to get wrong.
Here's the honest caveat: the ten above are a strong general-purpose starting point for a B2B services business, but schema includes well over a hundred types, and the right list is genuinely content-by-content.
A handful of examples of where the list above simply doesn't apply:
An e-commerce site needs Product, Merchant listing and Product variant schema far more urgently than JobPosting
A course provider needs Course info, Course list and Education Q&A
A recipe site lives or dies by Recipe schema
A dealership needs Vehicle listing; a letting agency needs Vacation rental
A media-heavy site benefits from Video, Speakable (which flags sections suitable for voice/audio reading aloud) and Carousel
A publisher dealing in claims and statistics may need Fact Check
The right question for any given page template isn't "which items from a generic top-10 list are we missing?" It's "what is this page actually trying to be, and does the code say so as clearly as the content does?" That's a template-by-template audit question, not a checklist.
This is exactly the work we build into ongoing client maintenance and support, rather than treating it as a one-off launch exercise: auditing existing schema against what Google Search Console and structured data testing tools actually report, checking implementations against current best practice (which, as the FAQ example above shows, can shift meaningfully within a single year), and building new markup directly into editable Umbraco content blocks so it stays accurate as content changes.
If your last schema implementation was done at launch and hasn't been looked at since, it's worth an audit — particularly if you've added new page templates, service lines, or content types since.
Our Development and Technology team handles schema implementation and technical GEO audits, and it's included as standard within our Digital Support Services maintenance programmes. If you'd like a clear picture of where your site currently stands, get in touch — we're happy to talk through what we'd prioritise for your specific site.
GEO is the practice of structuring and writing content so that AI-powered search tools — ChatGPT, Perplexity, Gemini, Google's AI Overviews — can understand, trust and cite it. It complements traditional SEO rather than replacing it.
Not directly — Google has been consistent that structured data primarily aids understanding and eligibility for certain search features, rather than functioning as a direct ranking signal. Its value for GEO is different: it's about giving AI systems a confident, unambiguous read of your content.
The visible SERP dropdown is gone as of May 2026, but FAQPage remains valid schema.org markup, and clear question-and-answer content continues to be used by AI retrieval systems and other search engines. Keep it where it genuinely reflects visible content on the page; don't add it purely to chase a rich result that no longer exists.
There's no fixed number. It depends entirely on what each page template is trying to do. A homepage, a service page, a job listing and a blog article all warrant different schema — the aim is accuracy per template, not volume.
Schema.org (JSON-LD) is aimed primarily at search engines and AI systems trying to understand a page's content. Open Graph is aimed at how a page appears when shared on social platforms and in link previews. They serve different audiences but both deserve equal attention in an audit.
At minimum, whenever a page template changes, a new content type is introduced, or annually as part of a wider technical audit — structured data guidance itself changes (as the 2026 FAQ update shows), and markup that isn't reviewed tends to drift out of sync with the page it describes.
The NXT Digital Solutions team provides ongoing technical SEO and GEO audits as part of our website maintenance and support programmes, alongside our .NET, Umbraco CMS and AI development services. Learn more at nxtds.com.
16 Jun, 2026
Last week, the NXT team traded our daily stand-ups for the vibrant energy of Copenhagen! We spent four incredible days immersed in the heart of the Umbraco universe at Codegarden 2026 and we've outlined our event and product highlights.
12 Jun, 2026
NXT Digital Solutions has won Best Innovation with AI at the Umbraco Awards 2026, announced live at Codegarden in Copenhagen on Wednesday 10 June 2026 for our enterprise Umbraco rebuild of Exclusive Networks.
07 Apr, 2026
NXT Digital Solutions is officially shortlisted for 'Best Innovation with AI' at the Umbraco Awards 2026 and it's all thanks to our work with Exclusive Networks.
