Structured data for AI search: what to prioritise

Schema markup is one of the older bits of technical SEO, and for a long time the marginal value of adding more of it had been falling. AI search has changed that. We’re now seeing structured data play a real role in retrieval, disambiguation and citation, and the question for tech marketers in 2026 is no longer “should we add schema” but “which schema earns its place”.

This piece walks through the schema.org types we’d prioritise for B2B tech sites, the ones we’d skip and why. It’s grounded in what we’ve seen on real client sites rather than what the documentation suggests.

Why schema matters more in an AI search era

Three reasons we’d put forward.

First, disambiguation. LLMs need to know whether your page about “Aspire” is about Aspire Technology Solutions, Aspire the children’s charity or Aspire the e-cigarette brand. Organization and Person schema with sameAs references to LinkedIn, Wikipedia and Crunchbase is one of the cleanest ways to say “this is the entity we’re talking about”.

Second, retrieval. Google AI Overviews leans heavily on structured data for several query types. Pages with the right schema get pulled into AI Overviews more often, in our testing. Whether ChatGPT or Perplexity weight schema directly is murkier, but well-marked-up pages tend to be cleaner pages overall, which feeds into retrieval indirectly.

Third, citation. The fields you mark up (author, datePublished, headline, about) often map directly to what an LLM needs to attribute a quote. A page with Article schema with a named author is more cite-friendly than the same page without.

We’ve covered this from a more general angle in schema markup for SaaS websites and the tracking implications in tracking AI search traffic.

The schema we’d prioritise

Across most B2B tech sites, six types do the heavy lifting:

Organization

Top of the list. This describes who you are as an entity. Name, URL, logo, foundingDate, sameAs links to social profiles, Wikipedia entry if you have one, Crunchbase, Companies House. For tech firms with a UK presence, including the Companies House number under identifier is a small but useful disambiguation signal.

If you operate under multiple brands, mark up the parent organisation and any sub-brands with their own Organization entries. We’ve seen this resolve cases where an LLM was confusing one client’s two product lines.

Article

For every blog post and editorial page. At minimum: headline, datePublished, dateModified, author (as a Person, not a string), publisher (as your Organization), about (the topic), wordCount and articleSection.

The author field is the one most often skipped. Treat your authors as first-class entities. Each named contributor should have a Person schema with a description, jobTitle and sameAs links to LinkedIn and any other public profiles. We’ve seen Perplexity in particular weight authored content more heavily than unattributed posts.

Person

For your authors and named experts. This pairs with Article schema and improves how LLMs attribute quotes. Include jobTitle, worksFor (your Organization), description and sameAs.

Product or Service

For SaaS and service-led tech firms, mark up your products and services. SoftwareApplication for SaaS, Service for managed services. Include name, description, applicationCategory or serviceType, provider, areaServed and offers where pricing is public. Where reviews come from third-party platforms, our piece on G2 and Capterra in AI search covers how those mentions feed into citation alongside schema.

For MSPs in particular, marking up each named service offering helps Google AI Overviews surface your firm for relevant queries. We’ve covered this from a different angle in Google AI Overviews and MSP citation.

FAQPage

Useful but specific. FAQPage schema has fallen out of favour for some Google rich result types, but it still helps AI search retrieval for question-shaped queries. Use it where the questions are genuinely buyer questions, not where you’ve invented questions to game the markup.

BreadcrumbList

The unglamorous one. Helps with retrieval by giving the model a clear sense of where the page sits in the site’s hierarchy. Cheap to add.

Schema we’d be more cautious about

A few types we see clients consider that we’d be more measured about:

• HowTo. Useful where you genuinely have a step-by-step process. Often misused on listicles. Google has tightened how it treats this, and adding it to non-process content can look spammy.
• Review and AggregateRating. Only mark up reviews you actually display on the page. Markup for reviews that are not visible to users gets penalised, and it’s an easy way to lose trust.
• VideoObject. Worth adding if video is core to your content strategy and the videos are hosted in a way the markup can describe. Not worth retrofitting onto every page that happens to embed a YouTube clip.
• Event. Only if you actually run events. Webinars count, ad-hoc lunch-and-learns probably do not.

How this fits with the rest of your AI SEO work

Schema is plumbing. It supports the writing, not the other way round. The pages that benefit most from rich schema are the ones whose content has been written to be cited in the first place. We’ve covered the writing side in writing content that AI search engines actually cite. Without that foundation, schema alone will not move the needle.

It also pairs with the broader entity work we covered in how LLMs cite sources. LLMs are essentially trying to build a coherent picture of your business from a mix of your site, structured data, third-party mentions and links. Schema is one of the inputs the model can read most cleanly. The others matter too.

Implementation in practice

A practical sequence we use with B2B tech clients:

1. Audit current schema. Tools like Schema Markup Validator and the Search Console Rich Results report show what’s there and what’s broken. We’ve found broken schema is more common than missing schema.
2. Map entities. Decide which entities need a schema record. The Organization, the named senior people, the products or services, the major content hubs.
3. Implement Organization and Person first. These are the highest-yield disambiguation entries. Most CMS setups can render them in the site-wide footer or head.
4. Add Article schema to all editorial. Including authored bylines and dateModified that updates when the page is genuinely revised.
5. Mark up products and services. With honest pricing where you publish it, region targeting where it applies and explicit provider links to your Organization.
6. Add FAQPage and BreadcrumbList where they make sense. Skip them where they don’t.
7. Validate and monitor. Run Search Console’s Rich Results report monthly. Watch for schema errors after deploys.

This is plumbing work, but it’s worth doing once and properly. We typically run it as part of an AI SEO engagement alongside content rewrites, citation auditing and the technical foundation work.

A note on JSON-LD versus microdata

Use JSON-LD. Microdata still works but is harder to maintain, and most modern CMS or static site setups make JSON-LD trivial to template. For Astro sites, schema lives naturally in the layout component or in a per-page slot. Whichever you use, keep one canonical source of truth so a single edit propagates. The same discipline applies if you publish API references or technical documentation, which we cover in designing developer docs.

If you’re trying to get visible in LLM answers and not sure where to start, tell us about your business. Schema is rarely the most exciting part of the brief, but on most of the sites we audit it’s the bit that’s been left undone.