The one-page version
For a B2B firm building out AI-search-ready content that wants the technical base right without drowning in markup.
- 1Use JSON-LD only. Skip Microdata and RDFa. One script tag per page, in the head or body.
- 2Organization on every page: legal name, logo, URL, sameAs links to LinkedIn and anywhere else you exist.
- 3Article on every article: headline, author with a real name, datePublished, dateModified.
- 4FAQPage only where the questions are genuinely on the page. Marking up FAQs you have not written is spam.
- 5Validate with Google's Rich Results Test and Schema.org's validator, then stop fiddling with it.
The short answer
A B2B site needs four schema types for AI search: Organization, Article, FAQPage and Service, all in JSON-LD format. Organization tells retrieval systems who you are with total consistency. Article tells them who wrote what and when. FAQPage marks up question-and-answer structure so it can be lifted cleanly. Service states what you sell in plain, machine-readable terms.
Everything else is optional or a distraction. You do not need a dozen schema types, a plugin generating markup you have never read, or markup on pages with nothing worth citing. Schema is the grammar of the page, not the argument.
What schema actually does (and does not do)
When an assistant assembles an answer, it works harder to extract facts from pages it cannot parse cleanly. Schema removes that ambiguity. It tells the machine, explicitly, this is the company name, this is the author, this is when it was published, this is a question and this is its answer. No inference required.
What it does not do is make an unremarkable page quotable. A page that says nothing distinctive, wrapped in perfect markup, is still a page that says nothing distinctive. Schema is table stakes: the absence of it can hurt you, the presence of it will not save you.
Schema gets your facts read correctly. Only your facts get you cited.
"Add schema and AI engines will start recommending you."
No. Markup describes content. It cannot substitute for it.
"Clean schema means nothing about who we are or what we published gets lost in translation."
Yes. It is insurance against being misread, not a ticket to being read.
Organization: who you are, said once
This goes on every page, sitewide, and it should be identical everywhere. It answers the most basic question a retrieval system has about your site: who is speaking? If your company name appears three different ways across your own pages and your LinkedIn profile, you are making the machine guess.
The fields that matter. Anything beyond this is garnish.
@type: Organization (or a tighter subtype like ProfessionalService)
name · your trading name, exactly as you use it everywhere
url · your canonical homepage
logo · a real, crawlable image file
sameAs · LinkedIn company page, plus any other profiles that are genuinely you
description · one sentence, the same one you use off-site
The sameAs field is the quietly important one. It ties your site to your presence elsewhere, which is how corroboration gets connected.
Article: authorship and freshness
Every article should carry Article schema. The fields that matter are the ones that answer "who said this and how current is it", because those are the fields a model uses to decide whether your content is attributable and fresh.
- headline. Matches the visible H1, not a clickbait variant.
- author, as a Person with a name. "Marketing team" is a dead end. A named author with a job title is a citable entity.
- datePublished and dateModified. Keep dateModified honest. Updating it without updating the content is a small lie that gets noticed.
- publisher. References your Organization. This is what connects the article back to who you are.
Want more enquiries from the content you are already publishing?
Book a 30 minute call. We will look at where your leads come from now and what would move the number.
FAQPage: structure models can lift
FAQPage is the highest-value content schema for AI search, because question-and-answer pairs are the exact shape assistants work in. A marked-up FAQ block is a pre-packaged answer a model can lift with confidence.
The rule is simple: only mark up questions and answers that are visibly on the page. The questions should be phrased the way a buyer would actually ask them, and each answer should stand alone if lifted whole, because that is precisely what will happen to it.
- Write the question as a buyer would type it. "How much does LinkedIn advertising cost for a small consultancy" beats "Pricing information".
- Answer in two to four sentences. Direct answer first, qualification second. Long FAQ answers get truncated or skipped.
- Keep answers stable. A FAQ answer that changes every week is not a reliable thing to cite.
- Do not fake it. FAQ markup on questions that do not appear on the page is spam, and it is the fastest way to get your markup ignored entirely.
Service: what you sell, in plain terms
Service schema on your service pages states what you offer, who provides it and the area it covers. For niche B2B firms this matters because the difference between "marketing services" and "LinkedIn advertising management for UK manufacturers" is the difference between being retrievable for everything and being cited for something.
One per service page, matching what the page actually says.
name · the service in plain English, not the brand name for it
provider · your Organization
description · what it is and who it is for, in one or two sentences
areaServed · where you actually operate
serviceType · the category a buyer would search for
The description field should match the plain-language positioning you use everywhere else. Consistency across schema, page copy and off-site profiles is what builds a clear association.
What you can safely skip
Just as important is what not to bother with. None of these are worth the implementation time for a typical niche B2B firm:
- Product and Offer schema unless you sell something with a price on the page. Most B2B services do not, and forcing it reads as spam.
- Review and AggregateRating markup you have generated yourself. Self-serving reviews in schema are against Google's guidelines and erode trust everywhere else.
- HowTo schema for content that is not a genuine step-by-step process with real steps on the page.
- Every plugin-generated extra. SEO plugins love emitting markup you never asked for. Read what yours outputs and strip what you do not need.
Validate it, then leave it alone
Schema is set-and-forget infrastructure, not an ongoing optimisation channel. Implement it, validate it, fix what the validators flag, and then move on to the work that actually earns citations.
- Step 1Implement in JSON-LD
One script tag per page. Organization sitewide, Article on articles, FAQPage where real FAQs exist, Service on service pages. If your CMS or framework can generate it from the same data that renders the page, even better.
Output: Markup live on the templates, not just one test page. - Step 2Run the Rich Results Test
Google's Rich Results Test on a sample of each template type: homepage, one article, one service page, one FAQ block. Fix every error. Warnings on optional fields you deliberately omitted are fine.
Output: Zero errors on each template type. - Step 3Cross-check the Schema.org validator
The Rich Results Test only covers the types Google cares about. The Schema.org validator checks the rest of your markup for syntax problems that could trip up other parsers.
Output: Clean parse on both validators. - Step 4Add a review to your content process
When a template changes or a new content type ships, re-validate that template. Otherwise forget it exists and spend the time on content with actual substance.
Output: Schema re-checks as part of template changes, nothing more.
Where this goes wrong
- Treating schema as the strategy. It is the plumbing. The strategy is what questions you answer and what evidence you publish. Confusing the two is how firms spend a month on markup and wonder why nothing changed.
- Letting a plugin speak for you. Auto-generated markup often contradicts the visible page or pads fields with generic text. Read your own output once and you will usually find something to fix.
- Inconsistent identity. Company name in schema that differs from the footer, the LinkedIn page and the trade directory. Pick one version and use it absolutely everywhere.
- Gaming dateModified. Bumping the date without changing the content to look fresh. Retrieval systems compare; the ones that notice stop trusting the date.
- Endless tweaking. Once it validates, further fiddling has roughly zero return. The marginal hour is better spent on one more genuinely quotable paragraph.
Get the grammar right, then go back to having something worth saying.
Rather have someone build the pipeline while you run the business?
Book a 30 minute call. We will talk through where your leads come from now and what would move the number.
Read next
Common questions
If you want content and campaigns working together to bring in leads
- LinkedIn, Meta and Google campaigns run against a named target account list.
- Content and service pages built around the questions your buyers actually ask.
- Tracking and attribution wired up so you can see what is really driving enquiries.
- Creative built from your own material, shipped and tested every month.
- Reporting on cost per qualified pipeline rather than cost per click.
Starter £1,250/mo, Growth £1,725/mo, Scale £4,025+/mo. Content work sits inside the retainer where it is in scope. Anything outside that is quoted properly before it starts.