AI Search

The minimum schema setup for a startup that wants AI citations

Most schema advice is over-engineered. Here are the three types that actually move the needle for AI visibility, what to skip, and how to implement in under two hours.

30 May 2026 · 8 min read · by Ridho Putradi S'Gara

Most schema advice drowns you in options. Schema.org defines over 800 types. Vendors will sell you "complete AI schema packages" covering everything from recipes to medical conditions. The truth is simpler and cheaper: if you're a startup trying to get cited by ChatGPT, Perplexity, or Google AI Overviews, you need three schema types, not three hundred.

I'm writing this because the last four audits I ran all had the same problem. Founders either skipped schema entirely (thinking Google said it doesn't matter) or implemented ten types badly because an agency convinced them completeness equals results. Both paths waste time.

Here's what actually works.

Why schema matters (even though Google says it doesn't)

Google's official documentation states clearly: "There's no special schema.org structured data that you need to add" for AI Overviews or AI Mode. That's true. There is no magic AIPage schema type that unlocks citations.

But here's the operational reality: schema helps AI engines do three things faster and with more confidence.

First, it disambiguates entities. Organization schema with sameAs links to your LinkedIn, Wikipedia, or Crunchbase profile helps Google (and Perplexity, and ChatGPT) resolve that your brand is a known entity, not just a string of text. Resolved entities get higher trust scores.

Second, it signals authorship and freshness. Article schema with author, datePublished, and dateModified makes it unambiguous who wrote the page and when. That matters for YMYL (health, finance, legal) topics where AI engines weight expertise heavily.

Third, it packages answers in machine-readable units. FAQ schema doesn't guarantee a citation, but when an AI engine encounters a user question that matches your FAQ question, it can extract your answer directly without parsing paragraphs. That reduces the computational cost of retrieval, which matters when the engine is deciding between your page and a competitor's.

Schema is a tie-breaker, not a magic switch. Content quality, topical authority, and link equity still dominate. But when two pages are roughly equivalent, the one with clean schema wins.

The three schema types that actually matter

If you're a B2B SaaS startup, a DTC e-commerce brand, or a service business in Southeast Asia, this is your minimum viable schema stack.

1. Organization schema

This goes on your homepage and every page via your site-wide template. It tells AI engines who you are, where you exist online, and what you do.

Required properties:

Optional but useful:

Why it matters: this is your entity anchor. When ChatGPT or Perplexity tries to figure out if "Acme" in your content is the same "Acme" mentioned elsewhere, sameAs links give them the answer. Without this, you're a floating string of text, not a known entity.

2. Article schema

This goes on every blog post, case study, guide, or resource page. It marks the content as editorial, not commercial, and surfaces author and date signals that AI engines use to judge credibility.

Required properties:

Optional but useful:

Why it matters: AI engines rely on author signals to filter low-quality content farms from credible sources. If there's no Article schema, they have to infer authorship from bylines and page structure. That works sometimes. Schema removes the guesswork.

I've seen this firsthand. A maternal nutrition brand I worked with saw a 3.7x increase in AI Overview keywords after we added Article schema to 140 pillar posts. The content didn't change. The machine-readability did.

3. FAQ schema (when you actually have FAQs)

This is conditional. If your page genuinely answers a set of related questions (a product FAQ, a "how does X work" guide, a troubleshooting doc), FAQ schema is the highest-leverage type you can add.

Required properties:

Why it matters: when an AI engine sees a user query like "how do I connect Stripe to Shopify," and your FAQ schema has a question with that exact phrasing, the engine can pull your answer verbatim. I'm not saying it will. I'm saying the probability goes up.

The mistake I see: founders add FAQ schema to pages that don't have FAQs. They reformat a paragraph into a fake Q&A just to get the markup. Don't. Google can issue a manual action for mismatched structured data, and worse, the AI engine might cite the wrong information.

Only use FAQ schema if the visible page actually shows questions and answers.

What to skip (and why vendors won't tell you this)

Here's what you don't need if you're a startup with fewer than 50 pages and a two-person marketing team.

BreadcrumbList schema. Nice to have if you're an e-commerce site with deep category hierarchies. Overkill if you have a flat site structure. Skip it until you have 200+ pages.

Product schema. Required if you sell physical goods and want rich snippets (price, availability, reviews) in classic search. Not required for AI citations unless you're in the comparison-shopping vertical. If you're B2B SaaS, you don't need this.

HowTo schema. Only useful if your page is a literal step-by-step procedure (install a plugin, cook a recipe, assemble furniture). Most "how-to" content doesn't qualify. A blog post titled "How to pick a CRM" is not a HowTo schema candidate. A support doc titled "How to reset your password" is.

VideoObject, Event, Course, Review. Vertical-specific. If you run webinars, mark them up with Event schema. If you publish course content, use Course schema. If you don't, skip them.

Speakable schema. This marks content for voice assistants to read aloud. Theoretically useful. Practically untested. I haven't seen evidence it moves the needle for B2B startups. Save it for later.

The rule: implement schema that describes what your page actually is. Don't add schema to game an algorithm. It won't work, and Google's manual review team can penalize you for it.

How to implement this in under two hours

You don't need a developer for this. You need a text editor, a JSON-LD generator, and access to your CMS.

Step 1: Use a schema markup generator or write the JSON-LD by hand. Here's a starter template for Organization schema:

```json { "@context": "https://schema.org", "@type": "Organization", "name": "Your Company Name", "url": "https://yoursite.com", "logo": "https://yoursite.com/logo.png", "sameAs": [ "https://www.linkedin.com/company/yourcompany", "https://twitter.com/yourcompany", "https://www.crunchbase.com/organization/yourcompany" ] } ```

Step 2: Add the JSON-LD to your site's `<head>` or just before `</body>`. If you're on WordPress, use a plugin like Yoast SEO or RankMath. If you're on Webflow, Framer, or a headless CMS, paste the script tag into your global site settings.

Step 3: Validate it using Google's Rich Results Test or the Schema Markup Validator. Fix any errors (usually a missing required field or a broken URL).

Step 4: Repeat for Article schema on your blog template. Most CMS platforms let you set dynamic fields (pull the title, author, and publish date automatically). If you're on WordPress, Yoast and RankMath do this out of the box.

Step 5: For FAQ schema, hand-code it for your top 5-10 FAQ or support pages. Don't template it unless every page genuinely has FAQs.

Total time: 90 minutes if you've never touched schema before. 30 minutes if you have.

How to check if it's working

Schema doesn't show immediate results the way a title tag change might. You're playing a longer game.

Short-term check (1-4 weeks):

Medium-term check (2-3 months):

Long-term check (6+ months):

The metric that matters is not "did I implement schema" but "did it increase my citation rate for the queries that drive pipeline." If you're not tracking citations, you're guessing.

The one thing to remember

Schema is infrastructure, not a campaign. You implement it once, then forget about it until you publish new content types or Google updates their guidelines.

The three types I've listed here (Organization, Article, FAQ) cover 90% of what a startup needs to be citation-ready. If you're spending more than two hours on your first pass, you're over-engineering it.

If you want a second set of eyes on your schema setup, or you're not sure whether your pages qualify for FAQ or HowTo markup, book a 30-minute strategy call. I'll audit your top 10 pages and tell you what's missing, what's wrong, and what to ignore.

And if your team wants to learn how to do this in-house without hiring an agency every time Google ships a new AI feature, the training program walks through schema implementation, citation tracking, and the rest of the AI search stack in a format that works for two-person marketing teams.

The goal isn't perfect schema. The goal is good-enough schema that doesn't slow you down and gives AI engines one less reason to skip your content.

Related reading

30 May 2026 · AI Search

Building topic clusters when you only have one writer

Most topic cluster advice assumes a content team. Here's how to build a focused, citation-friendly structure when you have one writer and three months.

Read post
30 May 2026 · AI Search

How to pick the 20 queries that decide your AI search strategy

Most startups track too many queries or the wrong ones. Here's how to identify the 20 that actually determine whether you get cited in AI answers.

Read post
29 May 2026 · AI Search

How to write a content brief that gets cited in AI answers

Most briefs optimize for rankings. The ones that get cited optimize for extraction. Here's the operational difference.

Read post