Schema markup is one of those AEO topics where 80% of the advice is wrong because it's recycled SEO advice with "for AI" added at the end. Yes, schema helps. No, slapping FAQ schema on every page does not move the citation needle.

We've tested schema implementations across dozens of B2B SaaS clients over the last year. Here's what actually moves AI citations and what doesn't, ranked by impact, with the honest version of when to bother and when not to.

What AI engines actually do with schema

Schema markup is structured metadata embedded in your HTML — typically as JSON-LD — that tells crawlers what your page is about in a machine-readable format.

Search engines have used schema for years to populate rich results, knowledge panels, and featured answers. AI engines use it differently. The honest summary:

Schema makes you easier to crawl correctly. Pages with clean schema get parsed more accurately, especially for entity disambiguation — figuring out which "Acme" is your Acme.

Schema feeds entity profiles. Organization schema with sameAs links connecting your site to your LinkedIn, Crunchbase, Wikipedia, and other canonical profiles helps AI engines build a confident picture of you. This is the highest-leverage schema use case for AEO.

Schema does not magically rank you. Slapping schema on a thin page won't make AI cite that page. Schema amplifies good content; it doesn't create it.

Schema does not equally help all AI engines. Gemini benefits most because it's search-grounded. ChatGPT and Claude benefit indirectly through entity disambiguation but don't "read schema" the way Gemini does.

The schema types that pay off, in order of impact

If you only implement four schema types for AEO, do these in this order:

Organization schema. The single highest-leverage schema for B2B SaaS AEO. Implement it on your homepage with full sameAs links, accurate description, founding year, logo, and key people. This feeds AI entity profiles directly and is a foundational component of entity authority work.

WebSite + SearchAction schema. Helps with navigation and tells engines your domain structure. Quick to implement, broad benefit.

FAQPage schema. Useful when implemented on actually-useful FAQ content. Useless or harmful when implemented on thin or duplicated content. Be honest about whether your FAQs would be useful to a real visitor — that's the test.

Article schema for blog content. Helps your published content get attributed correctly to your brand and feeds AI's understanding of what you cover. Implement on every published post including author and publish date.

Everything else (Product, BreadcrumbList, HowTo, Course, Event) is situational. If you have specific content that fits, implement it. If you're forcing it, skip it.

Organization schema: the foundation everyone gets wrong

Most B2B SaaS sites have basic Organization schema, and most of them have it wrong in subtle ways that hurt AEO.

The mistakes we see most:

Missing or weak sameAs array. The sameAs property is a list of canonical URLs that represent the same entity — your LinkedIn, Crunchbase, Twitter/X, Wikipedia if applicable, GitHub, YouTube channel, Wikidata. Most sites have 2–3 entries. Strong AEO setups have 8–12+, including industry-specific directories like G2 and Capterra.

Generic description. "Software company" is not a useful description. AI engines need specifics — what category, what for whom, what makes you distinguishable. Treat the description like a positioning statement.

Inconsistent name. Your schema name should match what's used everywhere else. If you're "Acme Inc" on Crunchbase and "Acme" in schema, you're forcing AI to disambiguate. Pick one.

Missing logo URL. The logo property feeds knowledge panels and increases entity recognition.

Missing founding date. Helps AI engines distinguish you from similarly-named brands and validates your maturity in the category.

These details sound trivial until you measure their impact. We've watched share-of-voice climb meaningfully on AI engines after a single Organization schema cleanup, with no other changes.

FAQPage and HowTo: when they help, when they hurt

FAQ schema is the most over-implemented schema type in AEO. Some teams add it to every page reflexively. This often hurts more than helps.

When FAQ schema works:

The FAQs are real questions buyers ask, with substantive answers. Useful to real humans first, parseable by AI second.

The page's content already structures Q&A. Don't bolt FAQ schema onto a page that doesn't logically have FAQs — Google has been clear they treat that as misleading.

The questions match how buyers phrase real queries to ChatGPT and Google. "What is X?" framing tends to surface less than "How does X compare to Y?" or "When should I use X?" style questions.

When FAQ schema hurts:

Used on pages where the FAQs are filler or duplicated. AI engines and search engines both detect this and discount it. Your overall trust signal weakens.

Stuffed with keyword variations of the same question. Google has explicitly warned against this since 2022.

Implemented as a workaround for thin content. Schema doesn't rescue weak content — it amplifies whatever's already there.

HowTo schema is the same story. Useful for genuine step-by-step content. Counterproductive when forced.

The honest test: would a thoughtful editor approve adding "Frequently Asked Questions" as a section title on this page? If no, don't add FAQ schema either.

Article, Product, SoftwareApplication: the edge cases

Article schema. Implement on every blog post. Include author with link to author profile (with their own Person schema if possible, including sameAs to LinkedIn). Include datePublished and dateModified accurately. Include publisher reference back to your Organization schema.

Product schema. For B2B SaaS, this is more situational than e-commerce. If you have a product page that genuinely positions a discrete product (not "our platform"), Product schema with aggregateRating from a verified source can help. If you're forcing it on a marketing landing page, skip it.

SoftwareApplication schema. Specifically useful for B2B SaaS. Implement on your main product page with applicationCategory, operatingSystem (often "Web"), and aggregateRating if you have legitimate review data. This is one of the few schema types that explicitly says "this is software," which helps AI engines categorize you correctly.

Course, Event, Recipe, etc. Skip unless they actually apply. Forcing irrelevant schema is a credibility cost, not a benefit.

Schema mistakes that quietly hurt your AEO

A non-exhaustive list of schema implementations we see hurting AEO performance:

Schema that disagrees with page content. If your schema says one thing and the page says another (wrong product name, wrong description, wrong date), engines downweight the schema and sometimes the whole page.

Schema with broken sameAs URLs. Linking to a Twitter handle that's now defunct or a Wikipedia entry that doesn't exist erodes trust signals. Audit sameAs links quarterly.

JSON-LD that fails validation. Use Google's Rich Results Test to verify every schema implementation. Broken JSON-LD often gets ignored entirely, costing you signals you think you're sending.

Schema implemented inconsistently across pages. If your Organization schema on the homepage says one thing and a different version appears in the footer of every blog post, engines have to disambiguate. Pick one canonical source per schema type and reference it consistently.

Schema that's never updated. Founding dates change, key people leave, descriptions evolve. Stale schema is worse than no schema.

The 30-minute AEO schema audit

Spend 30 minutes and you'll catch most of the issues:

1. Run your homepage through Google's Rich Results Test. Confirm Organization schema validates and includes name, description, logo, founding year, and at least 6 sameAs entries.

2. Search your brand on Google. Do you get a knowledge panel? If yes, is the description accurate? If no, that's a strong signal your entity work needs attention before more schema work is meaningful.

3. Pick three blog posts. Confirm Article schema with author, datePublished, publisher reference. Confirm dates are accurate.

4. Check your main product page. Confirm SoftwareApplication or Product schema is present and correct (if appropriate for your structure).

5. Verify all sameAs URLs return 200 status codes. Fix any that don't.

6. Check whether FAQ schema is implemented on pages that actually have meaningful FAQ content, and not stuffed onto product or marketing pages where it doesn't belong.

If any of those fail, fix them before adding new schema types — or grab a free AI visibility audit and we'll do this pass for you.

Schema is plumbing for AEO. It's not glamorous and it doesn't show up in growth charts directly, but the brands that win consistently in AI citations have always done their schema work properly. The brands losing on AEO have almost always neglected it.

Treat it like infrastructure: get it right, automate validation in your deploy pipeline if you can, audit quarterly, and move on to the higher-leverage work that schema is supposed to enable — content depth, citation building, and entity authority. The schema is the foundation. The work that matters happens on top of it.