Answer
What schema markup do you need for AI search?
Last updated
Short answer
At minimum, publish JSON-LD for Organization (or LocalBusiness if you serve a location), Service or Product for what you sell, FAQPage for your questions and answers, Article or TechArticle for substantive guides, and BreadcrumbList for hierarchy. Link them with stable @id values so engines read one connected entity, and make every fact in the markup identical to what a human sees on the page.
Why schema matters more for AI than for search
Rich results were the old reason to add schema. The new reason is better: schema is the only part of your page where facts are unambiguous. Everything else is prose a model has to interpret.
When an engine has to choose between two businesses and one of them states its hours, service area, phone number and offerings in machine-readable form, that one is cheaper to trust.
The minimum viable set
Five types cover almost every business, and adding more without accuracy is worse than adding fewer done properly.
- Organization or LocalBusiness - name, url, logo, telephone, address, areaServed, sameAs
- Service or Product - one per distinct offering, with provider and areaServed
- FAQPage - every question and answer pair you publish
- Article or TechArticle - for guides, with author, datePublished and dateModified
- BreadcrumbList - a real hierarchy, matching your visible navigation
Use a connected @id graph
Isolated schema blocks on separate pages force an engine to guess that they describe the same company. Give your organization a stable @id such as https://yoursite.com/#organization and reference that same @id from every Service, Article and WebPage you publish.
One connected graph resolves to one confident entity. Fragments resolve to uncertainty.
sameAs is the field most people leave empty
sameAs is where you point at the independent profiles that corroborate you - LinkedIn, X, Crunchbase, Wikidata, industry associations, review platforms. It is how you convert "this company says it exists" into "several independent sources agree".
Only list URLs that are live and genuinely yours. A sameAs pointing at a dead or wrong profile is actively harmful, and pointing only at your own domains adds nothing at all.
Rules that keep schema from backfiring
Schema that contradicts the page or overstates reality is a liability. Two hard rules: every fact in your markup must be visible to a human on the same page, and never publish review or rating markup for reviews you did not actually receive.
Validate with Google Rich Results Test and Schema.org validator after every change.
Frequently asked questions
Is JSON-LD better than microdata for AI search?
Yes. JSON-LD is the recommended format, it is easier to keep consistent, and it can be emitted independently of your page markup.
Can I add FAQPage schema for questions not shown on the page?
No. Markup must reflect content visible to users on that page. Hidden-only FAQ markup is a guidelines violation and can get your structured data ignored.
Should I add AggregateRating to look better?
Only if you have genuinely collected those ratings and display them on the page. Fabricated ratings are the fastest way to lose structured data trust entirely.
Related reading
Written by Benjamin Heckendorf, Founder & CEO of MyFast.ai.
Find out what AI says about your business
Run 100 real buyer searches across all seven engines and see exactly where you are named, cited or missing. First report $99.
