Which schema types actually get your content cited by AI tools like ChatGPT and Perplexity?
Which schema types actually get your content cited by AI tools like ChatGPT and Perplexity?
Our team has been testing structured data across client projects for the past few years, and what follows is based on what we’ve actually observed, not theory. The old playbook was simple: add FAQ schema, get rich snippets, watch CTR climb. That world is mostly gone. Google stripped FAQ rich results for most sites in 2023, and a lot of people wrote schema off entirely.
Schema didn’t die. It changed jobs. The question is no longer “will this get me a star rating.” It’s “will an AI model understand my page well enough to cite it.
🪄Article, News Article, Blog Posting
The one everyone skips because it feels obvious. Don’t. It’s doing attribution work. When an AI Overview or Perplexity answer pulls something, it needs to know who wrote it and whether they have any business writing it. Article schema with a proper author field linked to a real Person entity carries the E-E-A-T signal.
One of our finance clients was getting scraped into AI answers with zero attribution. We added proper author markup, and roughly six weeks later the brand name started appearing alongside the pulled content. The same pattern showed up across their other properties.
🪄FAQ and How To
The rich snippet is gone; the schema isn’t useless. Clean question-answer pairs map almost exactly onto how people prompt models. FAQ markup essentially pre-chunks your content into the format retrieval systems want. How To does the same for step-based instructional answers.
🪄Product, Offer, Review
Non-negotiable if you sell anything. Shopping related AI results and Perplexity style comparison tables pull structured product data; specs, price, ratings. Without markup you’re betting a model will correctly parse your page layout. In our experience, that bet loses more often than people expect.
🪄Organization and Local Business
This one surprised the team. Entity level brand data directly affects how consistently AI tools describe a company. We’ve watched models state wrong founding years, wrong headquarters, wrong service descriptions for businesses with sloppy Organization markup. Clean up the schema and the descriptions tighten. You’re feeding the knowledge graph directly.
🪄Person and Author
Related to the Article point, but worth separating. If you’re positioning someone as a subject matter expert, Person schema with same As links to their profiles gives the model something to anchor to. Otherwise “Sarah Chen, marketing consultant” is a text string connected to nothing.
🪄Event and Webinar
Narrower, but if you run events, ChatGPT Pulse and similar feed products are surfacing them. Low effort, occasionally high reward.
🪄Dataset and Research Study
The most underrated of the group. If you publish original research, benchmarks, or survey data, mark it up as a Dataset. AI answers citing statistics need a source, and structured proprietary data is exactly what they reach for. Most teams publishing survey results dump them in a blog post with no markup at all.
🔔The honest caveat
None of this works on thin content. Schema is a delivery mechanism, not a substitute for having something worth citing. We’ve seen sites bolt on every schema type imaginable to a 400-word listicle and wonder why nothing moved.
Content first. Machine readable second. In that order.
If you’re currently running nothing, start with Article, Organization, and Person. Those three cover the most ground for the least effort.