In the ever-evolving world of SEO, one question has sparked intense debate: Does adding schema to your website actually improve your chances of being featured in Google's AI Overviews? While it's tempting to assume that machine-readable data would help, AI Overviews are the result of "complex and layered processes," making it difficult to draw firm conclusions from logic alone.

Join researchers Molly Nogami and Ben Tannenbaum as they move beyond speculation and put this question to the test in a groundbreaking, head-to-head experiment. Motivated by the limitations of their own previous studies—one a correlational analysis that didn't assess schema quality, the other a test on ChatGPT that didn't address Google Search—they collaborated to find a more definitive answer.

The setup was a carefully controlled test designed to isolate one crucial variable: schema quality. The team built three nearly identical, single-page websites:

• One featured well-implemented schema, following all best practices with complete Article, FAQ, and Breadcrumb data.
• One had poorly implemented schema, riddled with deliberate errors like missing required fields and incorrect date formats.
• The third had no schema at all.

Aside from the schema, every other element was kept consistent, from the site build to the keywords, which were all chosen to have identical metrics in Ahrefs. After submitting the pages to Google on the same day, the results were stunningly clear.

The winner was unambiguous: the page with well-implemented schema was the only one to appear in an AI Overview. It also achieved the best organic ranking of the experiment, reaching Position 3 in traditional search. In stark contrast, the page with poor schema peaked at Position 8 and never surfaced in an AI Overview, while the page with no schema was crawled but never even indexed by Google.

While the authors are cautious not to call the results "absolute proof," they present a compelling story that schema quality—not just its presence—plays a vital role in AI Overview visibility. But could it have been an algorithmic fluke? Find out what's next as they plan a follow-up experiment to swap the schema treatments and see if the results can be replicated—a test that they believe could yield an even more "compelling result".