Structured Data & Schema Does Not Help With Visibility In AI Search

09/15/25
Source: Search Engine Roundtable by barry@rustybrick.com (Barry Schwartz). Read the original article

TL;DR Summary of Testing Structured Data Impact on AI Search Visibility

Recent experiments reveal that adding structured data/schema to webpages does not currently improve AI search visibility. Large language models (LLMs) like GPT tokenize content in a way that breaks down schema markup, making it indistinguishable from regular text. Tests confirm that LLMs primarily extract information from visible page text rather than embedded structured data. However, this could evolve as AI technologies advance.

Optimixed’s Overview: Evaluating the Real Impact of Structured Data on AI Search Rankings

Understanding the Relationship Between Schema Markup and AI Search

Structured data or schema markup has long been recommended for enhancing search engine optimization (SEO). However, its effectiveness in improving visibility in AI-driven search results has been put to the test recently. Two independent experiments shed light on how AI models interpret schema markup:

  • Tokenization Process Limits Schema Use: Mark Williams-Cook demonstrated that LLMs tokenize webpage content, which effectively dismantles schema markup into ordinary tokens. For example, the schema key-value pair "@type": "Organization" is split into separate tokens “type” and “Organization” without retaining schema context.
  • Visible Text Over Schema Data: Julio C. Guevara conducted a test with two product pages—one with visible text and schema, the other with only schema and no visible text. AI models like Gemini and ChatGPT only extracted meaningful information from the page with visible text, ignoring data contained solely in structured markup.

Implications for SEO and AI Search Optimization

These findings suggest that while schema markup remains valuable for traditional SEO and rich results, its direct impact on AI search visibility is currently minimal. Since LLMs rely heavily on token sequences derived from visible text, ensuring content is readable and accessible as plain text is crucial for AI search comprehension.

That said, the landscape of AI search is evolving rapidly. Future iterations of AI models and search engines may better utilize structured data, so staying informed and prepared to adapt SEO strategies is advisable.