Report: ChatGPT & Perplexity Treat Structured Data As Text On A Page

Why It Matters

If LLMs ignore schema semantics, SEO strategies that rely on structured data to influence AI answers may be ineffective, reshaping how marketers optimize for generative search.

Key Takeaways

• ChatGPT reads schema as regular page text.
• Invalid JSON‑LD still appears in LLM responses.
• LLMs currently ignore schema semantics for queries.
• Google and Microsoft claim limited schema benefits.
• SEO reliance on schema for AI may be overstated.

Pulse Analysis

Structured data, particularly JSON‑LD, has long been a cornerstone of modern SEO, enabling search engines to understand a page’s content and surface rich results. With the rise of large language models (LLMs) like ChatGPT, many marketers assumed that embedding schema would give AI‑driven search a similar boost, treating the markup as a shortcut to more accurate or featured answers. This expectation has driven a wave of schema‑first strategies, often prioritizing hidden markup over visible, user‑focused content.

The recent experiment by Mark Williams‑Cook challenges that premise. By creating a fake brand and hiding its address exclusively within deliberately broken JSON‑LD, he observed that both ChatGPT and Perplexity still retrieved the address when prompted. The models treated the markup as plain text, ignoring its invalid structure, which suggests they do not currently parse schema in the way traditional crawlers do. While OpenAI claims to pull shopping data from structured‑data feeds, and Google’s John Mueller notes that schema’s impact on LLMs is conditional, the test underscores a gap between official statements and observed behavior.

For businesses, the takeaway is clear: schema remains valuable for conventional search rankings and rich snippets, but it is not a silver bullet for influencing generative AI responses. Marketers should continue to prioritize clear, visible content that directly answers user intent, while monitoring LLM developments for any shifts in how structured data is consumed. A balanced approach—maintaining robust schema for traditional SEO while enhancing on‑page relevance—will future‑proof strategies as AI search evolves.