Does ChatGPT’s browsing tool extract JSON-LD (schema) along with visible HTML?

Documentation
1

Hi everyone, I’m trying to understand exactly how ChatGPT’s browsing feature works under the hood. In particular:

  1. When using the browsing plugin, does the tool fetch both visible page content (e.g. <h1>, <p>, etc.) and any <script type="application/ld+json"> blocks?
  2. If so, is there any official documentation or authoritative statement from OpenAI confirming that JSON-LD is scraped and merged into the text sent to the model?
  3. My own tests show that when a page includes schema markup, ChatGPT’s answers include details (emails, trial lengths, certifications) that are only present in the JSON-LD—not visible in plain HTML. This suggests the browsing tool is harvesting that structured data. Can anyone point me to definitive docs or clarify how the browsing tool handles schema?

Thanks in advance for any insights or links to relevant docs!

2

Not sure, but what I saw in responses it gave me that data was ignored unless available in body.