.
Google has fully deprecated support for FAQ markup Schema (May 7th).
As to the LLMs, the more I look into it, the more convinced I am that they are not reading the schema – both my personal deep dives into it and what I am reading/watching. They are reading the content within the schema – the visible HTML content that we wrap with schema. But they do not assess the content relative to the schema; they ignore the schema.
As for the JSON-D schema, they don’t even look at that. When the LLMs are in discovery mode, they take the visible text content. If you specifically ask them to review the schema or JSON, they will, but that is not what they are using to generate responses to questions.
I went through several hours with Clause, ChatGPT, Perplexity, and Gemini to cut through the crap. If you ask them straight up for SEO/AEO recommendations, they will spout off the crap from various sources – Reddit, industry pubs, etc. But that is their discovery mode, where they take what’s out on the web and regurgitate it. What they are not doing is examining their own processes for developing the recommendations.
When you spend time asking them (ad nauseam) what they are doing, where they are getting information, and how they are deriving the answers, and dive deeper into it, you eventually get to the heart of it. We did this for a service industry, and it was very clear that the schema played no role in the LLM’s responses to questions.
This kind of makes sense. Schema was created to add structure to unstructured data sources. The LLMs don’t need that. What they need is good content and differentiation to justify placing one company above another in the recommendations. They understand the content; they don’t need markup to tell them what it is.