When Claude uses web search, I sometimes wonder how much of a cited page it actually managed to read. If a result is behind a paywall, blocks automated access, or only exposes a short preview, can Claude still include it as a citation?

I’m trying to tell the difference between a source Claude inspected and one that was merely surfaced in search results. Does the response make that distinction clear, or is there a way to check what content was available to the model?