Garbled HTML pulled for page description “caption id=attachment…”

That’s shortcode with the brackets stripped. And since shortcode is either rendered or removed before output, it suggests that these descriptions are not merely being “scraped” from your site, but provided by a WordPress plugin or locally-hosted script.

I would first check your theme is not “leaking” raw shortcode; do you see something like [caption...] in articles or excerpts when browsing your site?

And if not, are you using any plugins that might be communicating with an aggregator-type service, not necessarily the ones you mentioned? For example, SEO-related plugins.