I tried clicking on one of the results to find the meta tags for myself in the page’s source, but it seems they only appear if you’re Google (or another crawler like Bingbot).
Google has a URL Inspector tool which allows you to see what Googlebot would see when scraping a page, which should be useful for something like this; however, it doesn’t allow inspecting urls on domains you don’t own. So let’s Google a workaround…
We can conclude that TikTok is experimenting with using a LLM (probably their own) for generating structured data, but it seems a step is missing in their process, and just raw JSON is being fed to search engines.