There’s just no compelling reason to do it that way.
LLMs like ChatGPT are getting data from the entire web and then having humans manually tag and identify everything. Getting data from the API is actually less useful to that end, and they’d need to integrate separately with every individual website.
Most websites don’t even have an API in the first place, either, so scraping would still be necessary for most sites.
There’s just no compelling reason to do it that way.
LLMs like ChatGPT are getting data from the entire web and then having humans manually tag and identify everything. Getting data from the API is actually less useful to that end, and they’d need to integrate separately with every individual website.
Most websites don’t even have an API in the first place, either, so scraping would still be necessary for most sites.