If you develop a chat bot and you want to integrate QA into it you have a variety of choices:
1 - Build/store/curate your own knowledge base. (Lots of work, as I am sure you know)
2 - Use server side processing to accept a query and search for an answer. You could screen scrape a search engine request or you might use a search engine API. If you use a search engine, the returned response may include the answer, but often does not have it in the form of an answer (it might be a list of links and a text blurb). The AI would need to parse the search engine response and try to distill an answer. This approach often suffers from lag time as the query round trips between the bot-host-search engine and back.
3 - Use cross site scripting from the client. This may be faster but it can be fragile if the bot uses an Iframe. It is subject to changes in website policies (does the search engine allow results in Iframes), and even with a direct API may require filtering of the responses.
#3 is the approach I have used with Skynet-AI.