I'm using whatever the current one is here,
https://chat.openai.com/chat
It says March 13th version.
I do wonder how big the challenge is, introducing ChatGPT to the WWW. Currently, I presume it is trained on data that can be considered trusted. Giving it access, to, well, everything means it sees misinformation, sarcasm, meme's etc. Is there an ability to rank sources? How does it decide which information is more valid? One risk of LLM's is the seem to present their answers with great confidence, even if they are wrong. You can ask "are you sure?" and it will change the answer, but may still get it wrong. If you can't trust it to get the you the correct answer, it loses a lot of value. Adding in all sorts of information that may be ambiguous could make the answers more likely to be wrong.