Incorrect information provided by the AI using the search results

Brave search AI is providing incorrect information about wireguard vpn.

Wireguard does support TCP. but even after correcting the AI in the followup question and refreshing the page it went back to insisting that it doesn’t support TCP.

Is this something that can be corrected?

@TDK just a FYI, I just asked Google:



Oh, and to note, the one is an official link:

TCP Mode

WireGuard explicitly does not support tunneling over TCP, due to the classically terrible network performance of tunneling TCP-over-TCP. Rather, transforming WireGuard’s UDP packets into TCP is the job of an upper layer of obfuscation (see previous point), and can be accomplished by projects like udptunnel and udp2raw.

Keep in mind that AI uses sources it can find. Such as Brave Search or even Google would be checking the sources it sees. They aren’t guaranteed to be 100% correct and takes people searching for more. These are tools just to see basic summaries.

And as I showed in the post above, including the official stance from Wireguard themselves, I wouldn’t quite say that it has been wrong. At least it can be debated and get down to people having to be a bit pedantic on things.

Lastly, just checked on Brave Search:

It details some extra information. Again, based on question asked. As it said and is confirmed by Wireguard themselves, they don’t explicitly support TCP.

But more important is the bottom of AI results:

Not only does it provide sources for you to fact check and see what you think about credibility, but it also emphasizes: AI-generated answer. Please verify critical facts.

protonvpn allows me to use it over TCP.
Also google’s result isn’t correct.
As far as i remember, Protonvpn uses its own wireguard implementation and it does allow it over TCP.

lol, what about Wireguard. You going to go tell them they are wrong on their website too? I think what you might have is as they describe, that the transforming of UDP into TCP is upper layer of obfuscation using other projects. But I’m not sure. Regardless, I’d say using the official website would be a good and authentic source.

FINAL NOTE

If you ever do see anything from Brave Search that you think is wrong or should be changed, it’s best to report it via Brave Search. If you don’t know how to do that, you should see something of a settings thing on the upper right. In this screenshot, you’ll see it on the far right. image

Clicking it will show a menu like below:

You’ll see on the bottom is a Share Feedback thing.

From there it pops up with a thing like below. You would input the information and give as much detail as you can. This goes directly to the Search team for them to consider.

The Project Manager for Search said this is always the best way to get information to Search for anything people think is incorrect or any other issues experienced with Search.

Wireshark is the source of truth.

here you go from a random server im connecting to using protonvpn.

Im a network engineer, and when someone wants to argue about something related to the network, i fire up wireshark to prove my point.

Packets don’t lie.

Obfuscation or not, we are talking about the support itself. the transport is TCP just like the udp based.

@TDK did you ignore links? Go check out https://www.wireguard.com/known-limitations/#tcp-mode

As said, this is the official website. It’s possible things have changed but how would you expect AI to know that? All it’s doing is checking things like the official pages and any discussions that have occurred online.

I just was even trying to skim through their official sites and all but can’t find anything. Every single source I’m finding explicitly says it doesn’t use TCP.

And heck, even article written 7 days ago says the same thing:

Main thing is I’m not pretending to have any firsthand knowledge. The only thing I’m communicating with you is that there don’t seem to be any sources that align with what you’re saying. So that would be the major disconnect.

1 Like

@TDK nice, but now other thing I’m seeing is that it’s differentiating. That referred to Stealth.

image

It’s definitely interesting. At least in talking I’m hoping you can kind of see where it would be confusing. Especially if you spend time looking online, you’ll see most results are saying it’s not supported.

Then on the one we did find, it’s making it sound like it’s a different version of WireGuard.

Though I did also just find https://www.reddit.com/r/ProtonVPN/comments/y1sjbm/stealth_vs_wireguard_tcp_for_better_anonymity/

And while things may have changed, at least back then they did say:

WireGuard over TCP is not designed to bypass censorship.

So depending on what your original question may be, it still would be accurate in saying ISP might block it since even official team had said back then that WireGuard over TCP was not designed to bypass censorship.

Honestly, I’m glad you asked as I’m learning a lot right now. And hoping you see I’m not arguing or saying wrong. Merely trying to explain that in terms of AI and what information is on the internet, it’s kind of diverse and complicated.

Well not to sound bashful and ungrateful. but at first you sounded like you are arguing as i backed up my information with first hand capture from my network.
Wiretap is what we use in network troubleshooting and Wireshark allows us to dispute claims by other IT teams when they want to push the blame on us network engineers.

I appreciate you responding quickly on the thread and we might be from different cultures and that’s why we are perceiving each other responses in different tones.

All i’m saying, i would appreciate if the AI can go to the source of the vendors first before it reaches out to reddit to provide inconclusive info.