I feel like the core of the XKCD message (IRL creations will still matter) is accurate but the reasoning is lukewarm. The bottleneck for convincing fakes has always been technical, which is why in the past it was never super common and required experts to distinguish and also not as much of a concern broadly.
The pacifying message in the comic is meant to be that since doctored media has existed in the past and real media has thrived during the same time that real media will still hold meaning. It kind of glosses over though that it's only because of the smaller volume of it and ability to identify it (via human-made mistakes) that it hasn't been an issue.
Deepfakes (and anything convincingly 'AI' generated) aren't necessarily a willingness to lie. Most deepfaked videos are disclosed as such and made for fun, curiosity or sexual reasons. Similarly for most neural net generated images.
There is a valid concern that perhaps it will became too difficult for most, and perhaps variously for experts, to eventually distinguish fiction from reality and with the lower barrier to creating convincing fictional media will mean it could be everywhere not just limited to select, higher stakes scenarios / those attempting to lie (situations which inherently have greater scrutiny/skepticism).
One could picture this could increase cynicism for everything. Video, image, audio, text, handwriting, art, news. When people have to second-guess whether something is 'real' for things they wouldn't have expected before. Real media (if provable) will perhaps even have more zest in such a scenario, like some have speculated. I suppose not dissimilar to how Tom Cruise now prefers to do as much practically in the face of a CG-normalized industry where people have 'seen it all'.
