I don't understand this argument, but would like to.
I run https://everytwoyears.org, a political non-profit focused on ending the warrantless metadata collection of U.S. citizens' communications. From everything I know about these programs, they are _explicitly_ not collecting content of communications. These programs only collect the metadata about a communication. As citizens, we don't get to have a clear definition of "metadata" (that is classified!) but we can assume anything that isn't the message itself is at risk of being considered metadata, especially if it was shared with a service provider in the normal course of conducting business (i.e. routing a request).
For HTTP requests, I assume the body of the request would require a warrant before it can be persisted on a government server. The HTTP headers, if unencrypted, _might_ be considered metadata but I would be surprised. The IPV4 headers are more than likely metadata. DNS queries are more than likely metadata.
If you are trying to avoid _active_ surveillance, where your government has a warrant, a VPN isn't going to help you. If you are trying to avoid _active_ surveillance where your adversary doesn't need/want a warrant to search you, a VPN isn't going to help you. But if you are trying to avoid having your internet activity ending up, de-anonymized, in a metadata database that your government does bulk analysis on, a VPN does seem like it would help. It seems like it would help a lot.