This post would be improved by discussing that their [threat model]( https://en.wikipedia.org/wiki/Threat_model ) is so different than Google's that it regards some of Google's business practices as threats. And that, in turn, there are threats that Google treats as much bigger threats, bringing their own world-class security team to. Calling this fundamental difference in approach "more secure" manipulates the less-…
A missing threat model gives this more credit than it merits. You can replace 'zero knowledge' with 'military grade' and it will be just as meaningful (if somewhat more obviously poor). Plus, you could say they are describing a threat model. If ProtonMail were compromised in this one particular way the confidentiality of your mail would be 'stronger' or 'improved'. This should be as reassuring as 'Switzerland', which…
Without a threat model (that is, the set of threats that one is trying to secure a system against), you have no idea what someone means by "secure". It could mean unpickable doorlocks, it could mean unbreakable windows, it could mean angry-Hippopotamus-proofing. It could mean that you smelly farts can't escape your pants.
Any claim of security without a threat-model is in the most literal sense meaningless. And don't get me started on "Military Grade Encryption", which is a term that at this point should give you a sense of concern, rather than safety.