The original question asks if Facebook, Twitter, etc authentication schemes are PCI-compliant.
The Payment Card Industry (PCI) Data Security Standards (DSS) specification cites two separate and distinct things about authentication and it is the context of use that discriminates between the requirements:
Section 8 states that a unique ID must be assigned to each person with computer access.
A username "principal" (from whatever service) meets that requirement (is unique).
This section goes on to discuss verifying authorized users - which the password component would comply with. So this covers basic "authentication" (is this who we think it is) for card-holder verification.
However, it depends on what access is granted. For a stored payment instrument, for the card-holder to make a purchase against it, this would suffice (provided the payment-instrument structure storage met the PCI requirements, etc).
However, for "remote access" to the payment system (generally) - 2nd factor authentication (out of band via voice/SMS one-time-passwords for example) is required. In this case, Google would be compliant for example if their phone number callback option was implemented as a means of 2nd-factor authentication for remote access to the payment system. If the phone number were not available or Facebook does not implement 2nd-factor authentication for remote access to the payment system then it would NOT be compliant.