The current HTML spec says a cite attribute's value has to be a valid URL [0]. I don't know if that's a point at which the WHATWG and W3C versions differed, like on the <cite> element. I do think it came from a semantic web mindset, that the attribute isn't really for the user, it's for programmatic use by the web host, like a CMS, or other systems.
It used to be the case that the values of `content` properties weren't read by screen readers so this technique would mean the information was solely conveyed visually, that blind screen reader users wouldn't get it, but screen readers have been reading `content` properties for a while.
In any case, if the citation is important information to convey, it should probably exist at a text node in the document, not as metadata that gets pulled out and displayed by the styling.
It used to be the case that the values of `content` properties weren't read by screen readers so this technique would mean the information was solely conveyed visually, that blind screen reader users wouldn't get it, but screen readers have been reading `content` properties for a while.
In any case, if the citation is important information to convey, it should probably exist at a text node in the document, not as metadata that gets pulled out and displayed by the styling.
[0] https://html.spec.whatwg.org/multipage/indices.html#attribut...