We can clear up the semantic problem by having a better definition of 'real' and 'virtual' photons. And there is one. Davies began grappling with this issue in his QED treatment of the Wheeler-Feynman theory (Davies 1971, 1972, email me if you want refs) but I don't think he quite solved it. The problem is solved unambiguously in the transactional picture as follows: virtual particles are unconfirmed, nascent offer waves; whereas real photons are confirmed offer waves resulting in actualized transactions. Virtual photons do not transmit real energy, while real photons do. But in TI, a 'real photon' is just an actualized transaction.This is a classic discussion, and there isn't really a right or wrong answer here. Just levels of approximation, and what you are or are not prepared to take as a true statement (always like this with why questions), as Demystifier correctly points out, there is definitely semantics here.
Still, in so far as it might be *useful* to picture the photon that you absorb as really being there, then you are allowed to make use of such an ontology, with the caveat that there are perfectly acceptable methods that make no use of perturbative methods at all and that you have to be careful ascribing reality to things that are strictly speaking mathematical fictions (for all the correct reasons that Tom pointed out).
More quantitatively, in terms of the Davies theory, a virtual photon is just the time-symmetric propagator while a real photon corresponds to the pole in the Feynman propagator. (Davies considers the difficulty of a 'real' photon being an internal line -- but never quite solves it. It is solved by defining the real photon as an actualized transaction.)
As Davies noted, it is misleading to try to define the real vs virtual distinction in terms of 'off-shell' vs 'on-shell', since any emitted and detected photon has a finite lifetime. But the photon discussed in the example above is certainly a real photon since it transferred real energy to your eyeball. However virtual photons need not be regarded as mere 'fictions' -- yes, they are sub-empirical, but that doesn't mean they don't exist. People sometimes want to throw them out as mere artifacts of a purely mathematical process (perturbation theory), but arguably they play a genuine physical role, for example, in the Kondo effect. The mistake is to equate 'real' with 'empirical' i.e. to say if something is not detected or does not transfer real energy it cannot be 'real'. This is just a metaphysical presupposition. Remember that Ernst Mach thought atoms were fictitious, and he turned out to be wrong.
I argue in my book that the fundamental message of quantum theory is that there is a level of physical possibility beneath physical actuality. That is, as Heisenberg said, "Standing in the middle between the idea of an event and the actual event, a strange kind of physical reality just in the middle between possibility and reality.” He wasn't just speaking figuratively here. He was onto something.