Think about all the portrait photos on the web (including all those billions of facebook poses). Rather than concentrate on the forced smiles or the pouting, today’s invention is a software tool which can draw conclusions about a webpage based on the backgrounds of the facial images.
Detecting portraits is pretty straightforward. Once this is achieved, a texture and colour analysis should be enough to determine whether the portraits are:
taken all at the same time, which may indicate eg a completely fresh board of directors
taken using that strange, brown, cloudy background that is almost only used by corporate officers in behemoth companies
seemingly made in a library. A sure sign of lawyers, since they still think sitting in front of hand-tooled vellum makes them look smart
a random mixture of colour and greyscale shots: indicative of a complete lack of taste (you are probably on the website of a research group in Physics)
easily identifiable as having photobooth backgrounds -which suggests a particularly low-rent outfit.
This automated approach provides an extra search filter, allowing one to rule out of consideration a large number of unattractive organisations.