This is from the World Wide Words mailing list:

Turns of Phrase: Deep Web

“The World Wide Web has not only become so big that search engines
can’t index it all (in fact, they manage only a small proportion),
but there’s lots of stuff out there – mostly in databases – that
can’t be reached at all by the conventional search technologies in
use since the Web began. The firm BrightPlanet has estimated that
this ‘deep Web’ (a term it seems to have invented) contains 7,500
terabytes of data, compared with about 19 terabytes of data on what
it calls the ‘surface Web’, numbers impossible to visual in other
than the vaguest way. Even if these figures are overestimates, it
still suggests that there is a lot of material out there that would
be useful if only one could find it. The firm also points out that
the deep data is usually of excellent quality, and that most of it
is publicly accessible without charge. Now all we have to do is
find a way of getting at it.”

The entire 41-page BrightPlanet paper is available for online reading or download (as an Acrobat .pdf) here. And, to delve deeper, a Google search on the term is here.