Actually, only people who read poorly do that. People who read well decode printed words directly into mental concepts, rather than sounding them all out, only sounding out a word when it is unfamiliar in print.
That (essentially correct) observation led to the creation of the "Look-Say" method of teaching English and its replacement of "Phonics" in the public schools.
Look-Say attempted to skip the "learn new words by sounding them out" step and teach students immediately to use the faster words-as-a-chunk technique of good readers.
But that ended up crippling them, because it left them with no way to acquire new words. They knew the handfull they'd encountered in class as a set of pictograms but didn't have the "secret code" to parse somethig they hadn't seen before. Result: Mostly illiterate graduates whose reading was so painful to them that they did little, getting farther and farther behind.
Turns out that good readers of substantially phonetic languages start with sounding-out (Phonics-style). Then as they gain skill and experience they start recognizing progressively more words at-a-glance, falling back to sounding-out when they hit words for which they hadn't yet built a neural-net recognizer. Eventually the "speed-bump" words become so rare that they blaze along familiar vocabulary without appearing to sound-out at all. But new or rare words bring out the old toolset, rather than bringing them to a full stop.
There are a corresponding pair of methods for learning a "second (i.e. additional) language: The "Grammatical Method" (learn and practice the lnguage rules) and the "Audiolingual Method" (repeat the samples). The latter came from an attempt to emulate the rapid language acquisition of children by modeling their environent
Tested right after a series of courses, college students taught by either method score about the same. Tested a year or so later (if they haven't been re-exposed to the second language meanwhile) those taught by the Grammatical Method had a significant skill loss, while those taught by the Audiolingual Method were unable to emit any sentence they hadn't encountered in class. Oops!
Turns out that (unless you learn two or more languages as a child) the neural structures that make kids little language acquisition machines literally die off, in several stages (at the ends of age ranges called "critical periods") as the neurons that weren't used by the language learned are "pruned". Once this has happened, learning a new language isn't impossible. But it's more like recovering from a stroke.