Autotune is essentially a computer program that can take someone's singing and put it "in tune."
On My Own, for example, starts in D minor and then modulates to Bb major, before ending in F major. (Typically speaking, it's also sometimes performed from E to B to F#, but that's not really important.) When Frances Ruffelle sings On My Own, because she's in key, she hits those notes.
Now, let's say I -- a not very good singer -- decide to sing On My Own. It doesn't sound very good, so someone kindly offers to autotune me to make it sound more in key. They program the autotune to take the song and make any adjustments required for it to go D- to Bb+ to F+. When they put in the Frances Ruffelle version and then autotune spits it back out you can't hear any difference because the program didn't have to do anything to make it fit that key. When they put in my version and autotune spits it back out, the program has to alter the recording to force the sound of my singing to fit. Now let's say I'm not as terrible as I think, the autotune doesn't have to do that
much work, so maybe an experienced ear can hear that my singing has been adjusted a little bit, but it didn't have to be adjusted a lot
so it's not distracting. And then let's be realistic and say that my singing is terrible and the autotune had to work really hard to make my voice sound like it was hitting the right notes, so now what it spits out is in the right key but sounds really tinny and computerized. All autotune does is alter sound to fit in the key of your choosing, so it doesn't necessarily make something better.
That's what it's designed to do, anyway. You can also use it to get creative and play around with how a voice sounds. Today's pop music is totally full of it.
Point is: if you're used to hearing live performances, (live-theatre especially), then you can usually tell the difference between someone who is singing and someone whose singing has been adjusted by a computer. And the reality is: if the singer was in key in the first place, you wouldn't be able to hear any autotune, which also means that it's not necessarily every single word/note that's been adjusted, just the ones that were off. So take this clip of Emma Watson singing "Belle - Reprise" for example.
It's not so
bad until about 00:38, but then it's like "oh."