Here's a fast one none tested that should work given that my knowledge of $ulevel counts has user level "200"

if user level is 200 and [question] is the first word then it writes the line to $1-

Then if a user types <data> it will search safely through the text file and treat it has plain text for <data> and if it matches should return data after [answer] as long has the word answer is spelled correctly.

Code:
on *:TEXT:*:#:{
if ($1 == [question]) && ($ulevel == 200) {
write ai.txt $1-
}
elseif (< isin $1) {
var %x = $read(ai.txt,nw,$+(*<,$1-,>*))
if (!%x) { msg # Sorry no match found for question $1- }
else {
msg # $gettok(%x,$+($calc($findtok(%x,[answer],1,32)+1),-),32)
}
}
}

Code:
if $reality > $fiction { set %sanity Sane }
Else { echo -a *voices* }