Hi Gabriel,
In Microsoft Word, the definition of a word is not exactly clear, but in AntConc, it is very precisely defined in the token settings. The default setting is to separate words like "doesn't" into "doesn" and "t", which is a common practice in corpus research. If you want a count that is closer to the count in Microsoft Word, I'd suggest you tokenize (or even better, POS tag) your files using a tool like my TagAnt tool, and then load the files into AntConc using the simple_word_pos_headword indexer in the corpus manager. If you do this, "doesn't" will probably get treated as one word or perhaps two words ("does" + "n't") depending on the tagger you use.
I hope that helps!
Laurence.