As long as you provide structural information, it's safe to assume that users of screen reader are familiar with variations in font presentation.
THIS. Basically, what you want to do is use structural, semantically meaningful markup. Trust the screenreaders (and the users of those screenreaders, who are responsible for the configuration) to do the right thing with that structural and semantically meaningful markup. They won't, all the time! Screen readers are imperfect tools, and they don't always react appropriately to the semantically meaningful markup. However, the tools are always improving, and the best thing you can do is trust the readers' tools to be doing what the readers want.
In other words, you don't want to be distinguishing between <cite> and <em> when you are formatting your text, avoiding using them sometimes because you don't want the emphasis to be spoken by a screenreader. Use the appropriate semantic markup, and trust that the presentation will take care of itself.
no subject
THIS. Basically, what you want to do is use structural, semantically meaningful markup. Trust the screenreaders (and the users of those screenreaders, who are responsible for the configuration) to do the right thing with that structural and semantically meaningful markup. They won't, all the time! Screen readers are imperfect tools, and they don't always react appropriately to the semantically meaningful markup. However, the tools are always improving, and the best thing you can do is trust the readers' tools to be doing what the readers want.
In other words, you don't want to be distinguishing between <cite> and <em> when you are formatting your text, avoiding using them sometimes because you don't want the emphasis to be spoken by a screenreader. Use the appropriate semantic markup, and trust that the presentation will take care of itself.