Google Analytics records a session whenever someone visits a site, timestamping when it starts and ends. Average session duration takes every session in a given period and averages how long they lasted, giving a rough read on how much time people spend with the site rather than just how many of them show up.
The metric works best as a comparison, not a standalone score. A landing page meant for a quick answer doesn't need a long average duration to be doing its job, while a long-form guide or a pricing page probably does. When duration comes in low somewhere it shouldn't, that's a cue to review the page the way a visitor would: is the content actually answering what brought them there, is a layout issue pushing people away before they read anything, or is the page simply mismatched to the query it's ranking for? One quirk worth knowing: duration calculations can undercount the very last page a visitor views in a session, since there's no follow-up interaction to timestamp its end, which can make single-page sessions look shorter than they really were.
