Yes. Always read the whole article before citing. Don't be sloppy. Don't take lazy shortcuts. Write responsibly. This is academia, not second-rate journalism.
If you cite something you don't understand, or which is riddled with nonsense, or is later retracted, and you show no sign that you were aware of it in your citation, then it will reflect badly on you.
Obviously, in any paper you cite, there will be things that are relevant, and things that are not. But that doesn't mean you can avoid reading the paper, and only cherry-pick reading the one sentence in it that is convenient for you to build your argument. If you don't read it all, you won't know everything in it that's relevant.
Building comprehensively on predecessors' work is how scholarship progresses. Being lazy, omitting hte reading of your predecessors' work, and citing it blindly without reading it or understanding it, is how bad scholarship propagates.
If it's a large reference work such as a dictionary, then no, you aren't expected to read the whole thing: in such cases, you'll just have to get by with an appeal to authority, by using the most reputable dictionary you can.