That Wikipedia page confirms that its widely disrespected, but read the Wikipedia on the actual studies performed. There is supporting research, some of it fairly impressive. The list of ratio of supportive studies to dismissive studies is very much skewed in support of NLP on this wikipedia page.
There are a few issues I see here.
1) NLP sets off big time "SCAM!" flags, since they seem to be trying to use NLP to sell NLP (to idiots).
2) Their theories can be useful, but are still crap. You can test it and "disprove" it by finding a flaw without finding the part that made it useful.
Because of these, it's going to take work to extract the value that's there.
3) It's hard to test things that have more than a couple causal factors. The hypnosis research, which is more respected, falls prey to this all the time. They measure one correlation resulting from a giant mess of factors without holding other factors constant (because they have failed to even identify them) and then are surprised when they cant' get consistent results for their oversimplified model.
4) "NLP" is being used too loosely. If they do a study that fails to find evidence for one theoretical claim NLP practitioners have made, its is interpreted as "NLP is bullshit". There are plenty of studies that go that way for various hypnosis theories, but it is interpreted as "Well then, hypnosis must work differently than that".
I ended up researching "hypnosis" instead of "NLP" because it seemed to be easier to extract useful information on the same subject (though there is still a huge pile of BS to sift through and piece together), but I will say that NLP can teach you to do stuff you couldn't do before.
Their "fast phobia cure" does work, for example. I've done it.
I just heard a comment by Braddock of Lovesystems that was brilliant: All that your brain does when you ask it a question is hit "search" and return the first hit it finds. So be careful how you phrase your question.
Say you just arrived at work, and realized you once again left your security pass at home. You ask yourself, "Why do I keep forgetting my security pass?"
If you believe you are a rational agent, you might think that you pass that question to your brain, and it parses it into its constituent parts and builds a query like
X such that cause(X, forget(me, securityPass))
and queries its knowledge base using logical inference for causal explanations specifically relevant to you and your security pass.
But you are not rational, and your brain is lazy; and as soon as you phrase your question and pass it on to your subconscious, your brain just Googles itself with a query like
why people forget things
looks at the first few hits it comes across, maybe finds their most-general unifier, checks that it's a syntactically valid answer to the question, and responds with,
"Because you are a moron."
Your inner Google has provided a plausible answer to the question, and it sits back, satisfied that it's done its job.
If you instead ask your brain something more specific, such as, "What can I do to help me remember my security pass tomorrow?", thus requiring its answer to refer to you and actions to remember things and tomorrow, your brain may come up with something useful, such as, "Set up a reminder now that will notify you tomorrow morning by cell phone to bring your security pass."
So, try to be at least as careful when asking questions of your brain, as when asking them of Google.