The Real Reason NLP Doesn't Work for Some People and Works for Others

There is a question that hangs over every conversation about Neuro-Linguistic Programming. Someone asks it in the back of every room I train in, and the loudest voices online are sure of the answer.

Does NLP actually work?

I could give you the salesman answer. The honest answer is more interesting than either side wants to admit. The question is not a question about technique. It is a question about epistemology. And once you see that, the whole debate rearranges itself.

Here is the lived version. In 1997 I flew to Duluth, Minnesota to train with Richard Bandler at the Black Bear Hotel and Casino. I was in my second year of university. I took extra days off school to do it. Two friends came with me, Devon and Wiki Tay.

When I came back to Singapore, I ran my first program with my business partner Jerry. Seven participants. Two of them were my parents.

My mother sat through the whole thing. At the end I asked her what she took away from it. She did not quote a technique. She did not name a pattern. She said: "State management. You can manage your emotions much better now."

That was the moment I knew this work was about giving people choice, not fixing what was broken.

But here is the part that matters for this lesson. My mother did not ask for a controlled trial before she accepted that she could manage her emotions differently. She felt the shift in her own nervous system, treated that feeling as real data, and acted on it. That is the entire division in the "does NLP work" debate, compressed into a single sentence.

The skeptic demands external proof before they will treat internal experience as data. The practitioner treats internal experience itself as the data. Those are two different operating systems. They will never agree, because they are not arguing about the same thing.

The honest fact is that NLP is a modeling discipline, not a body of clinical trial evidence. That is not a defense given by a salesman. It is the reality of what the field actually is. Richard Bandler and John Grinder built the original models in the 1970s by studying therapists who got results, most famously Milton Erickson, and writing down the structure of what they did so it could be taught. The Structure of Magic (1975) was an attempt to make explicit and transferable the patterns that experienced practitioners already ran. Frogs into Princes (1979) carried the same spirit into print for a general audience.

That is modeling. You study the behavior of someone who produces an outcome, extract the sequence, and teach the sequence. Modeling is how we learn most of what we do in life. You do not need a double-blind trial to learn to ride a bicycle; you need to watch someone, try it, and adjust on the feedback.

The problem arrives when people take a modeling discipline and demand it produce the evidence of a pharmaceutical trial. That is a category error, and it guarantees the debate never resolves. The skeptic is right that NLP has not been proven in the way a drug is proven. The practitioner is right that they have experienced real, repeatable shifts in their own and their clients' experience. Both statements are true at the same time. The argument only continues because each side insists the other is using the wrong standard of evidence.

There is a 200-plus comment thread on Reddit from May 2025 titled, plainly, "Does neuro-linguistic programming actually work?" Read it and you will see the exact split I am describing. Practitioners describe changes in internal experience and behavior. Skeptics respond that these are placebo effects and anecdote. Neither argument moves the other, because both are talking past the question of what counts as valid evidence in the first place.

The uncomfortable truth is that the whole field sits on a foundation borrowed from a man who made exactly this point half a century earlier. Alfred Korzybski, in Science and Sanity (1933), gave NLP its most famous line: the map is not the territory. The sentence is not a slogan. It is a foundational epistemological position: your map of reality is not reality itself, and any map is only useful to the degree it helps you navigate. Korzybski never claimed the map was the truth. He claimed the map was a tool. That is the exact posture NLP inherits.

Seen this way, the debate stops being about whether NLP is true and becomes about whether it is useful. And usefulness has a different standard of evidence than truth.

Let me now deal with the science directly, because an honest lesson cannot wave away the strongest objection. In 2010, psychologist Tomasz Witkowski published a systematic review in the Journal of Contemporary Psychotherapy examining claims made for NLP. His conclusion was blunt: the core claims had not survived controlled investigation, and what evidence existed was weak and inconsistent. Witkowski went on to argue, in later work, that NLP ought to be kept out of academic psychology until its premises met the same evidential bar as other interventions.

I am not going to argue with him, because he is describing a real and correctable condition. The field has a marketing problem, a certification inflation problem, and a research deficit. Too much of what passes for NLP training is packaged enthusiasm, and packaged enthusiasm does not survive contact with a controlled trial. Any honest practitioner has met the person who paid a large fee for a certificate and then could not reproduce a single reliable change in their life. That person is not lying when they say NLP did not work for them. They are reporting accurately on what they got.

But here is the part the review does not capture, and the part I think Witkowski himself would accept: the critique targets the overclaim, not the substrate. The substrate of NLP is the claim that vivid mental imagery, carefully structured and sequenced, changes emotion and behavior. That substrate is not fringe. It is documented across a broad mainstream literature. Mental simulations of an action activate overlapping neural networks with the actual action. Imagery-based rehearsal improves motor and emotional performance. The psychology of mental imagery and cognitive reappraisal supports the general mechanism, even when the specific NLP packaging has not been experimentally validated.

That distinction, substrate versus packaging, is the honest way to read the debate. The packaging has overclaimed and under-delivered in too many settings. The substrate, that the structure of your internal experience is changeable and that changing it changes your decisions, is a reasonable and usable hypothesis with everyday evidence behind it.

Now I want to bring this down to the founder's life, because that is where the debate either costs you or pays you.

Founders are running a company on a model of the world they built from their own experience. Nobody has run a controlled trial of your specific market, your specific customers, your specific personality before you ship. You decide based on the data you can gather internally: the knots in your stomach around a hire, the energy that appears around one product and dies around another, the pattern you notice in how you keep making the same strategic mistake. The founder who treats that internal experience as data moves faster and corrects sooner than the founder who waits for external proof that will never arrive in time.

Think about a hire you got wrong. You ran the interview process, the credentials checked out, the references were warm. All the external data said yes. And there was a hollow feeling you could not name, and you ignored it, because you told yourself feelings were not data. Months later the hire fails. The external data was correct at the moment you gathered it. The internal signal was correct too, and you threw it away because you had decided it did not count.

That is not a mystical story. That is a founder who has not yet admitted that their nervous system has been gathering information about people and situations for decades, and that the information is real even when it does not arrive in a spreadsheet.

Let me be direct about the mechanism, because this is where NLP is genuinely useful and not mystical.

Your internal experience is built from three ingredients: the pictures you make in your head, the sounds you run, and the feelings those produce in your body. NLP calls these representational systems. Every thought you have is assembled from them. When you replay a failure, you are running a specific sequence of pictures, sounds, and feelings. When you feel confident, you are running a different sequence. You are not a passive receiver of these; you are the operator of them, whether you know it or not.

Most people never notice they are running them at all. That is the whole game. Someone who knows NLP has trained themselves to notice the sequence, and that noticing is what gives them the option to change it. Change the sequence and you change the felt experience. Change the felt experience at the moment of a decision and you change the decision.

There is a detail that makes this concrete, and it is worth slowing down for. Within those three systems, each one has finer-grained settings that NLP calls submodalities. A picture is not only a picture. It is near or far, bright or dim, in color or in black and white, a still image or a movie, viewed from inside your own eyes or from outside looking at yourself. A sound has a location, a volume, a tone, a tempo. A feeling has a size, a temperature, a location in your body, a movement.

Two people can be replaying the exact same memory of the exact same failure, and one will feel terrible while the other feels nothing, because they are running the memory with different submodalities. The content is identical. The settings are not. This is the single most useful and least appreciated idea in the whole field, because it gives you a concrete handle on experience you thought you had no control over.

Let me give you a composite of the coaching room, built from several founders I have worked with, because this is the pattern that shows up again and again. Call her Emily. She was running a growing services business and she could not bring herself to raise her prices, even though every external signal said the work was undervalued and the clients would pay. She knew the economics. She had the spreadsheets. She could recite the margin math. She still delayed the conversation for six months.

When we slowed the moment down, she found she was running a picture of the conversation in a very specific way: a still image, in black and white, of a client's face looking disappointed, viewed from across the room as if she were an observer watching herself get rejected. Every ingredient was working against her. The still image froze her. The black and white drained the warmth. The external viewing position made her a spectator to her own failure.

The content of the image never changed. We changed the ingredients. We made it a movie, in color, viewed from her own eyes, with the client actually saying the words she was afraid to hear and then agreeing. The felt experience changed in the room. The knot in her stomach loosened before we even talked about price. She raised prices the following week and did not lose a single client. Some of it was the technique. Most of it was that she finally treated the contents of her own head as inspectable and changeable, instead of as weather she had to endure.

Emily is a composite, not one person, and I want to be plain about that, because the honesty gate of this lesson runs both ways. But the pattern is real, and I have seen the variation of it dozens of times.