The Problem With Letting AI Coach Your Freediving

Author: ALFC Team

Artificial intelligence has made it incredibly easy to get answers to almost anything. You can ask it to explain a complicated subject, create a training programme, analyse your progress or suggest ways to improve your performance. Within seconds, you'll receive a detailed response that sounds confident, organized and often surprisingly convincing. For someone trying to improve their freediving, that's an attractive proposition. Why spend money on coaching when you can describe your problems to an AI model and receive what looks like expert advice without leaving your house?

It's a reasonable question, especially when you consider how much information is now available online. Freedivers can watch instructional videos, read scientific papers, study training methods and discuss techniques with people from all over the world. AI makes that information even more accessible by bringing it together in a conversational format. You don't need to search through dozens of websites or understand complicated terminology. You can simply ask a question and receive an explanation tailored to your level of experience.

But there's an important difference between receiving information about freediving and receiving actual freediving instruction. One involves understanding what someone is telling you. The other involves understanding what someone is doing, how their body is responding and whether the decisions they're making are appropriate for their abilities and the conditions around them. Those are very different things, and confusing them can create problems that go well beyond a disappointing training session.

At ALFC, we believe technology can be extremely useful in freediving education. We also believe that knowing where its usefulness ends is just as important. An AI model can explain a technique, help you understand a training concept or organize information you've already collected. What it cannot do through a text conversation is replace the experience, observation and responsibility of a qualified instructor working with you in the water.

The problem isn't that AI has nothing valuable to offer. It's that the quality of its answers can make people forget how much it doesn't actually know about them.

A Convincing Answer Isn't Necessarily The Right Answer

One of the most interesting things about AI is how good it has become at explaining complicated subjects. Ask a model about the mammalian diving response, equalization, carbon dioxide tolerance or the differences between freediving disciplines, and it can produce a detailed explanation in seconds. Depending on the model, the question and the available information, that explanation might be very useful. It might also contain mistakes, outdated assumptions or recommendations that don't apply to your particular situation.

The difficulty is that both kinds of answers can sound equally convincing.

Imagine you're struggling with equalization at 25 metres. You've been diving comfortably to that depth for months, but recently you've started experiencing difficulties. You describe the problem to an AI model, explain that your ears feel blocked during the descent and ask what you should do. The model might suggest several possible explanations. Perhaps your technique needs adjustment, perhaps you're experiencing congestion or perhaps you're changing your head position during the descent.

Some of those possibilities might be reasonable. But the model doesn't know which one applies to you. It hasn't watched your descent, observed your equalization movements or examined your ears. It doesn't know whether you're describing ordinary difficulty, an injury or something that requires medical attention. It only knows what you've chosen to tell it, and even that information may be incomplete.

A qualified instructor has a different starting point. They can observe your technique, ask follow-up questions, assess what happens during your preparation and identify patterns that might not be obvious to you. They can also recognize when a problem falls outside their expertise and should be evaluated by an appropriate medical professional.

That doesn't mean every instructor will immediately identify the correct explanation. Freediving instruction isn't magic, and even experienced professionals can make mistakes. The difference is that an instructor has access to information that a text-based AI conversation simply doesn't provide.

There's another problem worth considering. People aren't always particularly good at describing their own technique. You might believe your head is perfectly aligned during a descent when you're actually looking down. You might think you're equalizing regularly when you're waiting too long between equalizations. You might describe a dive as relaxed even though your movements suggest you're working much harder than you realize.

If your description is inaccurate, an AI model can build an entire explanation around a problem that doesn't actually exist.

And because the response is usually well written, it can be easy to accept it without questioning the assumptions behind it.

This becomes even more complicated when people ask AI to confirm something they already believe. If you're convinced that your lack of progress comes from insufficient carbon dioxide tolerance, you might frame your question in a way that encourages the model to focus on that explanation. It may provide useful information about the subject without adequately challenging whether carbon dioxide tolerance is the problem in the first place.

A good instructor should be willing to tell you that you're focusing on the wrong thing.

Sometimes the most useful coaching advice isn't a complicated explanation. It's a simple observation that changes how you understand your diving.

You might spend weeks researching advanced training methods when the real issue is that your technique becomes inefficient as soon as you get tired. You might think you need to become mentally tougher when what you actually need is more recovery. You might be looking for ways to extend your breath-hold when your immediate priority should be improving basic skills and consistency.

An AI model can discuss all these possibilities. But without reliable information about your actual performance, it cannot determine which explanation is correct simply because its answer sounds authoritative.

And in a sport where incorrect assumptions can affect safety, that distinction matters.

Your Body Doesn't Follow A Training Programme Just Because AI Says It Should

Training programmes are one of the most appealing uses of artificial intelligence. Tell a model your age, experience, current personal best and training availability, and it can generate a detailed weekly schedule. You might receive recommendations for pool sessions, strength training, mobility work, recovery and different types of apnea exercises.

Everything can look beautifully organized. Monday has a purpose, Wednesday has another, and the entire programme appears to follow a logical progression towards your goal.

The problem is that a well-organized programme isn't necessarily an appropriate programme.

Freediving performance depends on many factors, including technique, experience, physical condition, recovery, psychological state and the environment in which someone trains. Two divers with identical personal bests can have completely different needs. One might be technically efficient but struggle with equalization. Another might equalize comfortably but waste energy during the descent. A third might have excellent physical abilities but become anxious when approaching unfamiliar depths.

Giving all three the same training programme because they share a similar personal best would make very little sense.

Yet this is exactly the kind of oversimplification that can happen when a training plan is created from a handful of numbers.

Even when an AI model asks detailed questions, the information remains largely self-reported. It doesn't know how you actually move underwater, whether your technique deteriorates under fatigue or how consistently you perform at different depths. It may not know about relevant medical conditions, previous incidents or recovery problems unless you mention them. And it cannot reliably infer those details from a personal best or a few descriptions of your recent sessions.

A qualified instructor can assess these things over time. They can observe how you respond to different exercises, adjust the difficulty of a session and recognize when the original plan no longer makes sense.

This ability to adapt is one of the most important parts of coaching.

Imagine you've planned a week of depth training. Your goal is to improve your consistency at 40 metres, and you've arrived feeling motivated. On the first day, everything goes well. On the second, your equalization becomes unreliable. On the third, you're unusually tired and your technique starts deteriorating.

An AI-generated schedule might still contain the same progression it recommended before the week began. You can certainly return to the model, explain what happened and ask it to adjust the plan. But once again, the quality of the adjustment depends on what you report and how accurately you understand the situation.

An instructor who has been working with you throughout the week has a much clearer picture. They know what your previous sessions looked like, how you behaved in the water and whether your performance has changed in ways you haven't noticed.

They might decide to reduce the training load, work on a different skill or recommend taking a break. They may recognize that the original goal is no longer appropriate for that week.

And sometimes that's exactly what good coaching looks like.

There's also a broader issue with treating training as a collection of exercises. People often assume that improvement comes from adding more work. More sessions, more repetitions, longer breath-holds or greater training intensity. But adaptation depends on how training stress is managed, not simply how much stress is accumulated.

In freediving, poorly chosen apnea exercises can be particularly problematic because hypoxia can develop without the warning signs people expect. Breath-hold training carries a risk of loss of consciousness, including in shallow water, and should never be treated as something that becomes safe simply because a programme looks scientifically structured.

A qualified instructor doesn't eliminate every risk, but they can provide appropriate supervision, safety procedures and individualized decisions that a written programme cannot supply.

The same principle applies to progress. Sometimes an athlete needs a more demanding session. Sometimes they need technical corrections. Sometimes they need to stop training and recover. The challenge is knowing which response is appropriate at that particular moment.

A programme can tell you what was planned.

A coach can help determine what should actually happen. 

An Instructor Notices Things You Don't Know To Ask About

One of the biggest limitations of using AI for freediving advice is that you usually need to identify the problem before you can ask about it.

That sounds obvious, but it's more important than it might initially appear.

Imagine you're training constant weight with bifins. You've been working on improving your efficiency, and you ask an AI model how to reduce energy consumption during your descent. It provides suggestions about streamlining, kicking technique, relaxation and equipment selection.

All reasonable subjects to explore.

But what if your main problem is something you haven't noticed? Perhaps your ankles are too rigid. Maybe your kick becomes unnecessarily wide when you approach your target depth. Perhaps your body position changes during the transition into freefall, creating resistance that you're not aware of.

You might spend hours discussing the wrong issue because you don't know what the real issue is.

An experienced instructor can observe those details. They can watch you from different angles, compare your movements between dives and explain what they're seeing. More importantly, they can give you immediate feedback and see whether a correction actually works.

That last part is particularly valuable.

Freediving technique is not always easy to understand through words. Someone can explain the ideal movement perfectly, and you might still struggle to reproduce it. You can understand the theory behind an efficient finning technique without being able to perform it consistently.

The difference between understanding a movement and executing it becomes obvious when you're in the water.

An instructor can demonstrate the movement, observe your attempt and make adjustments. They might change the exercise entirely if their original explanation isn't helping. They can recognize when you're trying too hard, becoming frustrated or compensating for one problem by creating another.

This kind of feedback is difficult to reproduce through a conversation with an AI model.

Video analysis can help bridge part of that gap. Some AI systems can examine visual information, and recorded dives can provide useful material for discussion. But even then, the analysis is limited by camera angles, image quality, what happens outside the frame and the system's ability to interpret the movement correctly. A video doesn't necessarily reveal how a diver feels, whether equalization is comfortable or what happened before and after the recorded section.

And none of that replaces someone being physically present to supervise the session.

There's also a psychological side to instruction that deserves more attention.

Freedivers don't always communicate what they're feeling particularly well. Some become quiet when they're nervous. Others talk more than usual. Some become frustrated after a disappointing dive, while others insist that everything is fine even when their behaviour suggests otherwise.

A good instructor gets to know these patterns.

They might notice that you're rushing your preparation, becoming distracted by another diver's performance or putting unnecessary pressure on yourself. They can recognize when encouragement is useful and when you need to be reminded that there's no reason to push further.

This is especially important because freedivers can become emotionally attached to their goals.

If you've spent months preparing for a particular depth, it's difficult to remain completely objective when you're close to achieving it. You may interpret a disappointing session as evidence that you need to try harder, even when a more conservative approach would be appropriate.

An instructor provides an outside perspective.

And sometimes that perspective is valuable precisely because it isn't telling you what you want to hear.

There's a tendency to think that good coaching should always feel encouraging. In reality, some of the most useful feedback can be uncomfortable. You might be told that your technique isn't ready for the depth you're targeting, that your preparation is inconsistent or that your expectations don't match your current abilities.

An AI model can certainly deliver critical feedback. But if it hasn't observed the underlying problem, its criticism may be based on assumptions rather than evidence.

An instructor who has watched you dive can explain what needs improvement and show you why.

That makes the conversation much more meaningful. 

AI Can Support Your Training, But It Shouldn't Be Your Instructor

None of this means artificial intelligence has no place in freediving. Quite the opposite. Used properly, it can be an excellent educational tool, particularly for people who enjoy understanding the theory behind their training.

You can use AI to explore the physiology of breath-holding, understand different training concepts or become familiar with terminology before attending a course. You can ask it to explain a scientific paper in simpler language, compare different approaches to training or help organize notes from a session with your instructor.

It can also be useful when you're trying to understand why certain recommendations are made. Perhaps your instructor has explained the importance of recovery, but you'd like to understand the underlying physiology. Or maybe you're working on a particular technique and want to read more about the principles involved.

These are reasonable uses of the technology.

The important thing is to treat the information as something to examine, not as an instruction that automatically applies to you.

For example, AI can help you organize a training diary. You can record your sessions, note how you felt and identify patterns worth discussing with your coach. If your training data is reasonably complete, it may help summarize changes in performance or highlight questions you haven't considered.

But identifying a pattern is not the same as identifying its cause.

If your performance has declined over several sessions, the explanation could involve recovery, technique, illness, stress or several other factors. An AI model might help you explore those possibilities, but it shouldn't turn a statistical observation into a confident diagnosis.

The same applies to equipment. AI can explain differences between fin materials, blade stiffnesses, wetsuit designs or dive computer features. It can help you understand what questions to ask before making a purchase. But it cannot guarantee that a particular product will suit your technique, body proportions or training needs without appropriate assessment.

At ALFC, we think the most useful approach is to combine accessible educational resources with proper instruction. There's no reason to reject technology simply because it cannot replace a coach. Books, videos, scientific articles and online seminars have always played a role in helping people understand freediving. AI is another tool that can make some of that information easier to access.

But there are boundaries that shouldn't be blurred.

Learning about freediving online is not the same as learning to freedive. Reading about equalization is not the same as developing reliable equalization skills. Understanding the theory of a rescue procedure is not the same as being able to perform it correctly when another person needs help.

Those skills require practical training, supervision and assessment.

And this is where the difference between an AI model and a qualified instructor becomes especially clear.

An instructor has responsibilities that extend beyond answering questions. They must consider the safety of the session, the abilities of the participants, the environmental conditions and the limitations of their own expertise. They need to recognize when an exercise should be modified, when a diver should stop and when additional professional assessment is necessary.

A text-based AI model doesn't carry out those responsibilities in the water. It cannot provide physical assistance, manage an emergency or take responsibility for the decisions being made during a dive.

It also doesn't know whether you've understood its advice correctly.

That's an important limitation because misunderstandings are common when people learn technical skills. You might read an explanation, believe you've understood it and then perform something completely differently from what was intended. In a classroom, that can be corrected through questions and demonstrations. In the water, an instructor can see the mistake and intervene.

Without that feedback, you may continue practising the wrong thing while becoming increasingly confident that you're doing it correctly.

And confidence without competence is not a particularly useful combination in freediving.

There's another reason we believe working with an instructor matters, and it has less to do with technology than with the nature of the sport itself.

Freediving is an activity that depends heavily on trust. You need to trust your training partner, your safety procedures and the people responsible for supervising your dives. You also need to develop an accurate understanding of your own abilities and limitations.

That understanding doesn't come from collecting information alone. It develops through experience, feedback, practice and sometimes difficult conversations about what you're ready to do.

A good instructor helps you build that understanding.

They don't simply give you a list of exercises or tell you how to reach your next personal best. They help you develop the skills and judgement needed to make better decisions in the water.

And importantly, they should be able to explain when you shouldn't do something.

This is where AI's tendency to provide an answer can become a disadvantage. When somebody asks for a training programme, they usually expect to receive one. When they ask how to improve their breath-hold, they expect practical suggestions. But sometimes the appropriate response is to ask for more information, question the premise or recommend against a particular activity.

Responsible AI systems can do those things, and they should. However, users can still misunderstand the limitations of the advice or seek another answer that better matches what they wanted to hear.

A qualified instructor can challenge those expectations in person, based on direct observation and an ongoing coaching relationship.

Of course, choosing an instructor matters too. Not everyone offering freediving advice has the same qualifications, experience or approach to safety. An instructor's title doesn't automatically guarantee excellent coaching, just as an AI-generated explanation isn't automatically wrong. It's worth looking for recognized qualifications, appropriate safety practices, experience with your level of training and a willingness to explain the reasoning behind recommendations.

Good instruction should make you more knowledgeable and independent over time, not simply more dependent on the person teaching you.

That's probably the best way to think about AI as well. It should help you ask better questions, understand more of what you're doing and participate more actively in your own training.

It shouldn't give you the impression that you've mastered a skill because you've read a convincing explanation of it.

At ALFC, we welcome people who arrive having researched freediving extensively. Some have read books, watched countless videos and spent hours exploring different training methods. That curiosity is a good thing. It often leads to interesting conversations and helps people understand the practical work we're doing together.

But when we get into the water, the conversation changes. We can see how someone actually dives, identify what needs attention and adapt the session to the person in front of us. We can work on skills in a controlled environment, provide feedback and make decisions based on what is happening rather than what was expected to happen.

That's the part technology cannot replace through a chat window.

Artificial intelligence will continue to improve, and some of its future applications in sports coaching may be genuinely impressive. Better movement analysis, more useful training data and improved educational tools could all help athletes and instructors make more informed decisions.

But better information doesn't automatically eliminate the need for practical expertise, supervision or human judgement.

Freediving isn't a subject you can master simply by asking enough questions. It's something you have to practise, experience and gradually understand through time in the water.

So use AI. Ask questions, explore the science, challenge your assumptions and take advantage of the information available to you. There's no reason not to.

Just remember that an AI model can tell you what a good dive is supposed to look like.

A good instructor can watch your dive, recognize what you're actually doing and help you improve it.

And when something goes wrong, the difference between those two becomes much more important than the quality of the explanation.