AI, physical training, and YOU

AI is on the forefront of people’s minds, and the impact of this technology is being felt. Its impact is far reaching, combination of making our lives easier but at the same time forcing change including on employment and the economy.  There is a growing realization that just like the railways did some 150 years ago, or the dot.com bubble of 25 years ago, over-investment in this great technology will consequential a risk to the downside to the greater economy.

As history has shown that after the dust settles on the inevitable economic bust years, the very same technology will be consequential in rebuilding the economy.  We will slowly forget AI’s association with the late 2020s downturn and embrace the upside as we rebuild in the 2030s.  It’s not going away; it’s only going to get more significant in our lives.

So where does that leave physical training?

As an end-user, you can grab a free workout program in a matter of seconds from AI that is far more impressive looking that most ‘professional’ physical preparation coaches can produce.

Here’s an advertisement example of this:

Impressive appearance aside,  will this AI delivered training program meet your training goals and overcome physical challenges any more effectively?

As a physical training ‘professional’ (and I will be placing inverted commas around these two words until the ‘profession’ actually becomes professional) you can take advantage of the AI to provide you with programs to give your clients in the same way they can get them. Which raises the question of impact on employment – will the consumer cut out the ‘middleperson’? i.e. will AI reduce the demand for physical coaches?

Most would agree AI presents unique challenges over and above the benefits to both the end user and the ‘professional’ in physical training.

I’ve witnessed the consumer demand for physical training grow at a far more rapid pace during the half a century than I had anticipated. This gave opportunity to newcomers to the ‘profession’ at a rate faster than appropriate training or industry guard rails on training has been able to keep up.

The question I asked as this trend became apparent post the turn of the recent century was ‘who’s training who?’, a reference to the growth in entrants to the ‘profession’ resulting in their ‘professional development’ being conducted by individuals who I suggest lacked experience to teach.

And that was before the internet really kicked in.

The arrival of the internet meant anyone with a keyboard and an internet became an online teacher.  No experience or evidence of competence required.

And that was before the rise of social media. The content development increased dramatically in the race for eyeballs and followers.  Definitely no objective or regulatory standards around who was doing the teaching.

And it is in this environment that AI has entered. So where does AI get its information?

I asked Google AI ‘where does AI get its information from’, and it told me:

AI gets its information from three main sources: pre-trained neural networks, live web searches, and licensed or private datasets. [1]

1. Training Data (Pre-Trained Knowledge)

When an AI model is built, it reads massive amounts of digital information to learn human language, logic, and general facts. [1, 2]

• Public Web Content: Articles, blogs, digital books, and open-source code libraries gathered via web scrapers like Common Crawl. [1, 2, 3]
• Reference Sites & Communities: Widely used encyclopedias and forum discussions like Wikipedia and Reddit. [1, 2, 3]
• Commercial Data Partners: Information licensed directly from publishers, stock photo libraries (like Getty Images), and social platforms. [1, 2, 3]
• Human Trainers: Feedback and review data provided by hired human evaluators to help the AI learn what polite or accurate answers look like. [1, 2, 3]

2. Live Web Search (Retrieval-Augmented Generation / RAG)

Because training data has a “knowledge cutoff” (a past date after which the model stopped learning new things), AI tools often use live search to answer questions about recent events. [1, 2, 3]

• Instead of relying only on its frozen memory, the AI looks up live search engines (like Google or Bing) at the moment you ask a question, reads the top pages, and summarizes the answer. [1]

3. Proprietary and Enterprise Data

• Companies building private AI tools connect them directly to their own internal files, customer service logs, databases, or cloud storage to answer questions specific to their business. [1, 2, 3]

Opinions on public web data are mixed. Some users on Reddit point out that training on open internet forums can copy social biases or internet rumors, while others note that models use them mostly for casual or pop-culture language rather than hard facts. [1]

The question you may want to ask yourself – is this source of information the source you want to base your physical training on?

There are a few things AI can’t replace, and one of them is experience. Even better, experiences guided by experienced coaches.

In reality many end users and coaches will accept the regurgitation of dominant beliefs and methods as collated by AI.  For those who want more, the search will be harder as you probably won’t find this non-AI value on AI search engines!

Next month, November 2026, we are offering that very experience for end users and coaches that is aimed to provide what AI cannot.

AI is here to stay. You will decide to what extent you will use it to shape your physical training decisions.