AI Disclosure

When you use Athlea’s coach, you are interacting with an AI system, not with a person. This page explains what that system does, what it cannot do, and how your data reaches the companies whose models we use.

Coaching guidance carries weight. If a plan tells you to run a hard session, or to hit a calorie target, you are more likely to do it than not — and you deserve to know that the thing recommending it is a language model working from your data, with the failure modes language models have.

Transparency about AI interaction is also becoming a legal requirement rather than a courtesy, and it is easier to build the habit now than to retrofit it.

These are not hypothetical risks. They are the normal behaviour of the technology, and knowing them changes how you should read the output.

• Confident and wrong. A model can produce a specific number, a citation, or a physiological explanation that is simply invented, in exactly the tone it uses when it is right. Fluency is not evidence.
• Out of date. Models have a training cutoff and may not reflect current sports science, current guidelines, or the current version of any list or standard.
• Missing context. The model sees the data it is given. It does not know about the injury you did not log, the medication you take, or how you actually felt this morning.
• Bad input, bad output. A mis-calibrated power meter, a watch that misreads sleep, or a mistyped weight will propagate straight through into a plan without anything flagging it as odd.
• Inconsistency. The same question can get different answers on different days. That is a property of the technology, not a sign that one of them is authoritative.
• Population bias. Training data under-represents women, older athletes, and people outside a fairly narrow athletic norm. Guidance may fit some people better than others, and menstrual cycle, menopause and masters-specific guidance deserve particular scepticism.

Prompts, guardrails and the boundaries of what the coaching assistants will say are designed and reviewed by people at Athlea. Individual conversations are not read live by a human, and no coach is monitoring your plan in real time — if you want a human coach in the loop, that is a different product.

You can always reach a person: email hello@athlea.ai.

To generate coaching, relevant parts of your training and health data are sent to a model provider acting as our processor. Our position is that providers may use it only to return a response to us, and not to train their own general-purpose models.

How the underlying data is handled generally is covered in the Privacy Policy.

Where Athlea answers from its research knowledge graph, the model is working over retrieved sources rather than memory, which makes invented facts less likely. Less likely is not impossible: retrieval can surface the wrong source, and a model can still misdescribe a paper it has in front of it.

Where a source is cited, follow the citation before treating the claim as established. Where no source is cited, treat the claim as the model’s summary rather than an evidenced finding.

Some of the photographic imagery of people on this website is generated by AI. The individuals shown are not real Athlea members, not real customers, and not real people at all. They are illustrations of a product experience, used the way a stock photograph would be.

• No AI-generated person on this site is presented as a customer, a member, or the source of a quote.
• AI-generated imagery is not used to depict a real named person, and is not used to suggest an endorsement that did not happen.
• Screenshots and interface imagery may be illustrative or simulated. Where a screen shows example data rather than real output, it says so on the page it appears on.

Any quote attributed to a named person on this site is from that person and given with their permission. Unattributed copy alongside an image is Athlea’s own writing, not a testimonial. Numbers and results shown in illustrations are examples unless they are explicitly presented as measured findings.

This matters legally as well as ethically: presenting invented endorsements as genuine is prohibited under UK consumer protection law, and the rules on fake reviews have recently been tightened.

If Athlea tells you something wrong — particularly something that could hurt you, or anything that phrases a supplement as permitted or safe — send it to hello@athlea.ai with enough detail to reproduce it. Bad output is a bug, and we would rather hear about it than not.