AI Transparency Notice

1. You are interacting with AI

The "Ask ChatGPT" and "Ask Claude" advisers are large language models operated by third-party providers (OpenAI and Anthropic respectively), reached either through their own API when the operator has configured a key, or through the Lovable AI gateway. Every answer they produce is machine-generated and is labelled as such at the point where you read it, not only in this notice.

An adviser that is not configured says "Not configured". It never fabricates an answer to hide a missing provider.

2. What the AI may and may not do

FishingPals are grounded: they may only choose from and explain the candidate set we supply, together with your fishing profile. They cannot invent an offer, a price, a stock level or a retailer, and they cannot place an order, approve a purchase intent, verify a catch or award points.

Where a candidate came from the seeded demo catalogue rather than a live check, the interface says so and FishingPal's answer inherits that limitation.

3. What is authoritative

Deterministic systems, not the AI, decide the facts in FishSmash. Gear compatibility comes from an explainable rule engine. Prices, stock and specifications come from timestamped merchant verification of the retailer's own published data. Where an AI answer conflicts with the compatibility engine or with a verified check, the deterministic result wins and you should follow it.

4. Known limitations — please verify

Language models can be confidently wrong, can be out of date, and can miss local rules. Always verify independently before acting on an AI answer about:

• safety — ice, boats, wading, weather, electrical or chemical hazards;
• legality — licences, permits, closed seasons, size and bag limits, protected species;
• current retailer facts — price, availability, delivery, returns and warranty;
• product suitability for a specific rod, reel, line class or species where the compatibility engine has not confirmed it.

5. What is sent to the provider

Your question, your relevant profile answers and the candidate set are sent to the model provider. Your exact catch coordinates, your email address and your payment references are never sent. See the Privacy Notice for the legal basis and retention.

6. Feedback and reporting a bad answer

If an adviser gives you a wrong, unsafe or offensive answer, report it through the contact form using the "Content report" or "Support" category and include the question you asked. We review reports, adjust the grounding and, where necessary, disable a provider.

General EU guidance on AI transparency obligations is published at https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai.