AI Disclosure — ITCSCore

ITCSCore uses large language models and statistical models to generate match summaries, predictions, and editorial briefings. This page discloses what we use, how, and your control over AI-generated content.

1. Models Used

Primary text model: gpt-5.5 (via OpenAI-compatible gateway). Fallback chain: tier B (Grok-3) → tier C (Qwen-2.5) → tier D (DeepSeek-V3). Statistical model: Poisson goal-distribution baseline, version p0-v1. The specific model version used for any prediction is recorded in the prediction's metadata (visible in the admin audit log).

2. Provider

LLM access is via kanqiumei OpenAI-compatible gateway and direct OpenAI API. Providers process inference inputs subject to their respective privacy policies. We do not send personally identifying user information to providers.

3. Use

(a) Generating prediction probabilities (combined with statistical baseline). (b) Generating natural-language match briefings ("explanationMd"). (c) Generating SEO meta descriptions and editorial summaries. We do not use AI to generate medical, financial, or legal advice.

4. Limitations

(a) AI output may contain factual errors ("hallucinations"). We apply a fact-grounding score (minimum 90) to filter low-confidence briefs, but errors can still occur. (b) Predictions cannot account for events after publication (lineup changes, injuries during play). (c) Probabilities are estimates, not guarantees. (d) Models can exhibit bias from training data.

5. How to Use AI Output

(a) Treat as one of many information inputs. (b) Cross-reference with official league sources before drawing conclusions. (c) Do not use as betting advice. (d) Report errors to admin@itcscore.com.

6. Data Sources

Match data: API-Football (https://www.api-football.com) live and historical. Training data: providers train on broad public corpora; we do not fine-tune on user data. We do not feed personal data into models.

7. User Control

Cookie consent allows you to opt out of analytics that would track AI feature engagement. P1: a future setting will allow you to hide AI-generated content site-wide.

8. Visual AI Identification

Every AI-generated text on the Service carries an "AI" badge. Hover or tap the badge to see the model name, generation timestamp, and a link back to this disclosure.

9. Feedback

Report inaccuracies or concerns: admin@itcscore.com. We log all feedback and review weekly.

Effective Date: 2026-05-01