TransparencIA receives an article as a URL, extracts its text and metadata, and runs it through an LLM-based pipeline that applies a journalistic quality rubric. It returns a composite quality score, a rationale for each of the eight criteria, a clickbait flag, a political-orientation label and a natural-language justification. Political orientation is kept out of the Quality Seal, so it never affects certification, and every assessment is stored with its evidence so any score can be audited and reproduced. The system is human-in-the-loop: it supports editors and fact-checkers and takes no editorial decisions on its own. Results are delivered through a web portal, a REST API with OpenAPI documentation and publishing on Telegram, X and Bluesky. A ten-month public deployment (October 2024 to August 2025) analysed more than 3 thousand real articles from different Spanish media outlets (TRL 7), and the system is now being extended to English-language press. The containerised stack is being split into five AI Builder modules (extraction, clickbait, quality, bias and sources) that other AIoD providers can reuse.