AI Methodology & Limitations
Our pipeline: Claude models for analysis, research and translation, one screenshot reader, native specialist tools, and what our checks do and do not catch.
Quick Answer
Stock Expert AI's text is written by Anthropic's Claude models; native specialist tools handle safety screening, embeddings, images, raw video and audio, and speech. Scores are computed by published rules, not by a language model.
As of October 10, 2026, Stock Expert AI uses Claude Sonnet 5.5 for Moon AI and standard long-form text; Claude Haiku 5.5 for short interface translation; Claude Opus 5.5 for source-bound research, dossiers, portfolio analysis and review; Claude Haiku 5.5 for extraction and quick summaries; and one Claude Sonnet 5.5 reader for portfolio screenshots. All runtime LLM inference uses the shared Anthropic API. Other AI providers are disabled; images use licensed assets or local graphics. Dated research and earlier dossiers retain the model attribution stored on each document. Translated pages pass a 6-layer fidelity check against the English source. AI scores do not predict future performance; financial decisions should be made with a licensed advisor.