This paper introduces the Universal Core Framework v1.0, a domain-agnostic methodological architecture for the systematic evaluation of AI-generated consumer guidance. As large language models are increasingly consulted for guidance in high-stakes consumer decision settings — regulatory navigation, jurisdiction-sensitive compliance, contractual self-advocacy, and other safety-critical contexts — existing AI benchmarks focused on factual recall and academic reasoning fail to assess whether such guidance is procedurally valid, safe, transparent, or jurisdictionally accurate. The framework defines six immutable evaluation dimensions (Accuracy, Completeness, Actionability, Safety, Jurisdiction Sensitivity, and Transparency), standardized study classifications, prompt complexity tiers, a universal study and data pipeline, dual-track scoring and calibration procedures, reproducibility and metadata standards, behavioral signature detection rules, claim-strength governance, and deviation accounting procedures. It deliberately separates the methodological architecture from any single empirical deployment, enabling cross-domain comparability and reproducible extension by independent researchers. A companion Reference Implementation applying this framework across multiple consumer domains and model ecosystems is reported separately (Walcher 2026b). This paper specifies and justifies the methodology; it does not claim to have validated it. The framework is introduced as an iteratively developed methodological architecture rather than a fully validated psychometric instrument, with empirical validation intended to emerge through subsequent Framework Validation Studies and Domain Extension Studies. AI evaluation, large language models, consumer guidance, reproducibility, evaluation methodology, AI safety, prompt-response evaluation.