Independent Research
Observational LLM & Generative Engine Research
ASTONTO studies outputs that users can observe. We do not claim access to proprietary model internals or unreleased vendor algorithms.
Black-Box Evaluation Framework
Because commercial AI platform vendors do not publish real-time internal weights, empirical evaluation relies on controlled black-box testing. We isolate observable input signals, prompt variations, and output consistency across ChatGPT, Perplexity, Gemini, and Google AI Overviews.
Published Studies & Methodology Notes
Methodology2026-08-01
Black-Box Evaluation of Observable AI Recommendation Behaviours
An overview of ASTONTO’s methodology for measuring how large language models and AI-enabled search systems represent, compare and recommend organisations through observable public outputs.
Reliability: HighRead Study →