The lesson
Publish a web summary of your research in a fixed Overview, Context, Execution, Results, Conclusion order, so a reader sees the method before the numbers.
An 'AI REPORT' label sits over a heading with 'LLM-as-a-Judge:' highlighted in green, a one-paragraph summary and an 'Access full report' button, followed by a tab bar for Overview, Context, Execution, Results and Conclusion. Context names two problems with manual review, 'Operational fatigue' and 'Reviewer fatigue'. Execution states the dataset as over 6,000 physician-led threads and 136,000+ messages, and diagrams how the Viral URI and Acute sinusitis judges are built. Results shows a bar chart of accuracy, precision, recall and F1 for the two judges, such as 96.6% and 81.5% accuracy, and a grid comparing human physician and LLM judge ratings from Excellent to Near miss.
Don't copy Counsel's clinical conditions, judge designs or performance figures. Keep the tabbed section order, the dataset size stated up front, one diagram per method, and results that show weaker cases next to strong ones.
Tell us what you want to make. We’ll talk through how to put these ideas to work for your brand.