How Accurate Is Bella?

Bella’s Accuracy

Bella, our Human Design AI assistant, is deeply versed in the intricacies of the Human Design system and adept at interpreting charts. With ongoing improvements to our prompt engineering, chart-data presentation, and the underlying AI models, Bella now achieves an impressive 98–99% accuracy when reading Bodygraph data.

Why Is Accuracy So Impressive?

The Bodygraph contains over 1,000 different pieces of information — gates, channels, centres, profiles, lines, variables, and more. Bella processes all of this data with the deep Human Design knowledge already inside the AI models she runs on to provide accurate interpretations.

Multiple AI Models

You can choose which AI model powers Bella. Currently supported models: ChatGPT 5.6 Luna (the no-cost Auto default); Claude Haiku 4.5, ChatGPT 5.6 Terra, Claude Sonnet 5 and Claude Opus 5 (premium, usage-based).

• ChatGPT 5.6 – Fast everyday chat (Luna) and structured reasoning (Terra).
• Claude – Haiku 4.5, Sonnet 5 and Opus 5 for progressively deeper analysis.

Different models may provide slightly different perspectives, but all are calibrated to deliver accurate Human Design readings.

What If Bella Gets Something Wrong?

No AI system is flawless, and occasional discrepancies may occur. If you notice an error:

• Use the thumbs down button on any message to flag it. This helps our team improve Bella.
• Try rephrasing your question or providing more context.
• Submit a bug report through the Help Widget if the issue persists.

Continuous Improvement

Our team continually refines Bella’s prompts and how chart data is presented to each model, and evaluates response quality. We use anonymised feedback from thumbs up/down ratings to prioritise improvements.