Our AI Prompts

We believe in radical transparency. We publish our core prompts and prompt annotations here, and continue expanding coverage. You can see exactly what instructions we give to AI models, how we prevent hallucinations, and what safeguards ensure accuracy.

30Total

14Active

16Planned

Quality Validation

Prompts that verify accuracy and consistency of generated content

Score Rationale JudgeVerifies that narrative rationale logically supports the assigned scores. Catches mismatches between glowing language and low scores.Active Citation Verification JudgeVerifies that citations actually support their claims. Checks URL accessibility and content accuracy.Active Factual Accuracy JudgeCross-references facts in narratives against source data. Catches hallucinated numbers, dates, and claims.Active Zakat Classification JudgeValidates zakat eligibility claims against source evidence. Prevents false zakat-eligible classifications.Active

Data Extraction

Prompts that extract structured data from charity websites

Homepage ExtractionExtracts mission, target populations, and founding year from charity homepages.Active About Page ExtractionExtracts detailed mission, vision, values, and organizational information.Active Programs Page ExtractionExtracts program descriptions and service areas from charity program pages.Active Impact Page ExtractionExtracts impact metrics, beneficiary counts, and outcome statistics.Active Donation Page ExtractionExtracts tax-deductibility status and payment methods from donation pages.Active Contact/Leadership ExtractionExtracts leadership team members, titles, and organizational structure.Active Zakat Page ExtractionExtracts zakat eligibility claims, asnaf categories served, and scholarly endorsements.Active Charity Navigator Financial ExtractorExtracts detailed financial metrics from Charity Navigator pages into structured schema.Active

Narrative Generation

Prompts that generate human-readable evaluations

Baseline Narrative GeneratorGenerates baseline evaluation narratives with Wikipedia-style citations. Core narrative seen by all users.Active Rich Narrative Generator (v2)Investment memo-style deep analysis with inline citations. Premium content for logged-in users.Active

Category Calibration

Cause-specific prompts that adjust scoring benchmarks (planned)