LLM Responses to Pediatric Depression FAQs: Readability, Accuracy, and Sentiment Scores

15 standardized pediatric depression FAQs were submitted to three large language models (ChatGPT-5, Microsoft Copilot GPT-5, DeepSeek 3.1V). Responses were scored for readability using seven indices, accuracy and completeness on a 0-6 scale, and sentiment. The dataset was created by RongQi Jiao and last updated on April 29, 2026.

Use Cases

• Benchmarking LLM readability for patient-facing materials based on Flesch Reading Ease and Flesch-Kincaid Grade Level scores
• Evaluating factual accuracy of AI-generated clinical guidance based on rubric scores from 0 to 6
• Comparing completeness of AI responses across different clinical topics (definition, causes, treatment)
• Analyzing sentiment tone of AI responses to mental health questions based on sentiment scores

Strengths

• Includes scores from seven established readability indices for each model
• Provides accuracy and completeness scores on a predefined 0 to 6 scale for 15 standardized questions
• Compares three contemporary LLMs (ChatGPT-5, Copilot GPT-5, DeepSeek 3.1V)

Limitations

• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download
• The dataset is small (90.6 KB), indicating limited scope

Provenance

Source

figshare

Collection Method

Cross-sectional analytical study; responses collected verbatim from LLMs and scored using predefined rubrics.

Freshness

Last updated 2026-04-29 05:57:40; freshness should be verified