Truace
Subscribe

Four AI chatbots were tested on 37 common parent questions about plagiocephaly

Study title: Quality, Transparency, and Readability of Artificial Intelligence Chatbots as Sources of Plagiocephaly Information: A Cross-Sectional Comparative Study

Researchers submitted 37 frequently searched plagiocephaly questions to four chatbots (ChatGPT 5.5, Gemini 3.5 Flash, Perplexity and Copilot) on August 18, 2026, producing 148 responses. They rated each response with DISCERN, EQIP, JAMA Benchmark and Global Quality Scale tools and with six readability indices.

Read the study abstract

Parents increasingly use artificial intelligence chatbots for information about plagiocephaly, but the quality and readability of their responses remain uncertain. This cross-sectional study compared 4 mainstream artificial intelligence chatbots using 37 high-frequency English queries identified through Google Trends for "plagiocephaly" worldwide over the previous 5 years. On August 18, 2026, the queries were submitted to ChatGPT 5.5, Gemini 3.5 Flash, Perplexity, and Copilot, generating 148 responses. Informati…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2254· cite & verify
Peer-reviewed study: Journal of Craniofacial Surgery
Trace impact readingContested
P 63The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 65The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

Parents are turning to chatbots for health information, so uneven quality and text pitched at a high reading level could limit how useful and trustworthy the answers are in practice. The study tested only English queries at a single point in time, and the source does not say how well the findings would hold for other models, languages or later versions.

Main points

  1. Perplexity had the highest median DISCERN score (50.00) and Gemini the lowest (38.00).
  2. Copilot had the highest median EQIP score (85.00) and Gemini the lowest (60.00).
  3. Median JAMA Benchmark scores were 1.00 for ChatGPT, Perplexity and Copilot and 0.00 for Gemini.
  4. Flesch-Kincaid Grade Level ranged from 11.66 to 13.14, and Flesch Reading Ease from 34.60 to 45.71, so no model met recommended readability thresholds.

The gain

Chatbots gave coherent, relatively complete plagiocephaly information, with a median Global Quality Scale score of 4.00 across all four models.

The problem

Chatbot plagiocephaly answers were written above recommended reading levels and offered limited source transparency, with quality varying widely between models.

The rundown

The study was cross-sectional and drew its 37 high-frequency English queries from Google Trends for the term plagiocephaly worldwide over the previous five years.

Between-model differences were tested with the Friedman test, Kendall W effect sizes and Holm-adjusted paired Wilcoxon signed-rank tests, and quality scores differed significantly among the models.

The authors conclude chatbot responses should supplement rather than replace professional assessment, and that patient-facing content needs professional review, source traceability and lower reading levels.

What this doesn’t fix

The study used only English queries submitted on a single date to four chatbot versions, so the findings reflect a snapshot of those models.

Sources

  1. Peer-reviewedJournal of Craniofacial Surgery2026-10-23

The debate