ChatGPT as a Digital Pharmacist: A Systematic Review and Meta-Analysis of Drug-Counselling Accuracy
Bibliographic record
Abstract
Abstract Background The emergence of Large Language Models (LLMs) like ChatGPT presents significant opportunities for healthcare, yet raises concerns about accuracy, especially in high-risk areas such as medication counseling. A comprehensive evaluation of ChatGPT’s reliability in providing drug information is crucial for its safe integration into clinical practice. This systematic review and meta-analysis aimed to assess the accuracy of drug-counseling information provided by ChatGPT 4. Methods Following PRISMA, we systematically searched PubMed, Embase, Scopus, and Web of Science on May 9, 2025, for original research evaluating the accuracy of ChatGPT (version 4 or newer) in drug-counseling queries. Included studies compared the AI’s output against standard comparators like pharmacists or drug databases. A random-effects meta-analysis was performed to calculate the pooled proportion of accurate responses, and study quality was assessed using a customized Newcastle-Ottawa Scale (NOS). Results The search identified 17 eligible studies. Of these, 15 were included in the meta-analysis, which showed a pooled accuracy rate of 86% (95% CI: 0.75–0.95). However, significant heterogeneity was observed across studies (I2=98.5%, p<0.0001). Quality of the studies was a concern, with only four studies (24%) rated as high quality. No evidence of publication bias was found (p=0.91). Conclusion ChatGPT demonstrates substantial promise in drug counseling, with an 86% accuracy rate that surpasses its performance in other medical domains. However, the high heterogeneity and a non-trivial 14% error rate, coupled with methodological weaknesses in the primary literature, indicate that ChatGPT is not yet ready for autonomous clinical use. Its current role should be as a supplementary tool under the strict supervision of qualified healthcare professionals to ensure patient safety.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.044 | 0.134 |
| Meta-epidemiology (narrow) | 0.003 | 0.002 |
| Meta-epidemiology (broad) | 0.017 | 0.043 |
| Bibliometrics | 0.008 | 0.009 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.005 | 0.004 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.004 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".