01Corpus
4,358 unique records
7,933 records became 4,358 after DOI-based and secondary title–author–year deduplication; 45.06% of the initial pool were duplicates.
BRAIN Journal Interactive Companion
The Post-ChatGPT Surge: Does AI-Health-Education Research Growth Reflect Evidence or Enthusiasm?
Production × impact2023
Selected year348 publications · 42.41 mean citations per article
Overview
A fast-growing field still consolidating its evidentiary core.
Three bibliographic databases were merged to trace publication volume, citation impact, authorship, sources, and thematic composition before and after ChatGPT’s November 2022 release.
02Production
56.48% annual growth
Output climbed from 228 publications in 2022 to 1,550 in 2025. The 2026 count of 1,367 covers only January through August 12.
03Impact
2023 citation peak
Mean citations per publication peaked at 42.41 in 2023, then fell across newer cohorts even after time normalisation.
Central finding
Neither purely evidence-driven nor purely hype-driven
Growth outpaced consolidation, yet a genuine tool-evaluation cluster paired named generative AI systems with accuracy and readability assessment. The landscape is best understood as transitional.
Interactive Search Lab
Build a database-ready Boolean search.
Choose a research purpose, database, and concept set. The generator applies database-specific syntax while keeping the article’s recommendation—specific generative AI terms combined with accuracy, readability, or validation—visible and explicit.
Factorial-analysis cluster terms combined
PubMed-ready query
("ChatGPT"[Title/Abstract] OR "GPT-4"[Title/Abstract]) AND (accuracy[Title/Abstract] OR validity[Title/Abstract]) AND ("patient education"[Title/Abstract] OR "health education"[Title/Abstract])
Specific GenAI terminology is combined with accuracy and an education context, matching the evaluative direction identified in the article.
Balanced groupsDatabase fields appliedEligibility filters active
Specific GenAI termArticle-supported evaluation termEducation context
What the generator changed
Production & Impact
Volume rose. Impact per publication moved the other way.
Select a year to compare output with raw and time-normalised citation performance. Citation indicators exclude PubMed-only records because PubMed did not provide citation data.
YearProductionImpact pattern
2022 baseline228 publications23.5 mean citations; 4.7 per citable year.
2023348 publications · +52.6%Peak impact: 42.41 mean citations; 10.6 per year.
2024–2025851 → 1,550 publicationsMean citations fell from 18.47 to 5.35.
2026 partial1,367 publications0.68 mean citations with one citable year.
Keyword Evolution
Generative AI language overtook machine learning in 2024.
Choose a cleaned author keyword to trace annual, non-cumulative mentions. The 2026 values cover January through August.
Annual non-cumulative mentions
ChatGPT
ChatGPT had 0 annual mentions in 2022 and 150 from January through August 2026.
Annual crossover
ChatGPT and large language model passed machine learning in 2024
2023 ML 45 · ChatGPT 35 · LLM 112024 ML 85 · ChatGPT 189 · LLM 1002026* ML 88 · ChatGPT 150 · LLM 309
Strategic Diagram
Prevalence did not equal thematic maturity.
Select a cluster to inspect how density and centrality locate it in the research landscape.
Niche themes
Motor themes
Emerging or declining themes
Basic themes
Relevance degree (centrality) →
Development degree (density) →
Co-word Structure
Two main keyword clusters, bridged by artificial intelligence.
Use the filters to isolate the two cluster memberships reported in the article. Node placement is schematic; the labels and cluster assignments follow the published findings.
Cluster 1Cluster 2Artificial intelligence: betweenness centrality 438.28
Analytical Pathway
From three databases to one deduplicated corpus.
Stage 01
Search three databases
Evidence Lens
What the landscape supports—and what it cannot resolve.
Switch between signals of substantive evaluation and signals of rapid, still-immature growth. Both sets are necessary to understand the transitional verdict.
Most defensible interpretation
A prolonged transitional state
Substantial conceptual and empirical work has begun, but publication growth has not yet been matched by equivalent consolidation of the field’s evidentiary core.