Pipeline to enable guideline question identification from GoogleTrends with artificial intelligence
Abstract
Objectives: For specific topics or keywords, Google Trends lists the most common related search queries performed by users. Although these queries may reflect public information needs from a public health perspective, they are rarely directly usable for clinical guideline development. We aimed to develop a reproducible pipeline that transforms Google Trends queries into structured guideline questions and other outputs for guideline development and public health surveillance.
Methods: The pipeline has five stages: (1) collection of Google Trends queries and metadata for keywords, location, and time period; (2) cleaning, normalization, deduplication, and noise removal; (3) identification of questions using rule-based keywords or large language models (LLMs); (4) classification into symptoms/diagnosis, treatment, safety, prevention, and prognosis; (5) extraction of PICO elements and generation of guideline questions. The system supports multiple LLM providers, including cloud and local Ollama inference, with typed outputs and CSV export for reproducibility.
Results: The pipeline produced two outputs: a classified question dataset and a PICO dataset. It enabled comparison between rule-based and LLM-based identification, with higher recall for LLMs. The system consistently transformed unstructured search queries into structured guideline questions relevant for evidence synthesis and public health monitoring.
Conclusions: This pipeline translates population search behavior into structured guideline questions. From a public health perspective, it enables insight into information needs and may support surveillance of emerging concerns. PICO extraction helps identify relevant populations and outcomes in a patient-centered way. The modular design ensures transparency and reproducibility and can be extended to social media and other infodemiology sources for public health intelligence applications. It supports timely evidence-informed decision making in public health.
Key messages: • AI pipeline converts Google Trends queries into structured, PICO-based guideline questions for public health use.
• LLMs outperform rules in detecting questions, enabling reproducible, scalable insights into population health information needs and surveillance.
Related
- Completeness and guideline concordance of artificial intelligence chatbot responses to patient questions on regenerative therapies for erectile dysfunction: a multi-platform analysis.
- Artificial Intelligence‐Optimized Image Biomarker Identification for Sinonasal Inverted Papilloma Diagnosis Using Image‐Enhanced Versus White‐Light Endoscopy
- Identification of adventitious lung sounds in children with respiratory illnesses in Bangladesh using artificial intelligence and digital auscultation
- Artificial intelligence-driven identification and mechanistic exploration of synergistic anti-aging compounds from Dengzhan Shengmai formulation
- Artificial Intelligence-Based Modeling to Aid in the Identification of Genetic Conditions in Infants with Congenital Heart Disease
BibTeX
@article{makowski2026pipeline,
title = {{Pipeline to enable guideline question identification from GoogleTrends with artificial intelligence}},
author = {T Makowski and W Neuman and B Sousa-Pinto and M Marques-Cruz and R Vieira and P Jemioło},
journal = {The European Journal of Public Health},
year = {2026},
month = oct,
doi = {10.1093/eurpub/ckag169.1236},
url = {https://doi.org/10.1093/eurpub/ckag169.1236},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - Pipeline to enable guideline question identification from GoogleTrends with artificial intelligence
AU - T Makowski
AU - W Neuman
AU - B Sousa-Pinto
AU - M Marques-Cruz
AU - R Vieira
AU - P Jemioło
PY - 2026
DA - 2026/10/01
JO - The European Journal of Public Health
DO - 10.1093/eurpub/ckag169.1236
UR - https://doi.org/10.1093/eurpub/ckag169.1236
AB - Objectives: For specific topics or keywords, Google Trends lists the most common related search queries performed by users. Although these queries may reflect public information needs from a public health perspective, they are rarely directly usable for clinical guideline development. We aimed to develop a reproducible pipeline that transforms Google Trends queries into structured guideline questions and other outputs for guideline development and public health surveillance. Methods: The pipeline has five stages: (1) collection of Google Trends queries and metadata for keywords, location, and time period; (2) cleaning, normalization, deduplication, and noise removal; (3) identification of questions using rule-based keywords or large language models (LLMs); (4) classification into symptoms/diagnosis, treatment, safety, prevention, and prognosis; (5) extraction of PICO elements and generation of guideline questions. The system supports multiple LLM providers, including cloud and local Ollama inference, with typed outputs and CSV export for reproducibility. Results: The pipeline produced two outputs: a classified question dataset and a PICO dataset. It enabled comparison between rule-based and LLM-based identification, with higher recall for LLMs. The system consistently transformed unstructured search queries into structured guideline questions relevant for evidence synthesis and public health monitoring. Conclusions: This pipeline translates population search behavior into structured guideline questions. From a public health perspective, it enables insight into information needs and may support surveillance of emerging concerns. PICO extraction helps identify relevant populations and outcomes in a patient-centered way. The modular design ensures transparency and reproducibility and can be extended to social media and other infodemiology sources for public health intelligence applications. It supports timely evidence-informed decision making in public health. Key messages: • AI pipeline converts Google Trends queries into structured, PICO-based guideline questions for public health use. • LLMs outperform rules in detecting questions, enabling reproducible, scalable insights into population health information needs and surveillance.
SN - 1101-1262
SN - 1464-360X
ER -