Google Trend Archive: A Year-Long Archive of Real-Time Web Search Trends Worldwide

Dataset on Hugging Face: aurman/GoogleTrendArchive
DOI: 10.57967/hf/7531
License: CC-BY-4.0

A comprehensive archive of 7.6 million+ trending search instances from Google's Trending Now feature, collected continuously from November 28, 2024 to January 3, 2026 across 200+ countries and regions.

Unlike aggregated retrospective tools like Google Trends, Trending Now captures search queries experiencing real-time surges, offering unprecedented temporal granularity for studying collective attention dynamics.

---

Contents

File	Description
load_data_demo.py	Python demo: download and explore the dataset
load_data_demo.ipynb	Jupyter notebook version of the demo
Processing_CleanedUp_Commented.R	R script for preprocessing the raw daily CSV files
Datasheet_GoogleTrendArchive.pdf	Dataset datasheet (methodology, limitations, etc.)

The actual data files (daily_compressed.zip, googletrendarchive_preprocessed.csv, Trends_LocationList.csv) are hosted on Hugging Face due to their size (1.5 GB+).

---

Dataset at a Glance

Property	Value
Total instances	7,600,000+ trending searches
Temporal coverage	Nov 28, 2024 – Jan 3, 2026
Geographic coverage	1,358 countries and regions
Missing data	~14 days (technical collection gaps)
Format	CSV (UTF-8)

Fields per instance

Field	Description	Example
Trends	Search query / cluster label	"man united vs bodø/glimt"
Search volume	Bucketed traffic range	"50K+"
Started	Trend emergence timestamp (ISO)	"2024-12-01T14:23:00+00:00"
Ended	Trend end timestamp (ISO)	"2024-12-01T16:45:00+00:00"
Trend breakdown	Related query variations (comma-separated)	"man utd vs bodo, man united bodo glimt"
Explore link	Google Trends URL for this trend	"https://trends.google.com/..."

---

Quick Start

Python

pip install datasets pandas

from datasets import load_dataset

ds = load_dataset("aurman/GoogleTrendArchive", split="train", streaming=True)

for row in ds.take(5):
    print(row)

See load_data_demo.py or load_data_demo.ipynb for a full walkthrough including filtering by country, parsing timestamps, and plotting trend durations.

R

# Install once
install.packages(c("data.table", "lubridate", "ggplot2"))

See Processing_CleanedUp_Commented.R for the full preprocessing pipeline used to generate googletrendarchive_preprocessed.csv from the raw daily files.

---

---

Original Data Collection

See trendsccraper_public.py for the Python Playwright implementation of the original scraping code used to collect the raw daily files.

---

Use Cases

• Information Diffusion — track how topics cascade across regions
• Event Detection — identify breaking news and crises from search surges
• Comparative Cultural Studies — analyze collective attention across countries
• Crisis Communication — understand information needs during emergencies
• Temporal Pattern Analysis — daily, weekly, and seasonal rhythms
• Predictive Modeling — forecast trend emergence, duration, and spread
• Media Ecosystem Analysis — compare search trends with news/social media

---

Limitations

• Search volumes are bucketed, not exact counts
• Google's trend algorithm is proprietary (clustering decisions are opaque)
• ~14 days of missing data due to collection outages
• Trends represent relative surges, not absolute search volume
• Digital divide: internet penetration varies widely across regions

---

Citation

@dataset{urman2026googletrendarchive,
  title   = {Google Trend Archive: Global Real-Time Search Trends},
  author  = {Urman, Aleksandra and Hann{\'a}k, Anik{\'o} and Baumann, Joachim},
  year    = {2026},
  publisher = {Hugging Face},
  doi     = {10.57967/hf/7531},
  url     = {https://huggingface.co/datasets/aurman/GoogleTrendArchive}
}

---

Authors

Aleksandra Urman, Anikó Hannák, Joachim Baumann
Social Computing Group, University of Zurich & Stanford Artificial Intelligence Laboratory
Contact: urman@ifi.uzh.ch

Funding

• Swiss National Science Foundation – PostDoc Mobility fellowship P500-2 235328 (JB)
• SNSF Project Grant 215354 (AU and AH)