What the chans are saying,
before it becomes the news.
chanalyse is a continuous 4chan discourse-monitoring and narrative-analysis engine. It scrapes /pol/, /biz/ and /g/ around the clock, clusters posts into stories, detects attention spikes against each story's baseline, classifies content with LLMs, and turns significant spikes into multi-perspective articles.
A research and observability system for studying online discourse — not a content aggregator or mirror. Continuous coverage with 100% board capture.
Corpus as of 2026-08-25. The scraper runs continuously; the published dashboard snapshot rebuilds roughly every 3 hours.
Latest articles
most recent per categoryEvery significant attention spike becomes a multi-perspective article — critical, neutral and supportive readings of the same story, with fact-checking against verified sources. Colours follow the globe: politics, technology, business.
Politics & Society
155Market calls & signals
/biz/ narratives → tickersWhen a /biz/ narrative spike fires, the engine distils it into directional calls — companies the discussion favours (long) and disfavours (short) — anchored to the article that produced them. Each row links to its source article.
Calls are LLM-generated interpretations of forum narrative sentiment, published for research transparency. They are not investment advice and carry no performance record.
Start here
The Data
The live dashboard: system health, board activity, top stories, themes and entities across /pol/, /biz/ and /g/ — plus the deeper Analysis views.
The Globe
News stories mapped onto a 3D globe by geographic origin, with recency gauges and a submarine-cable overlay showing where each narrative is being discussed.
The Articles
Multi-perspective coverage generated from significant attention spikes — each story written from critical, neutral and supportive angles, with an archive going back months.
Data for algorithm training
free · static · CORS-friendly
Every analysis view on this site is also published as a static, no-auth JSON API — roughly
8,061 baked endpoint files under /data/api/. With 4.5M+ classified
posts across 84k stories, the corpus suits training and benchmarking
narrative-detection, sentiment and
moderation-classification models.
- /data/api/analysis.json — full all-boards analysis snapshot (7d window)
- /data/api/analysis__board-pol__window-7d.json — per-board, per-window variant
- /data/api/activity-calendar.json — posting-activity calendar
- /data/api/coverage__window-7d.json — data-health / capture coverage
# grab a snapshot with curl
curl -s https://chanalyse.org/data/api/analysis__board-pol__window-7d.json | jq '.top_stories[0]'
# or with python
import json, urllib.request
d = json.load(urllib.request.urlopen(
"https://chanalyse.org/data/api/analysis__board-pol__window-7d.json"))
print(d["board_comparison"])
Licensing & attribution: the data is derived from public 4chan posts, machine-processed by chanalyse, and provided as-is for research purposes with no warranty. If you use it, attribute chanalyse.org (and note the underlying source is public forum content). Files are static snapshots refreshed on each publish.
Social media monitoring — how it works
- Continuous scraping of /pol/, /biz/ and /g/ — threads and posts captured as they appear (see Data health for capture transparency).
- Story clustering — posts are grouped into recurring stories and themes, tracked over time with mention counts and novelty scores.
- Spike detection — attention velocity is compared against each story's own baseline; statistically significant accelerations become spikes (905 detected so far).
- LLM classification — narratives, sentiment and stance are classified (18,071 classifier calls), powering the Analysis views.
- Authenticity & astroturf signals — coordinated-inauthenticity indicators surface on the 🕵️ Authenticity page, with moderation tracking on 🛡️ Moderation.
- Market signals — /biz/ narrative shifts are distilled into the 📈 Signals view and the calls above.
The pipeline
Spikes that cross the article threshold trigger multi-perspective article generation and appear on the globe; everything is baked to static HTML + JSON and deployed here.