Fake GitHub Star
Detector
Detect bot stars, paid stars, and coordinated fake-star campaigns on any public repo. Open methodology from dagster-io/fake-star-detector + BigQuery clustering. Free, no signup.
Free 30s scan first; only spends BigQuery $ if it sees something suspicious.
If anyone has analyzed this repo recently, we serve their result instantly.
Your analysis runs inside a Trusted Execution Environment. No data retention.
Built on Dagster's open research
Every heuristic on this site comes from dagster-io/fake-star-detector — an open-source Dagster project from Pete Hunt, Sandy Ryza, and the Dagster Labs team that first quantified the fake-star problem on GitHub in their 2023 investigation.
We're a thin hosted layer: GraphQL batching for speed, BigQuery cost guardrails, a public UI, and a shared cache. The detection logic — the per-account heuristic, the dbt clustering pipeline, the methodology — is theirs. If you find this useful, give them a star.
How fake GitHub stars work
Fake stars on GitHub come from two sources: bot networks (newly-created accounts that star and disappear) and paid star farms (clusters of accounts that move together across many repos in coordinated bursts). Both inflate a repo's apparent popularity and mislead potential users, contributors, investors, and hiring managers.
GitHub itself doesn't expose any "is this star fake?" signal. To answer it, you have to look at the profile of every account that starred the repo (followers, account age, activity), and ideally their full event history across GitHub to spot cluster patterns. That's expensive, which is why no public tool exists — until now.
Two detection models
Fast (per-account heuristic): Pulls every recent stargazer's profile via GitHub GraphQL API. Scores each against 8 hardcoded conditions ported from the open-source dagster-io/fake-star-detector: follower count, account age, bio/email/blog presence, and same-day account-creation-and-star pattern. Free, runs in 30 seconds for most repos.
Deep (behavioral clustering): Runs a 7-step dbt pipeline against GH Archivein BigQuery, pulling every action by every stargazer in a 365-day window. Detects two fake-star signatures: clusters of accounts that move together across many repos (the classic paid-farm signature), and accounts with near-zero non-star activity. Costs ~$5-15 per fresh repo (cached 30 days, served free after).
The fast model is precise but strict — it only flags the most obvious fakes. The deep model catches sophisticated farms that use aged accounts and backfilled activity. Use Smart mode to run fast first and auto-fire deep only if the fast scan shows ≥20% suspect.
Why this matters
In 2024, multiple high-profile GitHub repos were called out for buying stars to inflate traction. VCs, recruiters, and developers use star count as a popularity signal — making it economically rational for repo owners to fake. This tool gives you the data to verify before you draw conclusions.