I audited the 12 fastest-growing new GitHub repos for fake stars. Here's the base rate.
New repositories are where bought stars cluster, because the point of buying stars is to fake early traction. So I pointed an open-source auditor at a reproducible slice of GitHub: the most-starred public repos created i
New repositories are where bought stars cluster, because the point of buying stars is to fake early traction. So I pointed an open-source auditor at a reproducible slice of GitHub: the most-starred public repos created in the last 90 days.
Method
- Population: GitHub Search
created:>2026-02-26 stars:>1000, sorted by stars (684 matched). - Sample: top 12 by star count (20kβ192k stars).
- Tool:
fake-star-auditβ one Python file, standard library only, anonymous GitHub API. Five timing/identity axes plus extended signals. Every verdict prints the evidence behind it.
Result
| verdict | count | meaning |
|---|---|---|
| LOW | 9/12 | looks organic |
| MEDIUM | 3/12 | two signatures consistent with purchased stars |
| HIGH | 0/12 | β |
The 3 MEDIUM repos each tripped the same two axes: a page-1 sliding-window burst β a dense spike of early stars β and same-second star clusters: multiple stars in a single second, which is hard to do by hand.
These are signatures consistent with star buying, not proof. A MEDIUM means βworth a closer look,β not a verdict. I'm withholding repo names on purpose; the point is the method and the base rate, not an accusation.
Check any repo yourself
One file, no dependencies:
bash
curl -sO https://raw.githubusercontent.com/Armada735/fake-star-audit/main/audit.py
python3 audit.py --repo owner/name
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.