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 (20k–192k stars).
- Tool:
fake-star-audit— 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 | — |
The 3 MEDIUM repos each tripped the same two axes: a page-1 sliding-window burst — a dense spike of early stars — 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 “worth a closer look,” 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/ardev-lab/fake-star-audit/main/audit.py
python3 audit.py --repo owner/name
