Methodology

We believe anti-fraud platforms should be as transparent as the fraudsters are opaque. Here is exactly how a report becomes a published scam alert on GACS — and exactly how every score and color on this site is calculated.

1. Report intake

Anyone — victim or witness — can file a report at gacs.app/report. We collect: the suspect identifier (URL, wallet, phone, handle), category, evidence (screenshots, transaction IDs, chat logs), the reporter's relationship to the case, and timestamps. Reporter identity is optional and never published.

2. Triage

Every report is auto-triaged using deterministic checks: blockchain analysis of wallet flow, URL-pattern matching against known scam kits, similarity to existing entries in our database, and language models flagging classic pig-butchering / recovery-scam phrasing.

3. Human moderation

Reports flagged as plausible enter a human moderation queue. Moderators verify evidence is internally consistent, look for duplicate reports (a strong signal of legitimacy), and dismiss reports that appear to be retaliatory or competitive in nature.

4. The risk score, in plain English

Every public scam entry shows a single 0–100 risk number. It is deterministic — no randomness, no AI hallucination, no time-of-day jitter. The exact formula lives in src/lib/scam-risk-score.ts and is open for anyone to audit. Here is what feeds it:

+40Entry exists in the GACS scam database. Baseline. We don't publish a row unless triage + moderation cleared it.
+30On an active blacklist. The single strongest signal. Confirmed by our team or a partner feed.
+6 to +15Severity (medium / high / critical). Medium adds 6, High adds 12, Critical adds 15. Severity reflects financial loss, sophistication, and whether the operation is still active.
up to +15Community reports (diminishing returns). A logarithmic curve so 100 reports doesn't drown out a careful 5-report case. Cap is 15 points.
+5Active 30+ days. Persistence means the operation is still running, not a one-off.
+5Confirmed by an external feed. Independent corroboration from a non-community source.

The total is clamped to 0–100 and bucketed into a band. We also publish a 1–5 trust rating (inverse: 100 risk = 1 star, 0 risk = 5 stars) so search engines can render it as a star rating.

5. What the colors and bands mean

Colors on GACS are not decorative. They follow a strict semantic scale so a user can read a verdict at a glance, without parsing text.

Safe / Verified clean (green)

No risk signals after a full scan. Green is reserved for verdicts where the scanner actively cleared the target — it is never used decoratively elsewhere.

Informational / Trust (indigo)

Indigo is our trust color. It marks the GACS identity, primary actions ("Scan", "Report"), and informational states. Indigo on its own never means "safe" — that is green's job. Indigo means "this is the GACS voice talking."

Unverified (below 40)

A row exists but signals are sparse. Treat as "unknown," not "safe."

Caution (40 – 59)

Single-source or thin reports. Worth pausing on; not enough to publish a hard warning.

Elevated (60 – 74)

Multiple independent signals. Treat as likely fraud.

High (75 – 89)

Confirmed or heavy community reports. Do not engage.

Critical (90 – 100)

Active blacklist plus multiple corroborating signals. Active threat.

6. Category classification

Every entry is tagged with one of: fake-websites, fake-brokers, rugpull-tokens, memecoin-scams, social-impersonation, phone-scams, wallet-drainers, recovery-scams, or pig-butchering. Categories are visible at /scams/c.

7. Publication

Once approved, the entry gets a permanent URL at /scam/<slug>, appears in the live feed, and is pushed to subscribers via push, email, and RSS within minutes.

8. Disputes & corrections

If you believe an entry is incorrect, email disputes@gacs.app with evidence. We respond within 72 hours. Confirmed errors result in retraction or correction with a public changelog entry.

9. What we don't claim

GACS aggregates community intelligence. We do not claim every entry constitutes legal proof of fraud — only that enough credible signals exist that you should not send money. Always verify independently before making financial decisions.