A protection desk that finds every fake page, copied reel and scam site using an institution's name, and hands staff a finished complaint for each one.

A large public institution with millions of followers was being copied faster than anyone could report it: impersonator accounts, re-uploaded reels, look-alike websites asking for payments, and misleading videos. We built the infrastructure that watches for all of it, proves what is theirs, and turns every finding into a ready-to-file complaint routed to the right platform, registrar or authority.

3,200+look-alike domains checked every night

24watch jobs running without anyone pressing a button

1 cardper problem: the route, the evidence and the filled-in complaint

The challenge

What we were up against

Fake accounts used the institution's registered logo, scam websites copied its official text and collected payments from followers, and edited reels spread faster than staff could find them. Each platform wanted a different form, different evidence and a different legal basis, so most problems were never reported at all, and the few that were often came back rejected.

Our approach

How we got from zero to live

01Registered every original photo and video with picture and sound fingerprints, so a cropped, flipped or re-voiced re-upload still matches its source.

02Set up early warning: newly registered domains, live certificate feeds, ads using the name, and nightly checks of around 3,200 look-alike spellings, each site scored on logo use, copied text, payment requests and claims to be official.

03Mapped every takedown route (Meta IP reports, platform forms, registrars, hosts, payment apps, domain disputes, cybercrime, legal notices) to the evidence and wording each one accepts.

04Added a paste-a-link investigation: AI reads the post, decides what kind of problem it is, and drafts the report, while staff keep the final say on every filing.

The solution

What we shipped

One case desk: every finding becomes a case with hashed evidence, the best reporting route picked for it, and a single 'Report it' card with every field already filled in.

Copy, impersonation and look-alike finders across video platforms, social accounts, search, reverse image search and new websites, scored and matched to the originals.

Scam-network linking: the same phone number, payment ID or site across fakes is joined into one stronger complaint, with a printable 'check before you pay' poster for the public.

Tracking to the outcome, follow-up reminders and escalations, a daily summary and a monthly report, with 24 scheduled jobs running the watch on their own.

Tech stack

PythonClaudeGeminiMeta IP ReportingYouTube Data APICertStreamffmpegTesseractHeadless ChromeGoogle Cloud Run

PreviousNuvero AI

NextSouthwest Gases

Every week you wait costs another 31.5 hours

Ready to run on AI infrastructure?

Book a 30-minute call. We'll map your workflows, show you which manual work disappears first, and tell you honestly whether we're the right fit. No slides, no sales theatre.

Book a 30-min call Read the infrastructure booklet

loading_agent…