How we know what
we tell you.
This page exists so a skeptical person, or a compliance officer, can check our method without watching a single video.
Where material comes from
We use four tiers of source, in this order of preference.
- Federal agency publications. FBI IC3 alerts, FTC consumer notices and enforcement actions, U.S. Postal Inspection Service releases, Social Security Administration OIG notices, CISA advisories. Material created by federal employees as part of their official duties is generally not protected by U.S. copyright. That is not true of everything inside a federal publication: contractor content, licensed photographs, third-party exhibits, seals and trademarks can retain protection, and we check those separately. We cite the document identifier and date on screen.
- Federal court records. Public prosecutions, used when we need a real, sourced figure for what a scam actually cost someone. We never estimate a loss.
- Viewer submissions, used only with explicit permission and only after identifying detail is redacted and reviewed under our published process before anything is shown. See the submission policy below.
- Our own collection. We operate accounts that exist solely to receive scam messages, so that we are looking at what is actually circulating rather than what was circulating a year ago.
How a claim gets on screen
- Every factual claim carries a source, named on screen while the claim is being made.
- A dollar figure presented as fact — what a scam cost, what an agency recorded — is only ever one we can point to in a source. If we cannot source it, the script changes, the number does not. An amount inside a labeled illustrative example is a different thing, and is never attributed to a real person, case or source.
- We never publish a working link to a scam, or a number that can be tapped. The address appears on screen because it is the lesson, never anywhere it can be followed.
What is on screen, in three categories
A viewer should never have to guess which of these they are looking at, so the screen says so for as long as the material is visible.
- Reproduction. A real message — agency-published, or submitted to us with permission — shown verbatim, including typos, spacing and broken formatting. Those errors are frequently the giveaway; cleaning them up would delete the evidence.
- Reconstruction · exact wording. A real message rebuilt so it can be read on a phone screen. The words are unchanged; only the rendering is ours.
- Illustrative example. A message assembled from tactics an agency has documented, when no real artifact can be shown without exposing someone. Nothing in it is a claim about a real person or a real case, and any amount inside it is illustrative.
How we use AI, in full
We use it, we are specific about where, and nothing it produces reaches you without a human checking it first. We can promise you the check. No process catches every mistake, which is why the corrections log exists and why every claim carries a source you can go and read yourself.
Where AI is used
- Scripts. Drafted with AI assistance. Every material factual claim is then checked against its cited primary source, and the Owner reviews and approves the script before anything is recorded. AI never gets the last word, and no claim reaches a video because software produced it.
- Narration. A licensed synthetic voice. The same voice every time.
- Captions. Generated from the approved script — the words we wrote and checked — never from a transcription of the audio. Speech recognition is used the other way round: to audit the finished narration against that script and flag any drift for a human.
- Research assistance. Gathering and summarizing public agency material.
Where it is not
- No AI-generated people. No synthetic presenter, no generated faces.
- No evidence presented as something it is not. An artifact is either the real thing, or a reconstruction built from tactics a named agency has documented — and a reconstruction is labeled as one, on screen, for as long as it is shown. We do not invent a victim, and we never present an illustrative detail as a fact about a real person or a real case.
- No unreviewed publication. Every material factual claim is checked against its cited primary source before publication, and the Owner watches and approves every finished video before it is released.
- No AI-invented facts. Claims are checked against the primary source, not against a model's recollection of it.
Every video is flagged with the platform's altered-or-synthetic-content setting where one exists, and the disclosure appears in every description.
Viewer submissions
If you send us something, you decide how it may be used — with one limit we will not pretend away.
- Sending us something is not permission to publish it. We ask separately before anything is used, as two questions rather than one: our own channels, and material licensed to institutions. You can agree to the first and decline the second.
- Names, account numbers, card digits, addresses and phone numbers are redacted before a frame is rendered, and checked against the rendered frame rather than the source file. We redact with solid blocks, never blurs, because blurs have been reversed. We do not claim to catch everything, and we ask you to tell us what we have missed.
- We publish submissions anonymously unless you explicitly ask to be credited.
- You can withdraw permission at any time by emailing submit@decoysignal.com. Within seven days we stop using the material and remove it from this site and from videos on channels we control. We cannot recall a copy already delivered to a licensee or already downloaded by someone else, and we will tell you honestly which applies. The full process is on the submission page.
- Sending something does not mean we will use it. That is about editorial fit, not about you.
Money, and who pays us
- Affiliate links. We have none today. If we take one on, the disclosure will sit beside the link itself and in the video description, at the point of the recommendation rather than on a policy page.
- We do not accept payment to review or recommend a product. An affiliate relationship never buys a mention.
- Licensing. Institutions pay to use our content with their members. They do not get editorial control, and they cannot commission a video that says something we would not otherwise say.
- We will always tell you when we are being paid in connection with something we are showing you.
Corrections
We will get things wrong. When we do, we correct it publicly and say what changed. See the corrections log.