Chapter 11 — The People Who Aren’t People
You cannot count an enemy made of ghosts.
The last chapter ended on a hard turn, so here it is plainly: a great deal of what reaches you on a screen — in your feed, your inbox, your phone — was never sent by a person at all.
The principle
You can no longer assume that a “person” who contacts you is a person. The account, the face, the voice, even the video — any of them can now be manufactured. Authenticity, the thing you have leaned on your whole life without thinking about it, no longer comes for free.
The parallel — the ghost army
In the spring of 1944, a United States Army unit landed in Europe carrying a strange set of weapons: inflatable tanks. Trucks that broadcast recorded sound. Fake radio traffic. They were known, unofficially, as the Ghost Army, and their entire job was to make the enemy see a force that was not there — to pull enemy divisions toward an army made of rubber and noise.
It worked, because it has always worked. One of the oldest moves in war is to manufacture soldiers — to make your strength look larger than it is, or to make it appear where it is not, so the enemy spends his fear and his force in the wrong place.
The digital version of that trick is now pointed at you, in your living room. An adversary — a scammer, a propagandist, a foreign operator, it does not matter which — can fill your feed and your phone with “people” who do not exist. The purpose is exactly what it was in 1944: to make you believe a force is real, and move you because of it. You cannot count an enemy made of ghosts.
The layers of the fake
Start with the simplest layer: the bot, and the fake account. A great deal of internet traffic is automated — by some industry estimates, around half of it. I will be straight with you: that exact number is disputed, and the firms that publish it also sell the software to fight it. So do not anchor on a percentage. Anchor on the capability — a single operator, today, can run thousands of accounts. Which means the oldest signal you have, “everyone is saying this, so it must be true,” is broken. A handful of people can manufacture a crowd. A view held by almost no one can be dressed up to look like a tidal wave. That manufactured crowd has a name: astroturf — fake grassroots.
Now the layer that changed everything: artificial intelligence made the fakes good. It used to be you could spot the bot — the clumsy grammar, the reply that was slightly off. That tell is gone. AI now writes posts as fluent as your neighbor’s. It generates photographs of faces that have never belonged to anyone. And — the dangerous one — it clones voices. A short clip of a person’s voice, pulled from anywhere — a video, a voicemail, a podcast — is enough for an AI to make that person seem to say anything at all.
What it does to you
Here is that capability, walking up your front steps.
The phone rings. It is your grandchild’s voice — and the voice is crying. There has been an accident, or an arrest; they are frightened; they need money right now, and please do not tell their parents. Every instinct you own lights up. But it is not your grandchild. It is a cloned voice, built from a few seconds of audio, reading a script. The Federal Trade Commission has warned about this exact scam, by name.
And it scales far past families. In 2024, a finance worker at a large company joined a video call with what looked and sounded like his own senior colleagues — and, on their instruction, wired roughly twenty-five million dollars. Every person on that call except him was a deepfake. Romance scams run the same engine more slowly: a warm, attentive partner, built over months, who simply does not exist — and Americans report losing hundreds of millions of dollars a year to precisely that.
The reframe — stop detecting, start verifying
Now the most important thing in this chapter, and it cuts straight against your instinct.
Your instinct says: I will get better at spotting them. You will not. Nobody will. The fakes are improving faster than any human eye can track, and a year from now they will be better still. A defense built on “I will be able to tell” is already a defeated defense.
So you do not defend by detecting. You defend by verifying. You stop trying to judge whether the voice or the face or the message is real — that is the game you cannot win — and instead you move to a channel the faker does not control, and you check. An EOD technician does not decide a wire is safe by staring at it harder. He has a procedure. So, starting now, do you.
Make it actionable
DRILL — VERIFY, DON’T DETECT
The call-back. If “someone you know” contacts you with an urgent request, hang up — or close the message — and reach them yourself, on the number or account you already had for them. Not the number that just called you. The one you already trust.
The family code word. Sit down with your family and agree on a single word. If a panicked “relative” ever calls needing money fast, you ask for the word. A cloned voice does not know it. This one drill defeats the grandchild scam outright.
The second channel. Never act on a single channel alone. A call gets confirmed by a text to a known number; an email gets confirmed by a voice call. A faker almost never controls two.
The tell that never changes. AI can fake the voice and the face. It cannot change the shape of the ask. Urgency, plus secrecy, plus an unusual form of payment — a gift card, a wire, cryptocurrency — is a scam. Every time. The request itself is the tell.
Where this goes
The people who aren’t people use a fake human to move you. The next chapter is quieter, and in its way more common: you do not need a fake person to be tricked. Sometimes all it takes is a fake button. That is next.
When the systems designed to protect and defend against this kind of threat fail then resources like this Substack can educate and advise on strategies to prevent deceptive efforts from succeeding.