Every platform just told you the same thing: be a person

In the space of about three weeks this summer, five platforms that agree on almost nothing arrived at the same sentence. On July 31, Snapchat announced that wholly AI-generated videos would no longer be eligible for recommendation in Spotlight, explaining that it wanted the feed to remain “a place where people can discover authentic creativity from real people.” The day before, LinkedIn shipped a button, in the menu of every post, that reads “seems like AI slop,” and in the same breath retired its own “enhance your post” feature in favor of a proofreader that, in the company’s words, does not change your voice. A week earlier Substack handed readers a detector that scores any newsletter for machine authorship on demand. In mid-July YouTube clarified which videos its Partner Program will not pay for: the generic and repetitive kind that templates and generators produce at scale, the emotionally manipulative kind, and synthetic presenters discussing your money or your health. Instagram had already spent the spring removing recommendation reach from accounts that aggregate other people’s content.

Each of these was covered as a creator story, and each platform framed it as a matter of taste. Read together, they are something else: the first coordinated repricing of the human being on camera since the camera went into the phone.

What the platforms are protecting

We should be clear-eyed about motive. A feed drowning in synthetic video is a feed people leave, and a feed people leave is inventory nobody buys; the landlord guards the building, as we noted when YouTube doubled its bar two weeks ago. None of these companies has turned against machine-made content on principle. Every one of them drew the same careful line: the tool is fine, the author is not. Edit with AI, caption with AI, cut with AI, and you are still a person making a thing. Let the machine be the thing, and the distribution quietly disappears.

What is interesting is that the platforms are late. In one industry survey taken last winter, roughly four in five consumers said they trust video with real people in it, and about a third said that a video they judged to be AI-made lowered their opinion of the brand behind it. The audience decided this before the policy teams did. July was the month the policy teams caught up.

The platforms did not decide that a face is worth more than a render. Your customers did. The platforms are just the last to say it out loud.

The repricing, for a business

The economics are simple once stated. The cost of producing a video fell toward zero this year; a new photo-to-video generator launches most weeks, and any product shot can become a thirty-second commercial before lunch. When synthetic footage is infinite, it is also worthless, in the strict sense that nobody will pay attention to it. The scarce input, the one thing that cannot be generated at scale, is a verifiable person who knows the subject, looking into a lens and saying what they think. That was always the asset. It is now the only asset, and the platforms have started charging for the absence of it.

For a company selling something, this is less a warning than a reprieve. Most businesses never had a production budget to lose. What they had, and still have, is the founder who can explain the product in ninety seconds, the engineer who knows why it works, the account manager who has heard every objection. The reason so many of them reached for generated video anyway was not vanity. It was friction: the recording, the re-recording, the reading from a script without looking like you are reading from a script. The face was free. Getting it onto the screen was not.

The part we have a stake in

Here we should disclose an interest, because this is precisely the business StreamAgent is in, and you may discount what follows accordingly. We use machine intelligence liberally, for the work around the video: transcripts, the key moments in a watch curve, an assistant that answers which video is printing leads. We do not use it for the person in the frame, and the product is built on the assumption that you will not either. The studio runs in the browser. You write the script, and a teleprompter carries it up the glass at your pace, eyes near the lens, so a scripted take reads as a person talking. Camera, or camera with your screen shared mid-take, the way you would on a call. The take lands in your library ready to be the clip a branch plays, and because the clips a branching video needs are short, scripted, and single-message, the studio is the normal way they get made rather than a workaround for people without footage.

And the human on camera is not only a distribution tactic. It is the reason the rest of the machine works. A real person asking a real question inside the video gets answered; viewers do not confide in renders. The answer is the lead, the watch behavior and the choices around it become a score, and the score decides who gets called first on Monday. The face earns the trust. The platform turns the trust into a pipeline.

None of this says avoid the tools; the platforms did not say that either, and neither do we. Let the machines write captions, cut silences, and read your analytics. Then walk in front of the camera, because every feed you might want to appear in just agreed that this is the part they will reward, and because it is the part your customers were rewarding all along.

The machines can make the video. They cannot be the person in it.