Almost everyone selling AI content advice right now learned marketing from AI.
They can tell you what a model does. They've never had a piece of work fail in front of an audience that had no obligation to stay — and that failure, repeated for years, is where every useful instinct I have came from.
I spent twenty-five years in broadcast and media before any of this existed. Here's what actually transfers.
Three seconds, one pass, no retargeting
People assume the useful part of a broadcast career is the creative. It isn't. It's the constraint.
In television you have about three seconds before someone looks away. You cannot A/B test your way out of a weak opening, and there is no retargeting a person who already changed the channel. You get one pass, in front of an audience that owes you nothing, and then it's gone.
Work under that constraint long enough and it rewires the question you ask about everything you make. You stop asking is this good? — a question about you — and start asking does this survive contact with someone who wasn't looking for it? — a question about them.
Most AI-generated content dies on that second question. It's fluent, structured, grammatical — and it opens with throat-clearing no channel-surfer would sit through. The model wrote it for a reader who promised to finish. That reader doesn't exist.
The constraint didn't go away. It moved.
The feed replaced the channel. The answer box replaced the listing. And the three seconds got shorter.
When a buyer asks ChatGPT a question, the engine assembles one answer from whatever sources it can extract a direct claim from — and everything it can't extract, it skips. Which means your content now plays to two audiences with identical patience: a human who'll look away in seconds, and a machine that discards anything it can't parse into an answer.
Broadcast discipline serves both, because it was never really about television. It was about earning attention you weren't owed:
- The first line does all the work. A tease that buried its hook died at 6:58pm. A page that buries its answer under four paragraphs of preamble dies in the extraction step. Same failure, new venue.
- One claim at a time. Broadcast writing puts one idea per sentence because the viewer can't rewind. Engines reward the identical structure — one clean claim per paragraph is what they can lift and cite.
- Proof inside the sentence. "Ratings went up" convinced no general manager. Numbers live inside the argument or they don't count — for a GM, for a buyer, or for a model deciding whether your page is a source or filler.
- Say it the same way everywhere. A station brand repeated its promise identically across every daypart for years. Engines do the equivalent of checking your dayparts — and when your site, your directories, and your profiles disagree about what you are, they hedge or leave you out.
What it looked like when it worked
This isn't nostalgia; the discipline produced numbers, repeatedly, in both eras.
I led a station group to all-time ratings records and three consecutive Station of the Year awards — attention won nightly, against a remote control. When the game moved online, the same mechanics moved with it: I grew a regional broadcaster's site from 500 to 50,000 monthly visitors in six months by doing exactly what a promo does — figure out the question, put the answer where the machine can grab it, get vouched for somewhere that isn't your own air. And I drove over two million impressions on a minimal media budget, because multi-channel content strategy is just programming a broadcast day across platforms that don't call themselves dayparts.
Different scoreboard every time. Same game.
Nobody owes you their attention.
The broadcast constraint, unchanged · how it transfersCampaigns are episodes. Systems are the series.
The deepest broadcast lesson isn't about writing at all.
Any station can produce one great night of television. The stations that set records are the ones with a rundown — a system that produces the show every single day, at broadcast standard, regardless of who called in sick. The episode is never the asset. The series is.
Marketing teams adopting AI are currently producing episodes: a great pilot here, an impressive one-off there, each living in somebody's chat history. The teams that will own their categories are building the series — encoded standards, a repeatable pipeline, output that compounds because it ships every week in the same voice with the same facts.
You don't win a daypart with one good show. You win it by being unmissable at the same time every night.
The part that converts
"Converts" in broadcast was never a metaphor. The promo either put viewers in the chair at ten o'clock or it didn't, and Nielsen told everyone by morning.
AI content that converts obeys the same physics. It opens with the thing the audience came for. It makes one provable claim at a time. It carries its receipts inside the sentences. It shows up identically everywhere the buyer — human or machine — might check. And it ships on a system, not on inspiration.
None of that requires a broadcast career. It requires accepting what broadcast forces on you from day one: nobody owes you their attention, and the work either survives contact or it doesn't.
The audience changed. The channel changed. The three seconds are still the three seconds.
Questions people actually ask
Isn't broadcast experience outdated for digital and AI marketing?
The tools aged out; the constraint didn't. Broadcast is decades of forced practice at winning attention nobody owes you, in one pass, with the result measured nightly. Every environment since — the feed, search, the AI answer box — has moved closer to that constraint, not further from it.
What's the single biggest broadcast habit worth stealing?
Write the first line as if it's the only line anyone will experience — because for most of your audience it is. Broadcast calls it the tease; the discipline is identical for a LinkedIn hook, an email subject line, or the first sentence a model extracts from your page. If line one doesn't earn line two, nothing else you wrote exists.
How does 'one claim per paragraph' help with AI engines specifically?
Engines build answers by extracting claims they can attribute cleanly. A paragraph that makes one verifiable statement is liftable; a paragraph that braids three ideas together forces the model to paraphrase or skip. Structuring for extraction is the same skill as writing for a viewer who can't rewind — which is why broadcast writers already do it.
Find out what the machines say about you. One page, forty-eight hours, the prompts your buyers actually ask — and exactly what came back, including who got named instead of you.
Get my free snapshot