The Content Engine

The bottleneck in content is not writing. It's having something to say. So the economics change completely once you stop generating posts and start mining one real idea for everything it contains.

A single genuine insight (something you learned, a mistake you made, a question three customers asked this month) supports a long-form post, four short-form videos, an email, three social posts taking different angles, a carousel and a talking point for calls. Not by rewording it twelve times, which is obvious and boring, but by taking different angles: the story version, the contrarian version, the how-to version, the mistake version, the question version.

The source material has to be real. AI can angle, structure and draft. It cannot supply the thing you actually noticed this week, and content built on nothing you noticed reads exactly like content built on nothing you noticed.

Here's one real thing I learned recently:

[3-5 sentences. Specific. Include what you believed before and what changed.]

Mine it for a week of content. Give me:
- 1 long-form post (400-600 words)
- 4 short-form video hooks + 30-second scripts
- 3 social posts, each taking a DIFFERENT angle (story / contrarian / practical)
- 1 email to my list
- 1 talking point I can use in a live conversation

Rules:
- Different angles, not reworded versions. If two pieces make the same argument, replace one.
- Stay inside what I actually said. Do not add statistics, studies, or claims I didn't make.
- Match this voice profile: [paste voice profile]

Every piece of content has exactly one job in its first line: buy the second line. Nothing else matters if that fails, and most content fails there.

What reliably doesn't work: rhetorical questions ("Ever feel overwhelmed?"), throat-clearing ("In today's fast-paced world"), announcing your topic ("Let's talk about stress"), and the statistic opener that every account in your niche is also using this week.

What tends to work is specificity that implies a story. A number that's oddly precise. A confession. A flat contradiction of something the reader believes. The middle of a scene. "I lost $4,000 testing this so you don't have to" works because it's specific and it costs the writer something.

Never ask for one hook. Ask for fifteen. Hooks are cheap to generate and expensive to evaluate. You want to be choosing, not accepting.

Write 15 opening lines for this content:

[paste the content]

Requirements:
- Each under 12 words
- Each must make a specific promise or create a specific tension
- Vary the mechanism: confession, oddly-specific number, contradiction, mid-scene, direct address, uncomfortable truth

BANNED:
- Rhetorical questions
- "In today's world" / "Let's talk about" / "Here's the thing"
- Generic statistics
- Anything that could open a post in any other industry

After the 15, pick the 3 strongest and explain in one line each why that one earns the second line.

AI-written video scripts fail in a specific, recognizable way: they're written to be read, not said. Balanced clauses, no interruptions, no fragments, every sentence the same length. It sounds like a press release performed by a human.

Fix it in the constraints. Demand short sentences. Demand fragments. Ban any sentence over fifteen words. Tell it to write the way people actually talk, including the small inefficiencies: "and honestly," "look," a sentence that restarts.

Structure for 30 seconds is roughly: hook (2s), the tension or stakes (5s), the turn or insight (15s), one concrete takeaway (5s), a soft ask (3s). Don't over-engineer beyond that.

Two rules that matter more than the script: read it out loud before you shoot (anything you stumble on gets cut) and don't read it on camera. Learn the beats, then talk. The script is a map, not a teleprompter.

Write a 30-second video script.

TOPIC: [topic]
AUDIENCE: [who's watching and what they're skeptical of]
GOAL: [what they should do or feel]

STRUCTURE:
- Hook (2 sec)
- Stakes / tension (5 sec)
- The turn: the actual insight (15 sec)
- One concrete takeaway (5 sec)
- Soft ask (3 sec)

WRITE IT TO BE SPOKEN, NOT READ:
- No sentence over 15 words
- Use fragments
- Include at least two places a real person would break rhythm
- No list-of-three constructions
- No "in this video I'm going to"

Mark the timing beats in brackets. Then tell me which single line I should memorize and which parts I should improvise.

Sequences fail because every email is a pitch. The reader learns by email two that opening yields nothing but a sales page, and stops.

The fix is structural: each email must be worth reading even if the recipient never buys. One idea per email, one job per email. A working five-email shape:

1. Deliver something useful immediately, no ask.
2. A story that reframes the problem.
3. Address the objection they haven't voiced yet.
4. Proof: a specific example, not a testimonial wall.
5. The direct offer, short, with a clear reason to act now that isn't a fake countdown.

Subject lines: generate twenty, use two. Same economics as hooks.

And the rule that keeps you out of trouble: everything you write about results has to be true and typical, or clearly labeled as neither. The next lesson, and the compliance lesson in the income module, cover why this matters legally. It also matters practically: one oversold email costs you the list.

Write a 5-email follow-up sequence.

OFFER: [what you're selling]
AUDIENCE: [who, and what they're skeptical about]
VOICE: [paste voice profile]

SEQUENCE SHAPE:
1. Pure value, no ask
2. Story that reframes the problem
3. Handle the unspoken objection
4. Specific proof (an example, not a testimonial wall)
5. Direct offer, short, with a real reason to act now

RULES:
- Under 200 words each
- One idea per email
- Each must be worth reading even if they never buy
- No fake scarcity, no countdown timers, no "limited spots" unless I tell you the limit is real
- NO income claims, earnings figures, or health/disease claims of any kind. If the offer needs one to work, tell me instead of writing it.

Then give me 20 subject lines and mark your top 3.

This is the most important lesson in this module, and the one most courses skip. AI will cheerfully write things that get you sued, sanctioned or removed from your company.

It will invent a testimonial if you ask for "a customer quote." It will write "most people see results in two weeks" with no data behind it. It will write "clinically proven" about something that isn't. It will produce an income example that reads as typical when it isn't. It does this because those sentences are plausible, and plausible is what it optimizes for. It has no idea you're regulated.

Four hard lines. Put them in your project instructions so they apply everywhere:

• Never a fabricated testimonial or review. Not "for illustration," not "as a placeholder." Real words from real people or nothing.
• Never an income or earnings claim you can't substantiate and label with typical results.
• Never a health claim that a product treats, cures, prevents or mitigates a disease. Wellness language and medical claims are different categories, and the second one is regulated.
• Never a statistic you haven't personally verified at a source you could show someone.

Disclosure is separate and simpler: if a reasonable person would want to know AI was involved (in a testimonial, an image of a person, a "personal" message), tell them. The cost of disclosing is small. The cost of being caught not disclosing is your credibility, permanently.

Audit this content for claims that could get me in trouble.

CONTENT: [paste]

CONTEXT: [your situation, e.g. independent distributor, freelancer, coach, small business owner, plus your industry]. I am subject to FTC rules on endorsements and health claims and to [any company or industry policies that apply to me].

Flag every instance of:
1. Income or earnings claims, including implied ones ("financial freedom," lifestyle imagery, "replace your job")
2. Health claims: anything suggesting a product treats, cures, prevents, diagnoses, or mitigates a disease or condition
3. Testimonials or results presented as typical without substantiation
4. Statistics or study references I'd need to verify
5. Anything that needs a disclosure and doesn't have one

For each: quote the exact phrase, say why it's a problem, and offer a compliant rewrite that keeps the persuasive intent.

Be conservative. Flag borderline cases and let me decide.