The 2 prompts I never publish content without — and neither should you
Before anything goes live, I run it through two editorial layers: attention retention and audience trust. Both prompts in full, with an interactive walkthrough.
01 — The trap
Why is "seems fine" the most dangerous state of a text?
You finish a post. Re-read it. Seems fine. You publish — and silence. No reactions, no comments, no saves.
Sound familiar? "Seems fine" isn't an assessment of the text. It's a signal you've stopped seeing it through the reader's eyes. You know what you meant, so your brain fills in the gaps. The reader doesn't. They see only what's written — and leave at paragraph two if nothing holds them.
Genuinely bad texts get rewritten by their authors. It's the "seems fine" ones that get published — and they're what buries your reach. So I have an iron rule: no post, reel or story I have doubts about goes out without a run through two editorial layers. When I shared this approach on my channel, the post became one of the most viral I've ever published — apparently it's not just my pain.
02 — Two layers
What do these two editorial layers check?
A text fails for two independent reasons — and they need separate checks:
- Layer 1 — attention. The reader left because it got boring: filler, predictability, no hook. They don't "disagree" — they just can't be bothered to finish.
- Layer 2 — trust. The reader finished, but something clicked inside: "sounds too good," "marketing again," "this isn't about me." They won't buy, reply or save — even if it was interesting.
Finish → re-read → "seems fine" → publish → silence → "must be the algorithm."
Finish → prompt 1: where do I lose attention → prompt 2: where do I lose trust → rewrite the weak spots → publish a text that holds and doesn't repel.
Important: this is not "ask AI to improve my text." Improvement-by-default dissolves your voice. Both prompts work as a diagnostician: they show where and why the text weakens — the decisions stay yours.
03 — Prompt 1
How do you find the zones where a text loses attention?
Layer one — retention. The prompt, in full:
You are an expert in attention retention.
Here is the text: [paste your post, script, story, reel, email]
Break it down and show me where I lose the reader or viewer.
Find and describe the vulnerable zones where:
- attention drifts
- there is too much filler
- predictability kills interest
- there is no hook that would make someone keep reading/watching.
For each zone, specify:
1. What is going wrong here?
2. Why it hurts retention (name the cause: "the second paragraph sounds like a textbook intro", "no anchor in the middle of the story", "the rhythm collapses after a strong start").
3. What to change to bring the focus back.
Rewrite the problematic fragment using:
- hooks (questions, intrigue, contrasts, paradoxes),
- visual concreteness (images, specifics),
- emotional response (recognition, tension, curiosity).
Keep in mind this text is for [insert: Threads / a reel / Telegram / a newsletter etc.]; its job is to hold attention to the end and trigger a reaction (like, comment, full view, save).
The goal is not just to fix it, but to make the text work for the algorithms and for people at the same time.Note point 2: the prompt demands a cause ("the second paragraph sounds like a textbook intro"), not a verdict like "boring." The causes are what train you to see weak zones yourself — after a month of this, you'll catch them before the AI does.
04 — Prompt 2
How do you find where your audience stops believing you?
Layer two — trust. One hard requirement: this prompt needs a description of your audience. Without it, the AI judges your text through the eyes of "the average internet reader" — who has never bought anything from you and never will. Build your version:
The last three lines aren't politeness — they're protection. Without "keep my style," the model will happily rewrite you into smooth marketing tone — destroying exactly the thing people trust you for.
05 — The run
What does a pre-publish run actually look like?
The whole ritual takes 5–7 minutes. Walk through it on your next text:
Catch the "seems fine" moment — it's the trigger for the run, not for the publish button. Copy the full text, headline included: the headline is the first attention-loss zone.
Paste the text into prompt 1 (section above) and set the platform — Telegram, reel, newsletter. Read the breakdown and pick the 1–2 zones you agree with. Don't fix everything: you're the author, the AI is the diagnostician.
In a new message — prompt 2 with your audience description. Watch the "sounds too good" and "this isn't about me" items especially: those are the two main reasons a warm audience silently scrolls past.
Rewrite the weak spots in your own words — don't copy the AI's sentences verbatim. If the result "doesn't sound like you," roll it back: a live, slightly rough text beats a smooth borrowed one. Now publish.
And an honest warning from practice: after the run, I rewrite more often than not. Not because the text was bad — but because "seems fine" almost always hides 1–2 zones that would have cost reach.
06 — The system
What should you check before publishing any text?
To make the two layers a habit rather than a feat, keep a short checklist. Tick it right here — progress is saved, and you can download the list:
- The headline raises a "what happens next?" — it doesn't summarize the post
- Run through prompt 1: attention-loss zones found and rewritten
- Run through prompt 2: audience distrust points closed with specifics
- The text has at least one concrete detail: a number, an example, a case
- It sounds like me — the edits didn't eat my style
- I know what reaction I expect: a comment, a save, a reply
The two prompts aren't magic — they're an editing system: an attention layer plus a trust layer. It takes 5–7 minutes and usually ends in a rewrite — which is exactly why it works. Grab both prompts, run your next text — and compare the reactions.
FAQ
Does this work only for posts, or for reels too?
For any format with text: posts, reel scripts, stories, newsletters, even landing pages. Prompt 1 has a platform field — set it, because retention works differently in a reel and in an email, and the breakdown will account for that.
Why not just ask AI to "improve the text"?
Because "improve" dissolves your voice: the model rewrites everything into a smooth, averaged tone. These prompts act as a diagnostician — showing where and why the text weakens, with causes. The fixes and phrasing stay yours, so your style survives.
What do I put in the audience description if I barely know it?
Start with three lines: who they are, what they've tried, what they're skeptical of. Even a rough description beats an empty one — the model stops judging your text as "the average reader." Meanwhile, collect real phrasing: quotes from comments and client messages.
Isn't running every text through two prompts too slow?
The run takes 5–7 minutes. Post-publish silence costs more: the topic is burned and nobody saw the post. Plus, after a couple of weeks you start seeing the weak zones while writing — the prompts gradually train your own editorial eye.
