Four prompts we use to prepare Threads posts

• Threads
• Prompts
• ≈ 10 minutes

Four ready-made prompts and the order in which they work.

One prompt won't solve the task. A post is built from four steps, and at each one the neural network needs to know something different: who you're writing to, what makes you different, what to talk about, and in what voice.

Asking to "write a post for Threads" gives you a text that could be handed to any author. The neural network knows neither your audience nor your position, so it writes something average.

The four prompts below go in order and pass their result to each other. The first describes the reader, the second — your position, the third gives topics, the fourth writes in your voice.

The first prompt is needed to get a person instead of a category. Categories don't read posts, people do.

Pay attention to the second point of the answer — the pain map. It splits into three layers, and they work differently.

Act as an audience strategist. Help me build a detailed portrait
of the ideal reader on Threads.

My niche: [fill in]
What I sell or plan to sell: [fill in]
Who I roughly see as my audience: [fill in]

Give me:
1. One specific ideal reader (name, situation, why they opened Threads).
2. A pain map: surface, hidden, root, with examples of their real phrases.
3. 15–20 words and phrases they actually use, so I can put them in hooks.
4. 5 topics that will make them feel "this is about me".

Everyone in your niche has a topic. You get remembered for your angle — the position others are afraid to take.

So in the answer, look at point four: where you go against the majority. That's exactly what gets you reposted.

Act as a positioning strategist for Threads (a text platform for opinions).

My niche: [fill in]
My experience and results: [fill in]
The audience I want: [fill in]
What I want to sell: [fill in]

Give me:
1. 3 strong positioning angles (an angle and opinion, not a topic).
2. For each: who it's for, what pain it addresses, why it sparks debate and reposts.
3. 2 versions of a three-line profile bio (position, benefit, call to action).
4. My signature tension: 1–2 beliefs where I go against the majority.

The third prompt gives you volume. Thirty ideas is a month's worth of supply to pick from. You don't need to write all of them.

Ask it to sort by type and be sure to mark the ten strongest. Without this request, you'll get a flat list where it's unclear where to start.

Act as a content strategist for Threads. Generate 30 post ideas
for my niche, based on formats that have proven viral.

Niche: [fill in]
Audience and their pains: [fill in]
My strong opinions: [fill in]

Sort by type: opinions and provocations, personal stories, value and methods,
myth-busting, engagement. Give each idea a draft hook (first line).
Mark 10 with the highest potential for replies and reposts.

The fourth prompt does two things at once: first it extracts a profile of your voice from your own texts, then it rewrites someone else's working format for your topic.

You can't skip the first step. Without five to ten of your own texts, the neural network will write in its own voice, and that's recognizable from the very first line.

Act as an editor and copywriter for Threads.

Step 1. Here are 5–10 of my texts, extract my voice profile
(rhythm, phrase length, humor, vocabulary, tone of address): [paste texts]

Step 2. Here's a format or trend I want to repack: [paste trend]
My topic: [fill in]

Do this:
- Repack this format for my topic in my voice (not a literal translation).
- 10 variants of the opening line: opinion, promise, pain recognition,
  curiosity, number. Mark the 3 strongest.

Where to find the format itself for step two — in the breakdown about finding other people's viral posts.

This is a good place to check with the people who build these models. Anthropic states plainly in its prompting documentation: the model gravitates toward an averaged, "on distribution" result. Not a bad one — an average one.

And that's where the fix comes from too. If you want above-average results, you have to ask for them explicitly. They flat-out advise against expecting the model to figure it out from a vague brief.

This has a practical takeaway for your voice profile. Set up a separate project for your blog and keep the profile there: project memory doesn't mix with your other conversations.

If you're working in incognito or a different model, you'll have to paste the profile in by hand every single time.

From the same source come two formatting techniques that matter more than any wording tweak.

In our chain, this changes the order inside the third and fourth prompts. The answers from the first two steps go on top as context, while the task itself stays at the end.

First day free, no card, no signup. Set it up in ten minutes and test it on your own account.

Not ready for a tool — the prompts above work and are free, use them.