Using Artificial Intelligence to Write Viral Content on Threads

[brttheplugg]

Hello.

I have a problem. I’m growing a Threads account to sell digital products. The niche focuses on women’s relationship issues. Unfortunately, when I create my Threads posts using Claude, they aren’t generating audience engagement, etc. Out of the 25 posts on my profile, only 3 have gotten around 10k views. I used Apify to save popular posts from accounts in my niche along with their stats. Competitor accounts have a post that gets over 10k views practically every day. I’ve tried everything so far, but nothing’s working. I created a project in Claude with the best posts saved, and I’m crafting detailed prompts so that Claude can better understand the mechanics of competing posts and the language of the niche. But when I create posts, unfortunately, the results aren’t great. I also use a humanizer to eliminate AI-generated writing. What can I do to improve my results? (One more thing: The format of the posts I’m creating is a series of linked posts.)

 

[Arthonio]

in my experience threads is very random. I do have like 20 threads account, most of them are sitting at small numbers, a handful at like 30-60k followers.

What i figured out, u need to try a lots of stuff - once one type of format works, you can repost it until it stops working.
Posts which are triggering people, works the best
Likes are unimportant, only comments help you to go viral

 

[floqal]

Three out of twenty five landing isn't an AI problem, it's an input problem. Ask a model for a post about why she keeps going back to men who won't commit and you get the average of everything ever written on that subject, because averaging is what the thing does. That average is precisely what your reader has scrolled past a hundred times already, so the thumb keeps moving.

We write and test cold outreach copy at volume, and the pattern is the same every time. Copy a model produced from a topic underperforms. Copy built around one concrete thing the sender owns, a number, a mistake, a specific afternoon, gets replies.

So stop asking it to write and start giving it raw material. The message a reader actually sent you, an argument you had, the exact words someone used for what she felt. Let the model cut and tighten, never invent. And go back to those three posts, reply to every comment and post more on that same idea, because people who already noticed you engage far more easily than strangers do.

Best,
Floqal

 

[Chunkulus]

You scraped competitor posts with Apify including their stats, then went to Claude and asked it to write posts, which throws away the only asset you have. Feed it the scrape. Give it your fifteen or twenty highest performing competitor posts with the view counts attached, ask it to work out what the openers have in common structurally rather than topically, and only then ask for posts built on that pattern. You are moving the model from inventing to imitating something measured, which is a completely different task. Three of twenty five landing is also not necessarily a failure rate worth panicking over on Threads. Look at what those three share against the twenty two, because your own account is a smaller but far more relevant dataset than any competitor scrape.