Doomprompting: The New Doomscrolling That Looks Like Work

Why we lose ourselves in endless AI loops — and how to find our way out

We all know doomscrolling. The thumb swipes, the feed rolls, the time vanishes. We end up tired, irritable and no wiser than before. That behavior has now acquired a twin, and this one is far more cleverly disguised: doomprompting.

What Is Doomprompting?

The term has been circulating in English-language corners of the internet since 2025. It describes the endless back-and-forth with an AI: one more prompt, one more version, one more tweak. You rephrase, you refine, you sometimes contradict yourself. And with every round, you feel as though you\'re almost there.

The blog Working Theorys summed up the phenomenon in the summer of 2025: doomprompting is the new doomscrolling. Where social media turned us into passive consumers, AI turns us into passive conversationalists. Shortly afterward, the IT trade magazine CIO picked up the term and described how the constant reworking of AI outputs is devouring time and money across companies.

In German-speaking countries, the term has barely arrived yet. The behavior, on the other hand, has long since settled in.

Why the Loop Works So Well

Doomprompting follows the same patterns as any effective attention trap. It just feels better.

Variable rewards. Sometimes the answer is brilliant, sometimes mediocre, sometimes completely off. That very unpredictability keeps us hooked — just like a slot machine. The next attempt could be the one.

No natural endpoint. A book has a final page, a film has credits. AI has neither. It never grows tired, never impatient, and it tends to close with an offer: Should I create another version? Would you like it shorter? Every answer is an invitation to the next round.

The almost-good feeling. The result is always just one prompt away from perfect. That is the heart of the trap: we chase an improvement that often makes the output not better, merely different.

The illusion of productivity. This is what sets doomprompting apart from doomscrolling. When we scroll, we know we\'re wasting time. When we prompt, we\'re convinced we\'re working. We\'re typing, thinking, deciding. We only notice we\'ve been going in circles once the clock has already moved on.

Two Faces: Private Life and the Workplace

In private life, doomprompting costs mainly time and sleep. The idea for a blog post at midnight turns into a session that runs until half past two. Holiday planning stretches to a twelfth variation. The mind never settles, because it keeps receiving fresh material.

At work, the costs become financial. Spending hours polishing a prompt instead of thinking through the problem burns working time. With AI agents that consume computing power on every run, direct costs pile up as well. And the end result is often barely better than the third attempt.

Signs You\'re Caught in the Loop

You can no longer quite remember what you actually wanted.

Your prompts are getting shorter; the answers are getting longer.

You\'re comparing version seven with version four and find both acceptable.

You say \"one more, then I\'ll stop\" — for the third time.

You would have been done ages ago if you\'d just written it yourself.

Breaking Out of the Loop

The good news: doomprompting is not a law of nature. A few simple rules help.

Think first, then prompt. Before the first prompt, get clear on the goal. When is the result good enough? Anyone who can\'t answer that won\'t have an answer after twenty rounds either.

Brief, don\'t ping-pong. A thorough, complete brief replaces ten rounds of revision. Write the prompt as if you were briefing an experienced colleague.

Set a limit. Three versions, then decide. Not the perfect one — the usable one.

Take a break. When the loop starts, put the device down. Distance usually makes it immediately obvious which version was the right one.

Finish it yourself. Sometimes it\'s faster to add the final polish on your own than to explain to the AI for the fifth time what you mean.

Conclusion

AI is a remarkable tool. But like any tool that has no built-in closing time, it needs someone to put down the pen. Doomprompting is not the machine\'s fault. It is what happens when we hand control of our attention over to it.

The most important skill in working with AI is therefore not the perfect prompt. It is the moment when you say: that\'s enough.

This English version was created with the help of AI (translated from the German original) and may differ slightly from a professional human translation.