In visual carousels this trick is done with a photo sliced across two slides: the frame doesn't end, and your thumb reaches to swipe on its own. In a carousel without photos — like ours, where a slide is built from text and typography — the same effect comes from an unfinished sentence. Below: how an AI builds that "loop", a working prompt, and where the technique should stop.
The mechanics: why cutting a sentence works
The psychological effect is called the Zeigarnik effect: an unfinished action stays in memory more strongly than a completed one. In a visual carousel that's the sliced photo. In a text one it's a sentence that breaks on a meaningful word at the end of a slide and finishes on the next one. The reader physically can't stop before finding out how the sentence ends.
The prompt: marking up the open loop between slides
The prompt works on carousel text you already have — it doesn't write content from scratch, it rebuilds the boundaries between slides so that the end of one becomes curiosity about the next.
Here is the carousel text, slide by slide: [paste your draft].
For every pair of adjacent content slides, make the last
sentence of the first slide break on a meaningful word or
thought — without losing the meaning of the rest of the text.
Start the next slide as a direct continuation of that thought,
not as a new topic.
Don't break the hook and cta slides — the cover and the ending
must read on their own, even if they're opened separately.
What to fill in: your finished slide text from any draft — the technique is applied on top of existing content, it doesn't replace it.
Example: what it looks like in JSON
Below is a fragment of two adjacent content slides with the technique applied, in the format that maps straight onto the carousel JSON schema.
{"role": "content", "title": "Why this works",
"body": ["The brain hates a thought being cut off —
which is why the best way to hold attention is to..."]},
{"role": "content", "title": "...open the slide with the answer",
"body": ["not with a new topic, but with the direct
continuation of the previous slide's broken sentence."]}
This pairing works without photos too — the carousel's typography simply continues the thought visually, in the same typeface and the same tone, with no break in style between slides.
Where the technique breaks
It works well on short carousels — 4-6 content slides. On a carousel longer than 8-10 slides, holding an unclosed thought in your head gets tiring, and instead of curiosity the technique starts to irritate the reader with the sense of being strung along.
The second limit: it doesn't work without substance underneath. If every slide ends on a loop but nothing genuinely useful follows, the audience reads the manipulation fast and closes the carousel before ever reaching the cta.
How we do it
We apply the same open-loop principle when marking up our own carousels for qvib.pro — we just keep the length in check: no more than 5-6 content slides per carousel, and every slide carries value of its own beyond the hook. The full JSON format of our carousel slides — the same one our render pipeline runs on — is open in the prompts section of the arsenal.
Related links
- The "Viral carousels" arsenal section: qvib.pro/arsenal/virusnye-karuseli/
- Prompt engineering and post structure: qvib.pro/arsenal/promt-inzhiniring/
- How the qvib engine works: qvib.pro/