In a world where AI can generate almost anything, originality and strong visual storytelling matter more than ever. Audiences no longer respond to generic “content,” but to stories that feel authentic, human, and alive.

VG’s viral sleep experiment began with curiosity. Through iteration, it offered fresh insights into how small adjustments can keep readers engaged far longer.

The challenge

A sleep trick was exploding across TikTok and Instagram: Users swore they could fall asleep in seconds by moving their eyes in a simple pattern. One in three people struggle with falling asleep, making the promise of a quick fix especially tempting.

We were curious: Could this viral sleep trick actually do what people claimed?

The method itself couldn’t be simpler — small eye movements right, left, up, down, and in a circle, while lying still with eyes closed. We wanted to try it but also to understand it, so we invited a sleep expert to join in on the experiment.

The experiment

We started with a simple idea: What if we tested the trick the same way people experienced it online — with video, first-person storytelling, and a touch of playfulness?

We teamed up with a sleep professor, a respected academic from renowned the university Norwegian University of Science and Technology (NTNU) in Norway, who was willing not only to test the method himself but also film the process from his own bedroom.

In addition, five VG journalists agreed to try the same method over three nights, keeping logs and filming themselves along the way.

The article opens with a short mobile video filmed by the professor himself, lying in bed, doing the eye movements. That immediacy makes the experiment feel real from the very first second. Throughout the story, we use short video clips showing the actual eye movements, social posts from people praising the trick, and our own journalists trying it out.

We also aim to make the reading experience more visual and dynamic than a traditional article. Illustrations help guide the reader through the method, and an interactive poll — “How fast do you fall asleep?” — invites readers to take part as they scroll.

The goal is to let the audience feel like participants, not just observers.

The drop off

After publishing the story, we monitored closely how readers moved through it. The data showed a clear drop-off point halfway down the article — right before the section that explained the sleep trick in detail. That insight triggered a round of editorial adjustments.

We made seven key changes between the first and second versions:

1. Moved the explanation of the sleep trick higher up.
2. Clarified the steps of the eye trick using bullet points instead of dropdown menu.
3. Removed a reader poll.
4. Tightened subheads and removed “how to” phrasing.
5. Placed the professor’s results last, after the journalists’.
6. Moved the paywall further down — revealing what the trick was, but not the results.
7. Cut redundant paragraphs to improve pacing.

The new version performed noticeably better. The reader drop-off curve smoothed out, and significantly more people reached the section showing the professor’s results. Around 10% more readers got to the end of the article.

In the chart below, the orange bars show where readers dropped out of the article. On the left is the first version of the article; on the right, the updated one.

Before the changes, many readers stopped before the middle of the story. After the adjustments, the curve looks much smoother: Readers still drop off early, as they often do, but far fewer leave halfway through. More people now read all the way to the end.

The story also converted strongly to subscriptions, with particularly good traction among women and younger readers — audiences that don’t always respond to science or health content in traditional formats. The average read time hit 1 minute, 51 seconds.

In other words, this was the same journalism repackaged for a better and more engaging reader experience.

We also made a short social video version of the story to accompany the visual article on VG.no. It found strong traction on social media, reaching 477,990 viewers on TikTok, and 232,000 viewers on Instagram.

@vgnett Med dette trikset skal du sovne superraskt 👀

What we learned

The experiment reminded us good storytelling still depends on human curiosity — and on being willing to adjust once the story meets real readers. Data alone didn’t tell us what to do, but it helped us see where people lost attention and why.

Small structural changes had a surprisingly large effect. Moving key information higher up, clarifying the trick visually, and improving rhythm and pacing turned the story into one that people actually stayed with.

It also reinforced something we believe will matter even more in the years ahead: unique, lived journalism — stories that can’t be automated or copied. Letting journalists test things, film themselves, and invite experts to create these kinds of stories is a level of authenticity AI can’t replicate.

And the sleep trick itself? The results were mixed: Some of the reporters fell asleep faster, while others, like the professor, found themselves more awake.

In the end, the real discovery wasn’t about falling asleep; it was about how we keep readers awake.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .