Formatting may help people read a post, but this small experiment cannot show that formatting alone caused more reach. Shapeit's own three-pair comparison reported higher impressions and engagements after several changes were made together. Treat it as a useful observation, not a promise for your next post.
The experiment and its limits
The original Shapeit study describes a company page with approximately 12,000 followers, three post ideas and a four-week observation window. Each original post was followed 7–10 days later by a version with an AI-assisted opening, spacing, emphasis and a question or takeaway. The study reports consistent posting windows and no paid promotion.
The owner confirms that supporting data exists. The raw analytics exports are not attached to this article, and the results have not been independently audited. The table below preserves the reported observations. No new experiment was conducted for this update.
| Post idea | Original impressions | Revised impressions | Original engagements | Revised engagements |
|---|---|---|---|---|
| Personal lesson | 1,240 | 1,860 | 34 | 78 |
| How-to steps | 2,050 | 2,740 | 56 | 96 |
| Quick tip | 980 | 1,380 | 18 | 36 |
The arithmetic mean of the three pairwise changes is approximately +41% impressions and +100% engagements. These are averages of percentage changes, not a claim that every post doubled its engagement. Combining the totals gives different figures: 4,270 to 5,980 impressions (about +40%) and 108 to 210 engagements (about +94%).
What the numbers do not establish
Several variables changed together: the opening, layout, emphasis and ending. We cannot separate their effects. The posts were not randomized, the audience and feed conditions could change, and repeating an idea may itself affect performance. Three pairs are too few to establish a reliable expectation for another account.
We also did not measure the mechanism. It would be speculation to say these changes caused longer reading time, more expansion clicks or an algorithmic boost unless those metrics were collected. A larger number of impressions does not by itself demonstrate more qualified leads or customers.
What LinkedIn says about AI-assisted writing
In its June 2026 guidance, LinkedIn distinguishes useful AI assistance from generic content without an author's perspective. It says generic AI content may receive less distribution beyond an immediate network. That supports checking the substance and originality of a draft; it does not establish that any particular formatting style earns extra distribution.
Run a more useful check on your own account
Choose a question your audience already cares about. Write down the hypothesis before posting, such as “A concrete first sentence will make this explanation clearer.” Change one major element at a time where practical, and record the topic, audience context, posting time and metric definitions.
Compare several posts over a consistent window. Track relevant replies, saves or inquiries alongside impressions when those measures are available. Record what did not improve as well as what did. Do not treat a single successful post as proof of a repeatable formula.
Use Shapeit for the part it can help with
Prepare alternative openings, organize the explanation into short paragraphs, and inspect the approximate mobile preview. Keep an ordinary-text version so the message remains readable. A model's writing score is editing feedback, not a validated prediction of reach.
Prepare a clearer draft in Shapeit, then evaluate it against your own audience and goals.