MagicPost · LinkedIn Benchmark
April 2026
Impressions per post by follower count · 22,224 posts analyzed| Followers | Low | Median | High | Top 10% |
|---|---|---|---|---|
| 0-1K | 86-14% | 187-12% | 420-17% | 1,003-19% |
| 1K-5K | 191-13% | 433-10% | 1,043-8% | 2,778-6% |
| 5K-10K | 338-2% | 756-2% | 1,923-3% | 5,144-1% |
| 10K-25K | 705-8% | 1,679+0% | 4,778+14% | 14.3K+16% |
| 25K-50K | 1,320+14% | 3,210+20% | 9,264+30% | 29.7K+44% |
| 50K-100K | 3,321+22% | 8,009+25% | 22.1K+28% | 59.2K+32% |
| 100K+ | 6,237+10% | 14.1K+5% | 31.9K+8% | 87.1K+15% |
Low = bottom 25% · Median = middle post · High = top 25% · Top 10% = best decile
Six-month trend
Median impressions for a LinkedIn postimpressions on a image post
+NaN% vs the same author's other posts
+NaN% vs the same author's other weekdays
Posts built around contrast formulas like "it's not X, it's Y" reach on average 10% fewer people than the same author's usual posts. The pattern is consistent across creators and industries. These phrases have been used so often that audiences recognise them instantly and scroll past. The idea that LinkedIn penalises AI-sounding writing is not supported by the data. When comparing a creator's high-AI-score posts directly against their low-AI-score posts, no measurable difference in reach appears. The issue is not how a post is written, it is whether it says something that could only come from that particular person. The practical takeaway is straightforward: before posting, ask whether this sentence could have been written by anyone else covering the same topic. If the answer is yes, replace it with a specific angle or example that is yours alone.