For founders trying to crack the Hacker News algorithm, the temptation to lean on Large Language Models for launch copy is strong. But new research from Lettercamp suggests that shortcut might be costing you traction. By analyzing 108 Launch HN posts from the past year, the team discovered a clear correlation between manual writing and community engagement, challenging the notion that AI-generated text is indistinguishable to the discerning HN crowd.

The Data Breakdown

The study utilized Pangram’s classification tools to sort the posts into three buckets: AI-generated, mixed, and human-written. The distribution was surprisingly balanced, with 22 posts (20.4%) flagged as AI-generated, 44 (40.7%) as mixed, and 42 (38.9%) as human-written. However, the engagement metrics told a different story. Human-written posts achieved a median of 83 upvotes, while mixed posts trailed slightly at 80. AI-generated posts lagged significantly behind, securing a median of only 56 upvotes.

Voice and Authenticity

Lettercamp’s analysis points to a subtle but critical factor: voice. While YC guidelines don’t explicitly ban AI, they emphasize writing in your own voice. The data suggests HN users subconsciously detect the rhythmic, short-sentence patterns common in LLM outputs. For example, posts marked as AI, such as those from Chamber and TeamOut, exhibited stylistic markers that the author argues alienate readers. The gap between human and mixed posts is narrow, but the drop-off for fully AI-generated content is stark.

Outliers and Exceptions

No dataset is without its anomalies. The highest-scoring post overall was Tweeks, a browser extension, which fell into the 'mixed' category with 351 points. Among strictly human-written posts, Freestyle led with 322 points, while Sentrial sat at the bottom with 31. Interestingly, the top AI-generated post, RunAnywhere, managed 240 points, yet even this success story underperformed compared to the top human-written entries. This reinforces the idea that while AI can help, it rarely serves as the primary driver of viral success on this platform.

Key Takeaways

  • Human-written Launch HN posts have a median of 83 upvotes vs. 56 for AI-generated posts.
  • 20.4% of analyzed posts were fully AI-generated, while 38.9% were purely human.
  • Mixed posts perform nearly as well as human-written ones (80 median upvotes).
  • HN users appear sensitive to the specific syntactic patterns of LLM-generated text.

The Bottom Line

Stop letting an LLM ghostwrite your launch. The data shows that while AI can assist, authentic human voice is the primary driver of engagement on Hacker News. Your best tool for traction is still your own keyboard.