who wrote this?

my ap research study. 47 high schoolers read the exact same one-page analysis of world war 2, guessed whether a phd historian or an ai wrote it, and rated how much they trusted it. the byline changed everything.

the full title is “authority vs. automation: the effect of perceived source on student trust and critical scrutiny.” here's the version you'd actually read.

03 / the tldr

a
the setup

everyone read an identical historical analysis. no hints about the author. then two questions: who do you think wrote this, and how much do you trust it, 1 to 10. their written reasoning got coded from 0 (vibes) to 2 (pointed at actual evidence).
b
the trust gap

students who thought a human wrote it trusted it more: 7.48 out of 10 versus 6.59 for the ai group. honest caveat, with 47 students that gap misses statistical significance (p = .09, cohen's d ≈ 0.5). a lean, not a law.
c
the flip

the real finding. of the students who thought it was ai, 73% raised substantive critiques, missing citations, suspicious statistics, real gaps. of the ones who thought it was human, only 28% did. same words on the page, 2.6× more actual scrutiny.
d
the takeaway

the label does half the reading for you. call a text human and people nod along. call it ai and they start fact-checking. which means unlabeled ai text is the dangerous kind, and labeling it might quietly be a reading-comprehension hack.
e
the fine print

one school, one classroom, convenience sample. it's a direction worth chasing with a bigger n, not a settled fact. i know exactly where the cracks are, i wrote them into the paper myself.

why no full paper here? because it reads like it was written by a high school senior, which it was. the data is real though. if you want the methodology, the rubric, or the raw numbers, email me.