What the research actually found
AI fiction can sound fluent while making familiar narrative choices. StoryScope investigates those choices: how a story handles conflict, characters, time and meaning. That gives writers something more useful to inspect than a list of words to delete.
The University of Maryland and Google DeepMind team analyzed 61,608 stories. Their classifier reached 93.2% macro-F1 for human versus AI fiction using narrative features. That is a study-specific classification score, not a claim that 93.2% of arbitrary AI text can be detected.
The paper reports more explicit thematic commentary in AI stories, 77% versus 52% in human stories. It also identifies model-specific patterns, including flatter event escalation in Claude and gossip as a plot mechanism in GPT. These are population patterns, not rules for judging an individual passage.
The plot in the video is Figure 2 from the paper. Its axes are a projection of narrative features. It shows how stories cluster in that feature space; it does not depict handwriting or assign a writing-quality score.
Read the original paper: https://jenna-russell.github.io/assets/pdf/storyscope.pdf
The editorial tool I made from it
Human Scope turns that research into questions to ask of a draft. Does the obstacle develop before it disappears? Does the narrator repeat what the scene already shows? Does the ending resolve more than the story has earned?
I use those questions as an editing lens. A clear ending, direct explanation or linear plot can be exactly right. The skill preserves the intended genre, voice and point of view instead of forcing every story into the same idea of what sounds human.
Free, for your email
Get the full setup steps + the copy-paste setup prompt
The rest of this guide (7 more sections, 3 copy-paste prompts) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.
Unlocks: Download and install Human Scope · Start with a review before a rewrite · Turn the diagnosis into a targeted revision · A small example of the difference · Check the result as a writer · +2 more
50,000+ follow where these guides come from. No spam, reply “unsubscribe” any time.