Uses WildChat
AI Fiction in the Wild
arXiv preprint · First arXiv submission
Studies how people generate and consume fiction through chatbots, combining conversation classification with close reading of repeated story requests. The authors examine fanfiction, roleplay, and erotic fiction, and discuss how iterative generation can reshape the relationship between reader and writer. A companion viewer supports exploration of the fiction subset.
Data used
573,453 English-language conversations from the first WildChat release, covering April 2023–May 2024.
Findings
The study classifies 195,271 conversations (34%) as fiction. Sampling one conversation per inferred user reduces this share to 7.1%.
Sections 3–4: classification and user grouping by hashed IPs and similar prompts. Table 3: conversation counts. Read the source.
Dates and versions
First arXiv version; presented earlier at the MFS Cultural AI Conference on September 19, 2025. ArXiv version history.
2026 arXiv preprint, version 2.
Cite this paper
Neel Gupta, Maria Antoniak, Melanie Walsh. 2026. AI Fiction in the Wild. arXiv preprint.
Show BibTeX
@misc{gupta2026aifiction,
title = {{AI Fiction in the Wild}},
author = {Gupta, Neel and Antoniak, Maria and Walsh, Melanie},
year = {2026},
doi = {10.48550/arXiv.2606.22748},
eprint = {2606.22748},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.22748}
}