Research opportunity · Individual gain, collective narrowing? Does AI make science, creative work and shared knowledge more alike?
Is the diversity of what reaches the market falling, or only diversity inside experiments?
Question
In markets for creative and intellectual work, such as self-published fiction, has the diversity of published works, of works that sell and of works that reach the top ranks changed since generative AI spread, beyond what changes in volume explain?
Not yet measuredA PhD projectNeeds a budget (medium or higher)Partly addressed
Why it matters
Experiments show that work made with AI is more alike within a task, and one platform study shows average novelty falling while peak novelty rises. Readers and buyers meet a market, not an experiment. If markets select for distinct work, pool-level convergence may not reach the shelf; if AI-assisted work floods the catalogue and the top ranks, market diversity could fall even where individual works are rated well.
The debate lists this among what we still do not know: Does market-level diversity (books, apps, published papers) fall, or only diversity within experiments?
Evidence that motivates it
- On a large art-sharing platform, artists who adopted text-to-image AI raised their output by 25% and the favourites their works received per view by 50%; average content novelty and visual novelty fell over time, while peak content novelty rose and favourites became less concentrated among adopters.
On an art platform, adopters' average novelty fell while peak content novelty rose and favourites became less concentrated. - In a preregistered online experiment, writers given one or up to five story ideas from GPT-4 wrote stories rated 5.4% and 8.1% more novel and 3.7% and 9.0% more useful than writers without AI, with the largest gains for less creative writers; but their stories were 5.2% and 5.0% more similar to the AI ideas and more similar to each other.
AI-assisted stories were rated better and were more alike in an experiment. - Kaggle competition code from 2019 to mid-2026 has become more alike in its literal syntax, but the author finds no evidence of homogenization in the approach or intent of the code.
Code converged in syntax but not, by one measure, in approach.
A possible design
Build a panel of self-published genre fiction from before 2022 to 2026 with full text, release dates, genre and sales rank. Measure diversity at three levels (all releases, books with sales, top-ranked books) on plot and theme (structured summaries coded and validated by blind readers) and on style (embeddings), separately. Compare the trend in each genre with the genre's estimated share of AI-written text, holding the number of releases fixed by resampling.
Data needed
Full texts or long samples of published books with sales data; a validated distribution-level estimator of AI text; reader coding of themes.
Feasibility
A PhD project. Medium; data access and reader coding are the main costs. Copyright limits sharing the texts; the analysis can share derived features only. AI detection at the book level is uncertain and should be treated as a share, not a label.
Has this been done?
Partly done already: related work covers some of this question (first-pass screen of 15 October 2026; not yet checked in depth).
The nearest study measures dilution of sales and revenue per book by AI-written books in self-published fiction, and language overlap with existing books, but not the diversity of themes and plots across the catalogue or the top ranks.
How strong this evidence is
- Who searched and checked: .
- What was read: 2 of the 3 nearest works below were judged from an abstract, summary or search listing only; each is marked. Not every entry says how much of the work was read.
- Where the search could not reach: Screened by two web searches and an arXiv record lookup on the date shown; cultural-economics journals were not searched, and no study of app markets was found. Absence of a hit is not proof that the work does not exist.
- Before building a project on it, run your own search in a scholarly database such as Scopus, Web of Science, Google Scholar or SSRN.
Nearest work found
- Chakrabarty, Liu, Ginsburg & Dhillon, "Generative AI floods and dilutes the market for books," arXiv:2607.20349 (2026), working paper under review Judged from an abstract, summary or search listing only
Abstract read. 14,419 self-published genre-fiction books with daily sales; AI-text books are a large share of the catalogue, a smaller but growing share of sales, and draw on more distinctive language from existing books; no measure of thematic diversity. - Zhou & Lee, art-platform study (LD09-SL-C11)
Full text read for this debate. Adopters versus non-adopters on one platform; not the diversity of the whole market. - Slendebroek & Metaxa, "The Algorithmic Flattening of Sound: Computational Evidence and Justice Implications of AI Music Homogenization," AIES (2026) Judged from an abstract, summary or search listing only
Seen in search results only; the paper was not read. Reported to audit 100 generated tracks per genre from two music generators against human corpora; generator output, not the diversity of a market.
Search queries used
market-level diversity published books novels self-publishing after ChatGPT homogenization studygenerative AI music or app store market diversity homogenization new releases study 2026arXiv id_list: 2607.20349
Provenance
RO-LD09-03 v1. Drafted by Claude (SocioLens Project Lead), AI-drafted from the LD-09 debate on 2026-10-15. Not yet reviewed. Not yet reviewed by a domain expert. This is a research hypothesis, not a finding.