/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Analysis: scientists who appeared to use LLMs posted 33% more papers on arXiv than those who didn't, as concerns grow over AI slop in scientific publishing

Peer review has met its match.  —  On a frigid Norwegian afternoon earlier this month, Dan Quintana, a psychology professor …

The Atlantic Ross Andersen

Context & Ripple Effects

The reported association between apparent LLM use and higher arXiv posting volume turns a broad concern about scientific AI use into a measurable publishing-incentive problem. It matters because preprints circulate before conventional quality controls can fully assess them.

The issue sits alongside evidence that LLM-written text had already entered conference reviewing, with LLM-generated sentences detected in computer-science reviews, and later pressure on conferences to restrict LLM use in papers and reviews.

First-order effects

  • Researchers who use LLMs may be able to increase preprint output relative to peers who do not, while readers and moderators face a larger volume of work to assess.
  • The finding intensifies scrutiny of whether LLM assistance is supporting legitimate drafting or lowering the effective cost of low-quality submissions.

Second-order effects

  • Conference organizers and publishers have stronger reason to define and enforce disclosure or use rules, consistent with conferences moving to restrict LLM use in writing and reviewing.
  • More submissions can shift the bottleneck from producing manuscripts to screening them, raising the value of editorial triage and review processes that can identify fabricated or unreliable work.

Third-order effects

  • If AI-assisted output continues to outpace review capacity, scientific publishing may separate further into high-volume preprint distribution and more selective validation channels.
  • The durability of preprints as a discovery layer will depend on whether platforms can preserve trust as LLMs also enable fabricated-paper submissions by non-researchers.

The trend: Generative AI is reducing the cost of producing scientific text faster than the research ecosystem can verify its quality and provenance.

Discussion

  • @thiagokrause Thiago Krause on bluesky
    Horrifying - but less of a problem for history.  I still can't see a LLM-generated manuscript getting past peer-review, except in irrelevant/low-quality journals no one actually reads.
  • @dsquintana Dan Quintana on bluesky
    AI slop is everywhere in scientific publishing, we're only catching the easy-to-detect stuff (like when you happen to peer review a manuscript with a AI-hallucinated reference of a paper you apparently wrote)  —  www.theatlantic.com/science/ 2026...
  • @drjacekdebiec @drjacekdebiec on bluesky
    “AIs would write most papers, and review most of them.. This..back-and-forth would..train newer AI models.  Fraudulent images and phantom citations would embed themselves..in our..knowledge.  They'd become a permanent epistemological pollution..”  —  #AcademicSky  —  www.theatlan…
  • @kim_harding@mastodon.scot @kim_harding@mastodon.scot on mastodon
    Science Is Drowning in AI Slop  —  https://www.theatlantic.com/ ...  Peer review has met its match.