AI Slop Is Flooding Science. Here's the Playbook for Writers
Generative AI has supercharged output. It's also flooded the web with "AI slop" - low-effort text dressed up as expertise. A large-scale study of more than a million scientific preprints suggests the slop is winning on volume, and that's a warning shot for anyone who writes for a living.
The takeaway isn't "don't use AI." It's this: fluent language is no longer a proxy for quality. Substance, receipts, and honest clarity win.
What the research found (and why it matters to you)
Researchers from UC Berkeley and Cornell analyzed preprint abstracts from 2018-2024 and tracked what changed when authors started using AI. Two big shifts stood out:
- Productivity spiked: After adopting AI, monthly output jumped between 36.2% and 59.8%. The boost was largest for non-native English speakers (notably many Asian authors: 43%-89.3%). Authors at English-speaking institutions saw smaller gains (23.7%-46.2%).
- Complexity rose, quality signals fell: AI-aided articles used more complex language. But unlike human-written papers (where complex language correlated with acceptance), AI-driven complexity correlated with lower odds of journal publication. In short: complex wording was often masking weak work.
For writers, the signal is clear: editors and audiences are getting skeptical of polished fluff. Prose that reads "smart" but says little will work against you.
Search is shifting what gets read
Another finding: users of Bing's AI-enhanced search saw a wider mix of sources and more recent citations than Google users. That aligns with retrieval-augmented generation (RAG), where AI drafts against fresh indexed documents instead of fixed training data.
Translation for writers: newer, niche, and timely sources are more discoverable now. There's upside if you cite well - and downside if you lean on dated talking points.
The anti-slop checklist (use before you publish)
- State the thesis in one sentence. If you can't, the piece isn't ready.
- List 3-5 claims with sources. Prefer primary data over recycled summaries.
- Prove every claim. Quote the study. Link the dataset. Name the method.
- Cut decoration. Remove filler like "clearly," "significantly," and vague abstractions.
- Swap complexity for precision. Short sentences. Concrete nouns. Verbs that do work.
- Ban fake citations. Cross-check every reference. If you used AI to draft, verify manually.
- Read it aloud. You'll catch hedges, hedging, and hollow copy.
- Add receipts. If you can't show data, show process: how you tested, who you interviewed, what you measured.
How to use AI without turning your work into slop
- Draft for speed, edit for truth. Let AI give you scaffolding. You supply angle, examples, and evidence.
- Translate and clarify. Great for non-native writers: use AI to clean language, then re-check technical meaning.
- RAG your prompts. Paste in your notes, sources, and quotes so outputs anchor to real material.
- Quarantine style from substance. Ask AI for tone edits last. Don't let polish hide weak claims.
- Timebox dependence. 15-20 minutes for ideation, then switch to your own outline and reasoning.
- Track your sources as you write. Keep a mini bibliography to avoid post-hoc scrambling.
Editing signals that beat "smart-sounding" text
- Clarity beats cadence. If a sentence sings but doesn't move the argument, cut it.
- Specifics beat generalities. Numbers, names, dates, quotes.
- Recency beats familiarity. Pull a newer, credible source over the usual greatest hits.
- Context beats claims. Explain the "how" and "why," not just the "what."
For editors and content leads
- Change your first-pass filter. Don't equate complex language with merit. Scan for method, evidence, and originality.
- Require receipts. Links to sources, brief method notes, and data where applicable.
- Triage with AI, judge with humans. Use AI to cluster submissions and surface anomalies; rely on human editors for substance.
- Reward clarity. Make "explain the mechanism" a rubric item.
Why this matters for your career
As AI becomes embedded in docs, email, and search, avoiding it won't be practical. But using it lazily will sink trust. The edge goes to writers who ship clean thinking, current sources, and honest structure - with or without AI.
Volume is cheap now. Judgment isn't.
Tools and next steps
- Build a source habit: Keep a rolling list of credible journals, datasets, and experts in your beat.
- Standardize your RAG workflow: Collect notes and links first, then prompt against them.
- Audit your drafts: Once a week, pick one piece and trace every claim back to a source.
- AI tools for copywriting (curated list)
The bottom line
AI helps you write more. It doesn't think for you. Keep your edge by prioritizing verifiable ideas over ornate language, and use AI as a scaffold - not a crutch.
Your membership also unlocks:
