Economist Novosad Warns AI Writing Flood Will Crowd Out Human Readers

On October 22, Harvard development economist Paul Novosad posted a series of threads on X laying out his assessment of how AI-generated content will upend writing and academic publishing. He expects non-native English speakers to increasingly rely on AI for English writing, arguing that if AI's cost advantages are large enough, the trend is nearly inevitable.

Confirmed

• Novosad makes a slippery-slope argument: when AI pushes total text volume up by tens of times (he repeatedly cites "50x" as the order of magnitude), carefully reading each piece on its own merits becomes unsustainable in equilibrium—people will end up having AI read and review AI-written text, and human readers will inevitably be crowded out.
• He notes that journals are starting to use detection tools to filter submissions, a sign that the ideal review process is already breaking down under surging submissions—AI can mass-produce plausible-looking papers with minimal human involvement.
• He worries that learners who rely on AI for academic reading and writing from the start will likely see their academic English atrophy, and points out that current debate participants are all fluent English speakers; the more important question is the impact on beginners—AI summaries will become the norm for new learners.
• As an institutional response, he proposes dedicated "AI-written/AI-reviewed" spaces to explore fully AI-generated-plus-AI-reviewed workflows, kept separate from existing human venues (literary circles, review sections, academic journals) so human journals aren't destroyed first.

Why it matters

These arguments cut to the core of academic publishing, language education, and the content ecosystem: when production costs approach zero but review costs stay flat, traditional peer review and deep reading face structural pressure. Novosad's "partition" scheme offers one discussable institutional blueprint for embracing AI workflows while protecting spaces for human communication.

2026-10-22 ~ 2026-10-22 · 5 related posts

• AI will multiply text 50x, making 'judge each paper on its merits' impossible — paulnovosad · 2026-10-22
• [source] Economist: AI writing volumes will inevitably crowd human readers out of equilibrium — paulnovosad · 2026-10-22
• [source] Economist proposes separate AI-native writing spaces to shield human institutions — paulnovosad · 2026-10-22
• [source] Economist Paul Novosad: non-native English speakers will increasingly rely on AI writing — paulnovosad · 2026-10-22
• If AI does all academic reading and writing, will new learners' English regress? — paulnovosad · 2026-10-22