When economic sentiment diverged from actual fundamentals in 2022, a question emerged: why are people so down on the economy when unemployment is low and growth looks solid? One theory gaining traction was that AI anxiety—specifically fears about job displacement from large language models—was poisoning the national mood. A new investigation puts numbers to that hypothesis, and spoiler alert: the data doesn't back it up.

What Is the Vibecession Anyway

The 'vibecession' theory, popularized by economists on social media since 2022, argues that public perception of the economy became irrationally negative relative to measurable fundamentals. Three main camps have emerged in this debate. Will Stancil argues viral misinformation and fragmented news consumption explain the gap. G Elliot Morris contends that current price levels—everything costs more than it did in 2021—are driving the malaise rather than inflation rates themselves. A third camp insists the economy genuinely is struggling, pointing to rising credit card debt and surveys showing Americans unable to afford basics. Economist Matt Darling has become a central figure, largely arguing with the third camp while supporting Stancil's framing.

The Hypothesis: ChatGPT Freaked Everyone Out

The timing seemed suspicious. Consumer sentiment started tanking right around when OpenAI released ChatGPT in late 2022. The author behind this investigation—writing at Kevin on the Margin—noticed that the University of Michigan Survey of Consumers showed a massive divergence between predicted sentiment (based on unemployment, inflation, and Federal Funds rates) and actual reported sentiment beginning in 2022. Meanwhile, the same survey asks respondents to estimate their probability of losing their job within five years. By 2025, this mean probability reached remarkably high levels despite historically low unemployment—suggesting anxiety was decoupled from reality.

Testing AI Anxiety Against Economic Vibes

To test whether AI job fears were driving bad vibes, the author ran a regression model including the five-year job loss probability as an explanatory variable alongside traditional economic indicators. The hypothesis: if AI anxiety was poisoning sentiment, adding this variable should dramatically improve the model's predictive power. Instead, the two models—the original without job anxiety and the new one with it—overlapped almost perfectly. Adding the job loss variable moved virtually nothing. The theory was elegant, but the numbers didn't cooperate.

So What Actually Explains the Vibecession?

The investigation concludes that AI anxiety likely isn't a major driver of economic sentiment divergence—or at least not in any way captured by this particular survey question. This leaves the original mystery unsolved. Will Stancil's social media misinformation thesis remains plausible, Elliot Morris's price-level framing holds water, and advocates for the 'economy is actually bad' camp can still point to auto loan defaults at record highs. The vibecession wars continue unabated, now with one less potential explanation crossed off the list.

Key Takeaways

  • ChatGPT's release timing roughly aligns with when consumer sentiment diverged from fundamentals, but correlation isn't causation
  • Despite low unemployment in 2025, survey respondents reported historically high five-year job loss anxiety—a disconnect worth monitoring
  • Adding job loss anxiety to economic prediction models improved predictive accuracy by essentially zero percent
  • The vibecession remains unexplained; three competing theories still lack definitive empirical support

The Bottom Line

Look, I wanted AI anxiety to be the culprit. It would've been a clean narrative—our industry creates tools that terrify everyone, and now we're measuring the psychological fallout in consumer confidence indexes. Neat story. But science doesn't care about neat stories. The vibecession is real, people's vibes are bad, and apparently it's not because they're worried about getting replaced by language models. Someone tell the AI doom crowd to find a new talking point.