The AI Job Scare Needs Better Evidence

The AI jobs apocalypse was supposed to start with ChatGPT. Two years on, victims are hard to find. Recent studies from the Dallas Federal Reserve and economists at Pittsburgh, Stanford, and Columbia ("AI-exposed jobs deteriorated before ChatGPT") suggest artificial intelligence is altering labor markets—just not in the calamitous way the doomsayers predicted.

Mass displacement would leave obvious fingerprints: spiking layoffs, collapsing wages, unemployed workers unable to find new roles. The data show none of this. Separation rates in AI-exposed occupations haven't budged. Job-finding rates for the unemployed track normally regardless of AI exposure. And the weakness that does exist in fields like computer science began in early 2022—months before ChatGPT launched—consistent with Fed tightening and a post-pandemic tech correction, not technological rupture.

The data "do not give much indication of widespread labor market disruption due to AI," concludes the Dallas Fed, while the other research group says: "Unemployment risk in highly exposed occupations rose beginning in early 2022—well before ChatGPT—and in most occupations and states we observe no discrete break coincident with its introduction."

The real story so far is quieter and more selective than the one offered by too many click-bait headlines. Employment among 20-to-24-year-olds in high AI-exposure occupations has slipped nearly a percentage point since late 2022, while prime-age workers in the same roles are largely unaffected. But employers aren't firing people. Rather, they're raising the bar for new hires, perhaps preferring candidates who can contribute immediately over those who need training. The disruption is at the entrance, not the exit.

Granted: This creates a genuine problem for some young graduates, perhaps, but it's a different problem than obsolescence. LinkedIn data show recent cohorts entering AI-exposed jobs at lower rates and taking longer to land them. Yet those who do get hired—especially those whose coursework emphasized writing, coding, and information synthesis—are earning higher starting salaries than their predecessors. The skills aren't suddenly worthless so much as the competition for entry-level seats is just fiercer.

The bank Goldman Sachs offers a useful frame in a recent report: AI could automate a quarter of work tasks while displacing only six to seven percent of jobs over time. Even in pessimistic scenarios, unemployment rises modestly and temporarily. That's disruption, not apocalypse.

None of this means complacency is warranted. A generation struggling to land first jobs stores up trouble for later. But the current evidence suggests AI is making hiring more selective rather than making workers redundant. The job apocalypse may yet arrive. For now, it's just early—which, in forecasting, is another way of saying wrong.

The post The AI Job Scare Needs Better Evidence appeared first on American Enterprise Institute - AEI.