AI-Exposed Jobs Are Recovering Faster Than the Rest of the Market

AI exposed jobs are recovering, and doing so faster than the rest of the US labor market. Indeed Hiring Lab's July 2026 analysis of its own job postings data found that between May 2022 and May 2026, the more exposed to AI an occupation was, the more its postings declined. Since early 2025, that relationship has flipped: the occupations with the most AI exposure are now the ones rebounding fastest.

The old pattern: exposure meant decline

For roughly three years after ChatGPT's late 2022 release, Indeed's data showed a consistent relationship: occupations more exposed to AI saw their job postings fall further than less exposed ones. Software development, one of the most AI-exposed occupations by Indeed's own classification, led that decline. This is the pattern behind most of the "AI is taking jobs" headlines from that period, and it was real, measured in actual postings rather than survey sentiment or a forecast.

The reversal since early 2025

The same analysis found that relationship reversed starting in early 2025: on average, the more AI-exposed an occupation is now, the more its postings have rebounded. Software development postings on Indeed grew almost 15% between February 2025 and May 2026, while overall US job postings fell 7% over the same stretch, per Indeed Hiring Lab. The occupation that fell hardest during the decline phase is the same one leading the recovery.

Who is actually getting hired

The rebound is not spread evenly across the workforce. 71% of the increase in software development postings between May 2025 and May 2026 came from senior roles, and 37% came from postings that mention AI directly in the job title, the same report notes. Entry-level demand has not recovered the way senior demand has, which lines up with what this site has already covered on entry-level hiring: employers increasingly want judgment that used to come with a title, not a junior headcount.

Why this is not a promise, just a pattern

Indeed's own economists are careful to note that a correlation between AI exposure and posting growth is not proof that AI caused the rebound, and the reversal covers roughly eighteen months against three years of decline before it. It also does not mean every AI-exposed role is safe, or that the pattern holds in every sector; it means the occupations hit hardest are not permanently shrinking. That distinction between an exposure statistic and a guaranteed outcome is the same one worth applying to any "AI will take X% of jobs" claim: a real number from real data, read for exactly what it measures.

What this means for an AI-exposed role

A reversal at the occupation level does not tell an individual reader whether their specific tasks are becoming more valuable or more replaceable, and the split between automation-prone and augmentation-prone work still runs inside every exposed occupation, software development included. The postings data says the senior, judgment-heavy end of an AI-exposed field is hiring again. Showing that judgment, not just claiming the job title, is what the six-domain check is built to evidence.