One of the most common predictions of the AI era was that the technology would hollow out knowledge work and drag down pay for the jobs left over. Almost four years after ChatGPT launched, the biggest jobs dataset in the world says the opposite is happening to advertised wages. Indeed Hiring Lab’s analysis, published on 17 September 2026, found advertised pay growing fastest in the occupations most exposed to AI — software development, IT support, data and analytics, marketing and finance.
🔍 THE BOTTOM LINE: US advertised pay in the most AI-exposed occupations has climbed about 46 percent since 2021, against 25 percent in the least-exposed jobs. The premium survives statistical controls at 5.7 percent, and the Indeed researchers’ reading is that AI is acting more as a complement to skilled workers than a replacement. The squeeze is real, but it is landing on new entrants rather than pay packets.
What the Indeed Data Shows
Indeed sorted US occupations by how much of their skills generative AI could perform or substantially reshape, then tracked advertised salaries in each group since 2021. The most-exposed third — software development, IT systems and support, data and analytics, marketing, banking and finance — has seen advertised pay rise about 46 percent. The least-exposed third — nursing, caregiving, food service, cleaning and manufacturing — rose about 25 percent.
The gap barely moved in the first year after ChatGPT, then began widening around 2024 and has kept growing since. Controlling for the changing mix of occupations, Indeed measures a post-ChatGPT pay premium of 5.7 percent. Comparing each job title against its own past narrows it to 4.7 percent; holding the seniority mix constant trims it to 2.4 percent, a figure Indeed notes is no longer statistically significant. That honesty matters: the premium is real in the first two specifications and debated in the third.
The Seniority Split Is Where It Gets Complicated
The same dataset shows why the AI-jobs debate keeps producing contradictory headlines. At senior level, pay in more-exposed occupations grew 45 percent since 2021 versus 28 percent in less-exposed ones — a 17-point gap. At mid-level it is about 12 points. At entry level, just 2. Indeed treats the split as suggestive rather than settled, and notes that measured against the 2022 baseline the post-ChatGPT gap is actually fairly even, roughly five to seven points at every level.
The composition shift behind those numbers is stark. In the most-exposed occupations, the entry-level share of salaried postings fell from 29 percent to 10 percent between 2021 and 2026, while the senior share rose from 22 percent to 47 percent. That mirrors what our coverage of the Census Bureau’s graduate working paper documented: the people paying AI’s early costs are graduates trying to get on the ladder, not the people already standing on it. Indeed’s own postings data shows entry-level ads down 7.5 percent year on year while senior postings surged 14.7 percent — even as software development postings overall rebounded 15 percent from their post-2022 lows.
Reading It Alongside the Bad News
This finding does not cancel the displacement research; it sits next to it. The Census working paper found the most AI-exposed majors facing a 13 percent earnings drop in their first quarter of work and a five-point fall in the odds of immediate employment. The Dallas Fed reached a similar shape a different way: entry-level roles shrinking while experienced workers’ wages rise. Indeed’s wage data and the entry-level hiring data are, put together, a coherent picture of a market that still values AI-adjacent skills highly but has thinned the bottom rung.
The Indeed researchers land on the complement interpretation: AI is “reshaping which skills the market pays for,” with employers competing for people who can direct the tools. Their advice to workers is implicit in the data — the premium accrues to people operating at the frontier of what the tools can do, and it widens with experience.
What It Means in New Zealand
New Zealand’s salary data is less dramatic but pointing the same way with the same blind spot. The Hays Salary Guide FY26/27, published on 7 September, puts average pay rises in NZ technology at 3.8 percent — but 42 percent of tech professionals received little or no increase, and 82 percent of employers report capability gaps. Meanwhile 60 percent of NZ tech professionals say they use AI tools regularly, yet only 22 percent have had formal training. The wage premium is arriving unevenly here too, and the people best positioned to claim it are those who formalise skills their workplaces have not structured.
On demand, SEEK’s August data shows AI skills now mentioned in 4.0 percent of NZ job ads, up 93.3 percent year on year, concentrated in ICT, marketing and consulting — the same high-exposure occupations Indeed finds pulling ahead on pay in the US.
❓ FAQ
Are AI-exposed jobs being paid less because of AI? US postings data says no. Indeed Hiring Lab found advertised pay in the most AI-exposed occupations up about 46 percent since 2021 versus 25 percent in the least-exposed, with a statistically significant post-ChatGPT premium of 5.7 percent after controlling for occupation mix.
Does the pay premium apply to entry-level jobs? Barely. The seniority gap is about 17 points at senior level, 12 at mid-level and 2 at entry level — and Indeed notes its seniority split is suggestive, not statistically confirmed.
What does this mean for new graduates? It is consistent with Census Bureau findings that graduates of AI-exposed majors face lower initial employment and earnings, with the damage concentrated in the first years. The premium appears to reward experience and AI fluency together.
Is the same happening in New Zealand? The data is thinner, but Hays reports NZ tech pay up 3.8 percent on average with big variance, and SEEK shows AI skill mentions in job ads up 93.3 percent year on year.
🔍 THE BOTTOM LINE
The wage story of the AI era so far is not collapse and not a windfall — it is a widening spread. Pay is holding up best in exactly the occupations people predicted AI would devalue, worst for the newcomers trying to enter them. For anyone mid-career in tech, data, marketing or finance, the data says the market is paying more for AI fluency, not less. For students and graduates, it says the same skills pay, but the first rung of the ladder has moved.
📰 Sources
- Indeed Hiring Lab, “AI Exposure Isn’t Squeezing Advertised Pay in the US — It’s Boosting It” (17 September 2026)
- Indeed Hiring Lab, “AI and Job Postings: From Destruction to Creation?” (8 July 2026)
- Indeed Hiring Lab, “The Labor Market Is Tilting Toward Seniority” (23 July 2026)
- US Census Bureau CES Working Paper 26-56, “Graduating into Disruption” (September 2026)
- Hays Salary Guide FY26/27, New Zealand Technology Snapshot (7 September 2026)
- SEEK NZ Employment Report, August 2026
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
