ChatGPT became commercially available in late 2022 and then revolutionized the tech landscape. Every company under the sun prioritized AI software, and we’ve witnessed massive developments in the past two and a half years since ChatGPT went viral.
But as soon as ChatGPT arrived, we saw worries about AI replacing jobs. The fears worsened as OpenAI released better models, and competitors like Claude, Gemini, and DeepSeek arrived to challenge ChatGPT’s supremacy while delivering similarly powerful features.
The new theme in AI tech is agentic behavior for AI, which allows AI to work on tasks without human intervention. Some AI agents can browse the web and perform actions on your behalf. Others can help with coding. AI agents are still in the early days, but they’re a superior manifestation of AI sophistication, and they deepen the worry that AI might displace even more jobs.
While those fears are certainly warranted and need to be part of the conversation whenever an AI firm launches a potentially disruptive AI model, a new study says AIs like ChatGPT are hardly taking anyone’s job or improving productivity meaningfully. Using AI in the workplace might improve some tasks, but it’s creating an additional workload directly tied to the job the AI is performing.
Economists Anders Humlum and Emilie Vestergaard from the University of Chicago and the University of Copenhagen released a research paper examining the effects of AI, like ChatGPT, on the labor market. They concluded that “AI chatbots have had no significant impact on earnings or recorded hours in any occupation.”
The study (working paper) analyzed the effects of AI on 11 professions, all believed to be vulnerable to AI, per The Register. These are accountants, customer support specialists, financial advisors, HR professionals, IT support specialists, journalists, legal professionals, marketing professionals, office clerks, software developers, and teachers.
In total, the researchers looked at 25,000 workers in 7,000 workplaces in Denmark over the course of 2023 and 2024. I will note here the first obvious drawbacks of the study. First, the study might be representative of the Danish (and European by extent) market. Also, the 2023 and 2024 AI landscapes are widely different. AI tools in 2024 were significantly better than those available to workers a year prior.
Then there’s the fact that the study hasn’t been peer-reviewed, being released as a working paper for now.
Still, the research is relevant, as it shows that job displacement hasn’t been as massive as believed because the productivity gains expected from AI were countered by an increased workload.
“The adoption of these chatbots has been remarkably fast,” Humlum told The Register. “Most workers in the exposed occupations have now adopted these chatbots. Employers are also shifting gears and actively encouraging it. But then when we look at the economic outcomes, it really has not moved the needle.”
The researchers found that AI chatbots created new tasks for 8.4% of the workers in the study, even those who didn’t use AI. One example is teachers, who are now spending time trying to detect whether students are using AI like ChatGPT for homework.
Also, AI users spend more time reviewing the quality of work coming from chatbots, something that’s not surprising. I often tell you to ask for sources for ChatGPT claims, and that’s how I operate the AI. ChatGPT has to give me sources for everything it says, which I can check to ensure accuracy. AI hallucinations haven’t disappeared despite AI getting better. If anything, the newest ChatGPT models are more prone to offer incorrect information, even though they’re otherwise better at reasoning than their predecessors.
The researchers did find that AI like ChatGPT can save users time, but that amounts to just about 2.8% of work hours, or less than two hours a week. They say their findings contradict a February study claiming AI can enhance productivity by 15%, explaining that other research has focused on professions with the potential for high AI productivity, like customer support. Their study included real-world workers where the adoption of AI doesn’t lead to similar benefits.
Humlum told The Register that AI like ChatGPT can’t automate everything in the jobs they surveyed. Also, we’re in the “middle phase” of AI, where employees are still trying to figure out how and when the AI can help.
Finally, the researchers also found that if productivity gains happened, only between 3% and 7% of that benefit is passed on to workers via higher wages.
As I explained above, the study has limitations, so more research data is needed. However, these conclusions can’t be ignored, especially by AI firms like OpenAI. On the one hand, they can use it to say that AI adoption is high and that workers are using chatbots like ChatGPT, but they’re not losing their jobs. On the other hand, the study doesn’t deliver the productivity increases AI firms want to deliver with chatbots like ChatGPT.
We are still in the early years of AI, and we’re all getting used to it. It’ll be interesting to see what happens in the next few years when products like ChatGPT and Gemini start to get agentic capabilities that allow AI products to do a lot more for the user.
The post Study finds ChatGPT isn’t taking anyone’s job, at least not yet appeared first on BGR.
