ChatGPT, Gemini and Claude Disagree on Which Jobs AI Will Wipe Out

The three biggest language models, ChatGPT-5, Gemini 2.5 and Claude 4.5, produced sharply conflicting assessments of which professions and industries AI automation will hit hardest.

The models couldn't agree

Researchers asked each model to rate how exposed various professions are to artificial intelligence. The results diverged so sharply that they call into question the very idea of a universal exposure index.

ChatGPT flagged one set of professions as high-risk, Gemini picked an entirely different set, and Claude offered a third. The gaps were most visible for creative specialties, technical positions and management roles. One model placed programmers in the high-risk zone, while another treated the profession as relatively safe. The same goes for designers, analysts and consultants: the three models see their futures in completely different ways.

Why this is dangerous for policy and the economy

Ministries, governments and large corporations rely on these indexes to plan workforce retraining, education and social policy. When three authoritative models send contradictory signals, there is a real risk that public resources end up in the wrong place.

Imagine a government investing billions to retrain thousands of workers in certain fields based on ChatGPT's forecast, while the actual displacement happens in entirely different sectors, as Claude predicted. Workers would be retrained for a field nobody needs, and a social crisis would unfold at the same time in unprotected industries.

• Policymakers don't know which index to rely on when choosing which professions to prioritize for support
• Companies may misplan automation investments based on contradictory forecasts
• Educational institutions are unsure which skills to prioritize teaching

How this happened

One of the main reasons is that these models have different architectures and training data. ChatGPT, Gemini and Claude were trained on different text corpora and differ in size, so their assessments come out subjectively colored. Each model interprets what "AI exposure" means in its own way.

On top of that, exposure indexes are crude tools by nature. They try to reduce the complex reality of the economy and technological progress to a single number per profession. In practice, AI's impact depends on many factors: region, wage level, company specifics, the pace of technology adoption and governments' political will.

What this means

As long as AI is being asked to assess its own impact on the economy, the results will keep diverging. That doesn't mean the indexes are useless. It's just time to treat them as one signal among many, not as the final word.

Policymakers and employers need to combine data from different models, analyze local factors and avoid rushing into drastic decisions based on a single index.