AI Job Risk in Science
Science produces enormous amounts of text and data, and AI is already changing how much of that gets processed by hand. Literature review, data cleaning, and first-pass comparison of models or methods can now be done far faster than a researcher working alone. The tension is that generating a summary of prior work is not the same as knowing which question is worth asking, and running an analysis is not the same as knowing whether its result actually means what it appears to mean. Those judgment calls sit with a scientist's training and reputation, not with the tool that produced the draft.
Industry average exposure
76
Jobs Analyzed
9
How to read this page in practice
The notes below explain how to read the industry figure, where generative AI reaches the work first, and where human-led work is more likely to remain.
How to Read This Industry
Read AI's effect in science by separating information processing from the scientific judgment around it. Searching and summarizing literature, cleaning and preprocessing datasets, running standard statistical or model comparisons, and drafting routine lab records move much faster with current tools. Formulating a genuinely novel question, designing an experiment that actually tests it, and interpreting an ambiguous or surprising result do not move at the same pace, because they depend on domain judgment that a tool cannot originate.
What Automation Hits First
AI moves first through literature search and summarization across large numbers of papers, through data cleaning and preprocessing pipelines, through drafting and formatting of lab notebooks and routine reports, and through rapid first-pass comparison of models, parameters, or analysis approaches. Some experimental workflows already use automated instruments for repetitive sample processing. It stalls on the core of research: deciding which question is worth pursuing, designing an experiment or study that isolates the right variable, judging whether a surprising result reflects reality or an artifact, and taking responsibility for a claim published under a researcher's name.
What Still Depends on People
What stays durably human in science is originating and judging, not processing. Principal investigators and researchers framing which questions matter, experimentalists designing studies that control for the right variables, scientists interpreting ambiguous or conflicting results in context, and peer reviewers and mentors exercising judgment about what counts as a valid claim all depend on domain expertise and accountability that a model can assist but not replace.
How to Use the Gap
For science roles, weigh how much of the work is literature search, data processing, or routine analysis versus posing questions and interpreting results. Roles concentrated in literature review, data cleaning, or standard model comparisons should expect a higher score, since those tasks are already being compressed by current tools. Roles centered on original question framing, experimental design, or interpreting ambiguous findings should read a lower score as reflecting judgment that remains squarely human.
Jobs Most At Risk from AI
This table is a current snapshot of jobs in this industry that sit on the higher-risk side. Read it together with the fixed commentary above rather than as a permanent list of examples.
| Rank | Job | Generative-AI exposure |
|---|---|---|
| 1 | Sociologist | 90 |
| 2 | Meteorologist | 85 |
| 3 | Physicist | 84 |
| 4 | Astronomer | 81 |
| 5 | Biologist | 80 |
| 6 | Anthropologist | 80 |
| 7 | Chemist | 77 |
| 8 | Research Assistant | 71 |
| 9 | Geologist | 68 |
Exposure runs from 0 to 100 and compares occupations with each other, using public datasets (see sources). It is not a probability that the job disappears. Methodology
Sources
Exposure index calculated by ai-job-risk.net from the datasets below (data as of October 15, 2026). US labour figures are from the BLS, unchanged.
- Eloundou, Manning, Mishkin & Rock, "GPTs are GPTs: Labor market impact potential of LLMs", Science (2024); data MIT licence
- Gmyrek, Berg, Kamiński, Konopczyński, Ładna, Nafradi, Rosłaniec, Troszyński. Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140. Geneva: ILO, 2025. © ILO, CC BY 4.0
- Tomlinson, Jaffe, Wang, Counts & Suri, "Working with AI: Measuring the Occupational Implications of Generative AI", Microsoft Research; CC BY 4.0
- Anthropic Economic Index (job_exposure); CC BY
- U.S. Bureau of Labor Statistics, Employment Projections 2025–35 and AI exposure categories
The publishers did not prepare, review or endorse this site. Full citations, licences and method
Frequently asked questions
Q.Which jobs in Science are most exposed to AI?
Most exposed: Sociologist, Meteorologist, Physicist. AI moves first through literature search and summarization across large numbers of papers, through data cleaning and preprocessing pipelines, through drafting and formatting of lab notebooks and routine reports, and through rapid first-pass comparison of models, parameters, or analysis approaches. Some experimental workflows already use automated instruments for repetitive sample processing.
Q.Which Science jobs are safest from AI?
What stays durably human in science is originating and judging, not processing. Principal investigators and researchers framing which questions matter, experimentalists designing studies that control for the right variables, scientists interpreting ambiguous or conflicting results in context, and peer reviewers and mentors exercising judgment about what counts as a valid claim all depend on domain expertise and accountability that a model can assist but not replace.
Q.Is Science safe from AI?
For science roles, weigh how much of the work is literature search, data processing, or routine analysis versus posing questions and interpreting results. Roles concentrated in literature review, data cleaning, or standard model comparisons should expect a higher score, since those tasks are already being compressed by current tools. Roles centered on original question framing, experimental design, or interpreting ambiguous findings should read a lower score as reflecting judgment that remains squarely human.
Q.How is the Science figure calculated?
It is the average exposure index of the Science occupations we cover, weighted by US employment where BLS data exists. See the methodology page for the sources and how to read them.