# How Scientists Use Claude: 207,000 Sessions Studied

> Anthropic studied 207,049 scientist sessions on Claude. Data analysis and writing dominate, ideation lags. The numbers and the speed-up scenarios.

- Author: Nitish Garg, Founder & CEO, CellCog
- Published: 2026-11-05
- Canonical (HTML): https://cellcog.ai/blog/anthropic-how-scientists-use-claude/
- Section: Guides / Workflows & use cases
- Publisher: CellCog (https://cellcog.ai), the AI employee platform. Blog index for agents: https://cellcog.ai/blog/llms.txt

## Key points

- Anthropic published its study of how researchers use Claude on November 5, 2026, measuring 207,049 research sessions on Claude and Claude Code from August 2 to September 30.
- Data analysis and writing dominate: data analysis is 33.7% of Claude sessions and 48.1% on Claude Code, against 27.6% of researcher time.
- Scientists use Claude least for ideation and data gathering, the two tasks they rank most important for producing an important result.
- Computer science is 4.3% of the world's researchers but 32.7% of Claude Code research sessions.
- On Claude Science, ideation is 8% of sessions and over 70% run with high AI autonomy, against 23.7% on Claude.
- With today's AI widely adopted, Anthropic estimates research output could roughly double in statistics and computer science, while lab sciences stay capped by physical work.

## At a glance

- **What did Anthropic study?** 207,049 research sessions on Claude's consumer plans from August 2 to September 30, 2026, labelled by field and task, plus a survey of 528 scientists and a model of research speed-ups.
- **What do scientists use Claude for?** Mostly data analysis and writing. Data analysis is 33.7% of Claude sessions and 48.1% on Claude Code; ideation is 4.2% and 2.1%.
- **How much faster could research get?** With today's AI widely adopted, about 2x in statistics and computer science and about 1.4x in chemistry. Anthropic calls its scenarios illustrative, not forecasts.

Anthropic has published its first large measurement of how scientists actually use Claude, and the answer is lopsided. Researchers hand Claude the work it is good at, data analysis and writing, and keep the work they say matters most, coming up with ideas and gathering data, largely to themselves. The report came out on November 5, 2026, from a team led by Johannes Hermle, Ben Jones and Lukas Althoff.

It rests on 207,049 research sessions on Claude's consumer plans from August 2 to September 30, 2026, a survey of 528 scientists, and a model of how much faster research could go. Anthropic's one-line summary: "In short, scientists appear to use AI where it works best."

## How Anthropic measured it

A language model read each session and labelled it with a field and a research task from SciNet, a taxonomy of 7,262 tasks across 34 fields and 318 subfields. Only those labels were kept, and results are reported in aggregate, only for cells with a minimum number of distinct accounts. Of the 207,049 sessions, 128,914 came from Chat and Cowork (which the report calls Claude) and 78,135 from Claude Code. A smaller sample from Claude Science, Anthropic's research tool for scientists, was studied separately.

## Where scientists use Claude

![Animated diagram: a robot arm lifts analysis, writing and code blocks into a tray marked AI while flask and microscope blocks stay on the belt labelled Lab work stays human, then bars show output with today's AI at 1.96x for computer science and 1.41x for chemistry](https://cellcog.ai/blog/media/anthropic-how-scientists-use-claude/steps.svg)
*Scientists hand Claude the thinking steps; lab work stays human and caps the gain.*

On Claude, the largest fields by session share are computer science (11.9%), engineering (8.4%) and medicine (7.7%). Size explains much of that, but not all. The formal sciences, computer science, mathematics and statistics, use Claude far more than their headcount predicts, while chemistry and medicine, where work happens in a lab or with patients, use it less.

Claude Code concentrates even harder. Computer science is 4.3% of the world's researchers but 32.7% of Claude Code research sessions. The artificial intelligence and machine learning subfield alone is 1.2% of researchers, 5.0% of Claude sessions and 15.2% of Claude Code sessions.

## What they use it for

*Table: Share of sessions vs share of researcher time, by task (Anthropic, November 5, 2026)*

| Task category | Claude sessions | Claude Code sessions | Researcher time |
|---|---|---|---|
| Data analysis | 33.7% | 48.1% | 27.6% |
| Writing and communication | 32.0% | not reported | 16.0% |
| Data gathering | 11.5% | 11.9% | 28.6% |
| Ideation and hypothesis generation | 4.2% | 2.1% | 7.1% |

The gap is the story. In the survey, scientists ranked ideation and hypothesis generation as the most important category for producing an important research result, and data gathering second. Those are the two categories they use Claude for least. Anthropic offers reasons: much data gathering is physical, scientists may prefer to generate ideas themselves, ideation is hard to classify and may be undercounted, and AI may be genuinely weaker at choosing which question to pursue.

Specialized tools change the mix. On Claude Science, ideation is 8% of sessions, roughly in line with the time researchers spend on it. Over 80% of Claude Science use is estimated to need doctoral-level training, against about 50-70% on the other interfaces, and over 70% of its sessions run with a high degree of AI autonomy, against 23.7% on Claude and 64.1% on Claude Code.

## Use follows capability

Anthropic built a score for how well AI can, in theory, perform each research task. For tasks in the top 20% on that score, such as visualizing data and running statistical tests, Claude use is about twice what researcher time would predict on Claude and about four times on Claude Code. Tasks in the bottom 20%, like purifying reaction mixtures, barely get used at all.

The score tracks how easy a task is to train with reinforcement learning, where success is easy to check and feedback is fast: the correlation is 0.78. It rises with cognitive work (0.69) and falls with physical work (-0.57), interpersonal work (-0.31) and regulatory limits on AI (-0.38).

## How much faster research could get

The report then models what AI could do to research output for a fixed budget, in three scenarios. "These are illustrative scenarios, not forecasts."

*Table: Anthropic's three research acceleration scenarios (November 5, 2026)*

| Scenario | What AI does | Fastest fields | Slowest fields |
|---|---|---|---|
| 1. Today's AI, widely adopted | Takes on 27-52% of researcher time, median 38% | Statistics 2.02x, computer science 1.96x | Arts 1.35x, anthropology 1.36x, chemistry 1.41x |
| 2. 75% of non-physical steps automated | Automates the highest-scoring cognitive steps | Computer science about 8x, statistics 7x, economics 6x | History, anthropology and the arts about 2x or less |
| 3. All non-physical steps automated | Takes on nearly all cognitive research work | Statistics, computer science, economics, mathematics, psychology about 15-16x | Biology, chemistry, materials science, biomedical sciences 3-4x |

Two things cap the gains. Physical work is the first: in lab and field sciences such as chemistry and materials science, 38-41% of researcher time goes to physical steps, against 4% in the median field. "Absent automated labs, their physical steps, representing about 40% of researcher time, remain the bottleneck." The second is cost. The model assumes an hour of AI agent time costs $30 against $95 for a researcher's hour, with AI running 0.2 hours for each researcher hour it saves, based on today's public list prices.

The survey points the same direction: about 40% of researchers who actively use AI "expect AI models to carry out full research projects on their own by 2030." Anthropic also flags the risk side, from less human oversight of what gets researched to changes in what a research career feels like, and notes that the fields with the biggest gains include those that build AI itself.

## Where CellCog fits

CellCog is our product, an AI employee platform, and this study maps closely onto how our employees divide work with the people who hire them. An [AI research assistant](https://cellcog.ai/ai-employees/ai-research-assistant) takes the cognitive, checkable steps the report says AI does best: reading the literature, running the analysis, building the charts and drafting the write-up, while keeping its own memory of the project between conversations. The question, the judgment and anything physical stay with you, which is exactly where the study says scientists keep them today. For how Anthropic is widening its own science work, see our record on [Anthropic's AI for Science expansion](https://cellcog.ai/blog/anthropic-ai-for-science-expansion/).

## What is not established yet

- **Beyond consumer plans.** The sample covers consumer plans only; enterprise and API research use is not in it.
- **Ideation counts.** Anthropic says ideation is woven into other tasks and may be undercounted.
- **Classifier scores.** Capability scores, time shares and task labels all come from language-model classifiers, validated by Anthropic but not by outside researchers yet.
- **The scenarios.** Anthropic calls them illustrative, and they depend heavily on assumptions about AI's reach and cost.

## Where to watch

- Whether specialized tools like Claude Science keep pulling ideation share toward researcher time.
- Automated labs, the one change that would lift the physical cap in bench sciences.
- A repeat of this study on a later model, to see whether use keeps tracking capability.

## Sources

- Anthropic, [How researchers use Claude](https://www.anthropic.com/research/how-researchers-use-claude), November 5, 2026, read 18:50 UTC.
- Anthropic, [Economic Index](https://www.anthropic.com/economic-index), the usage research series this report sits beside.
- Anthropic, [Claude Code](https://claude.com/product/claude-code), the agentic coding tool behind 78,135 of the sessions.
- Our record on [Anthropic's AI for Science expansion](https://cellcog.ai/blog/anthropic-ai-for-science-expansion/).

## FAQ

**When did Anthropic publish the study?**

On November 5, 2026, on Anthropic's research site, by a team led by Johannes Hermle, Ben Jones and Lukas Althoff.

**How did Anthropic protect researchers' privacy?**

A classifier labelled each session with a field and task; only those labels were kept, and results are reported in aggregate for cells with a minimum number of distinct accounts.

**Which fields use Claude most?**

On Claude, computer science (11.9% of sessions), engineering (8.4%) and medicine (7.7%). Relative to headcount, the formal sciences lead; computer science is 32.7% of Claude Code sessions.

**What holds the speed-up back?**

Physical work. In lab sciences 38-41% of researcher time goes to physical steps, so even full automation of cognitive work lifts those fields only about 3-4x.

**Where does CellCog fit?**

CellCog is our product, an AI employee platform. An AI research assistant takes the analysis and writing steps the study says AI does best and leaves the question to you.

## Related

- [Anthropic AI for Science Expands to 20 Research Partners](https://cellcog.ai/blog/anthropic-ai-for-science-expansion/index.md)
- [Claude Science Maps the Whole Sky in Ultraviolet](https://cellcog.ai/blog/claude-science-uv-sky-map/index.md)
- [Anthropic's Three Numbers: Claude Leads 26% of Its AI R&D](https://cellcog.ai/blog/anthropic-rd-automation-index/index.md)

## The AI employee for this read

[AI Research Assistant](https://cellcog.ai/ai-employees/ai-research-assistant): I built this page. For what it covers, hire a research assistant: it reads the primary sources and cites every line.

---

Markdown alternate of https://cellcog.ai/blog/anthropic-how-scientists-use-claude/. Try CellCog free, no credit card needed: https://cellcog.ai/signup