According to an education report Anthropic published on August 27, 2025, professors and instructors use Claude most for developing curricula — more than any other task. The report is based on an analysis of roughly 74,000 anonymized conversations from higher-education professionals on Claude.ai, offering a rare empirical look at how educators actually use AI day to day.
What did the report study, and how?
Anthropic analyzed conversations from Claude.ai Free and Pro accounts associated with higher-education email addresses, drawn from May and June 2025. Using a privacy-preserving analysis tool that surfaces broad usage patterns without exposing individual conversations, Anthropic filtered for educator-specific tasks — such as writing a syllabus or grading an assignment — narrowing the dataset to about 74,000 conversations.
Each conversation was then matched against O*NET — a standardized database of occupational tasks maintained by the U.S. Department of Labor that breaks down what a given job actually involves — to identify which postsecondary teaching or administrative task it corresponded to. Anthropic also partnered with Northeastern University, surveying and interviewing 22 faculty members who are early AI adopters.
The report also cites a separate Gallup survey finding that teachers saved an average of 5.9 hours per week using AI tools. Worth noting: that figure refers to AI tools broadly, not specifically Claude.
Which tasks do professors use Claude for most?
Curriculum development accounts for the largest share of educator conversations by far, at 57% of the total analyzed.
| Task | Share | What it covers |
|---|---|---|
| Curriculum development | 57% | Syllabi, course materials, assignments |
| Academic research | 13% | Literature review, research support |
| Assessing student performance | 7% | Grading assignments and exams |
- Mock legal scenarios — case studies for hands-on courses like law
- Vocational and workforce training content — designing practice-oriented curricula
- Recommendation letter drafts — supporting students' applications
- Administrative documents like meeting agendas — handling routine paperwork
How do you decide which tasks to automate and which need a human?
The report frames educator AI use along a spectrum between augmentation — where AI assists but a human makes the final call — and automation — where AI performs the task directly. Tasks requiring context, creativity, or direct student interaction lean toward augmentation, while routine administrative work leans toward automation.
- Automation-heavy tasks — financial management, record-keeping, and other routine administrative work
- Augmentation-heavy tasks — designing lessons, advising students, writing grant proposals, and other work needing context and creativity
Interviews with Northeastern faculty surfaced why they draw this line.
- Automating tedious work — to cut down repetitive workload, as one faculty member put it, AI "takes care of the tedious tasks"
- A collaborative thought partner — for developing ideas together, as one put it, "AI can find effective ways to explain concepts to students that I had not thought of myself"
- Personalized learning experiences — for building individualized materials, as one put it, "AI is useful for giving students and me individualized, interactive learning experiences beyond what one instructor could provide"
Is it okay to let AI handle grading on its own?
The report's own findings suggest caution. Grading and evaluation came up less often than other tasks overall, but when it did, 48.9% of the time faculty used AI in an automation-heavy way — letting it perform the grading directly. Notably, those same faculty rated grading as the area where they felt AI was least effective.
A caution worth flagging — a high automation share doesn't mean the practice is well-validated. The report itself notes faculty concerns about automating assessment. Before automating any grading, it's worth checking your institution's fairness and student-privacy policies, and keeping a human review step before grades are finalized.
Educators aren't just chatting — they're building tools
The report finds that educators go beyond simple question-and-answer use of Claude. Using Artifacts — a Claude feature that displays generated output (code, documents, visualizations) in a separate panel next to the chat, so it can be reviewed and edited directly — faculty have built chemistry simulations, automated grading rubrics, and data visualization dashboards.
Northeastern faculty also reported using AI for their own learning, averaging 29% of their AI time, though the report notes this wasn't captured in the Claude.ai conversation analysis, since it's difficult to distinguish this kind of use from student usage patterns.
Frequently asked questions
Q. How much data does this report cover?
Anthropic analyzed roughly 74,000 conversations from higher-education professionals on Claude.ai from May and June 2025, plus a survey and interviews with 22 Northeastern University faculty.
Q. Should I let Claude automate grading?
Proceed carefully. The report found that grading — an area faculty themselves rated as where AI is least effective — still saw a 48.9% automation-heavy usage rate. Check your institution's grading policy before relying on automation.
Q. Does this apply outside the U.S. higher-education context studied here?
The conversation analysis covered higher-education professionals globally, but the qualitative interviews were limited to Northeastern University faculty in the U.S. Institutional policies on grading, records, and privacy vary, so treat the overall patterns as a reference rather than a direct policy recommendation for your institution.
Q. Is this report useful if I'm not a professor?
Yes. It also covers administrative use cases like admissions and financial planning, which can be relevant for university staff handling repetitive administrative work.
For related reading, see our guides on using Claude Artifacts and Claude.ai pricing plans.
Related articles
- How to Use Claude: A Step-by-Step Guide for Beginners
- How to Use Claude Projects: Knowledge & Instructions to Cut Repetition
- How to Use Claude Code: Real Usage After Install (Commands, VS Code)
- Claude Artifacts: How to Create, Edit, Publish & Share (2026)
- How Canada Uses Claude: Findings from the Anthropic Economic Index