AI in Academic Writing: Who Is Actually Using It and How? (2026 Data)

The AI in academic writing statistics for 2026 reveal a seismic shift: 92% of students now report using AI tools in their studies, up from 66% just two years earlier. But headline adoption figures mask a more complex reality. How students use AI — and for which writing tasks — varies dramatically by degree level, discipline, and institution type. This data roundup goes beyond the headlines to examine task-specific adoption rates, faculty responses, institutional policy adoption, and what the evidence shows about AI’s actual impact on writing quality and academic integrity.

The questions that matter for thesis and dissertation writers specifically are: which AI tools are students using for long-form academic writing, what do universities allow, and what does the emerging research show about outcomes? The data provides some clear answers — and some surprising nuances.

Overall AI Adoption in Higher Education

Multiple major surveys now track AI adoption in higher education. The key data points from 2025–2026 surveys paint a consistent picture of near-universal student engagement:

The jump from 53% to 88% generative AI use for assessments in a single year represents one of the fastest adoption curves ever recorded in education technology. For context, the internet itself took over a decade to reach comparable penetration in academic settings.

AI Use by Writing Task Type

The aggregate adoption numbers become more instructive when broken down by specific writing task. Research from Zendy’s 2025 AI in Research survey and the Frontiers in Education synthesis of 2023–2025 studies reveals distinct use patterns:

The 46% writing-and-editing adoption rate for body text production is particularly significant for thesis writers. This figure includes students at all stages, but among final-year doctoral students specifically, a 2025 survey by Zendy found usage rates as high as 67% — with the highest concentration in the early drafting phases where blank-page paralysis is most acute.

Adoption by Degree Level and Discipline

AI adoption varies meaningfully by degree level, with graduate students showing both higher adoption rates and more sophisticated use patterns:

• Undergraduate students: 89% aware of ChatGPT; primary use is essay drafts and assignment completion; lower use for learning or research
• Master’s students: 82% use AI for academic work; higher rates for literature review assistance and writing support; growing use of specialized academic tools
• Doctoral students: 74% use AI tools; more likely to use specialized tools (Elicit, Consensus, Tesify) rather than ChatGPT alone; concerns about academic integrity are higher but use is not lower

Disciplinary variation in AI use for writing shows a counterintuitive pattern. STEM students, who have more structured programs, are actually more likely to use AI for writing tasks — partly because their institutions adopted AI policies earlier and clearer guidance reduced ambiguity about permitted use. Humanities students report higher rates of uncertainty about AI policy, which correlates with both avoidance and covert use.

Which AI Tools Are Students Actually Using?

The most important distinction in 2026 is between general-purpose LLMs and purpose-built academic writing tools. Both categories have grown substantially, but their use patterns and integrity implications differ significantly.

General-Purpose Tools (Higher Risk for Academic Integrity)

• ChatGPT: Used by ~89% of AI-using students at some point
• Claude: Growing adoption, particularly among graduate students
• Google Gemini: Rapid growth since integration with Google Docs

Purpose-Built Academic Tools (Lower Risk)

• Tesify: Thesis-specific AI with source-based drafting and citation integrity
• Jenni AI: Focused on academic writing with citation integration
• Paperpal: Manuscript editing with academic language awareness
• Elicit: Research synthesis and literature review
• Consensus: Academic claim verification

The distinction matters because general-purpose LLMs hallucinate citations and fabricate research findings at significant rates. A 2024 Nature study found ChatGPT-generated academic citations were fabricated in 47% of cases. Purpose-built tools like Tesify are designed specifically to prevent this — working only with sources the student uploads, preventing fabrication at the source. See our full comparison of the best AI thesis writing tools for a feature-by-feature breakdown.

For Spanish-language students, our guide on Tesify vs ChatGPT for TFG writing covers the same comparison with Spanish university context. French-speaking students can consult our guide to anti-plagiat tools.

Faculty and Institutional Responses

Faculty responses to student AI use in 2025–2026 are considerably more nuanced than the early “ban everything” stance of 2022–2023:

• 61% of faculty have used AI in their own teaching preparation (up from 34% in 2023)
• 88% of those who use AI do so “minimally” — primarily for drafting communications, feedback templates, and course materials
• Only 34% of faculty consistently enforce AI-use restrictions in assignments
• 52% of faculty now explicitly permit AI for certain thesis tasks (literature searching, grammar checking) while restricting others (argument development, conclusion writing)

The shift toward nuanced permission frameworks — rather than blanket bans — reflects the empirical reality that blanket bans are both unenforceable and counterproductive. Students who use AI transparently and skillfully with purpose-built tools generally produce higher quality work than students who use general-purpose LLMs covertly.

The Policy Landscape in 2026

UNESCO’s 2024 survey found that two-thirds of higher education institutions globally have developed or are actively developing AI use guidance. The distribution by region:

The majority of institutions with policies permit AI for research assistance, grammar checking, and brainstorming while restricting it for argument generation and conclusion writing. Very few institutions maintain blanket prohibitions — and those that do report among the lowest rates of formal compliance.

What the Research Shows About Outcomes

The most recent research synthesis (Frontiers in Education, 2025, covering studies from 2023–2025) finds mixed but broadly positive outcomes from AI use in academic writing when used appropriately:

• Students using AI for grammar and editing show measurable improvements in written clarity without reduction in original thinking
• AI-assisted literature review reduces time spent on search tasks by 40–60% without reducing comprehensiveness when used with verified databases
• Students using AI for complete draft generation show reduced learning outcomes — as measured by retention and transfer tasks — compared to students who write independently with AI assistance only for specific bottlenecks
• The most consistent predictor of positive outcomes is intentional use: students who use AI as a scaffold for their own ideas outperform both students who use AI to replace their thinking and students who avoid AI entirely

This evidence base supports the approach taken by Tesify — providing writing assistance that works with the student’s research and arguments, rather than generating generic academic content. The data increasingly suggests this distinction is not just an academic integrity question but a genuine learning and quality question.

Frequently Asked Questions