Over the last three years I’ve been talking with groups of students to understand how they see artificial intelligence. How they use it. What excites them. What worries them. That work has led to three reports so far, in 2023, 2024 and most recently our June 2025 Student Perceptions of AI report.
In this blog, I want to take you behind the numbers and focus on a particular group of students who often face unique challenges: international students on pre-sessional programmes, preparing for their main academic studies in the UK. Between April and August 2025, I spoke with 77 students in three online pre-sessional programmes at the University of Manchester. The majority of these students were from China. This shaped both the tools they used and their experiences of access. Others from different countries and backgrounds brought valuable perspectives.
We met in three facilitated group discussions, shared ideas on Padlet, and I analysed anonymised transcripts for themes. Here’s what they told me about how they’re using AI in their academic and everyday lives. Their confidence in applying it. And what they expect from universities.
What felt different about pre-sessional students
Compared with the wider 2025 student cohort, this group of pre-sessional students:
- used AI more intensively for language support: grammar correction, vocabulary building, live translation of lectures, conversational practice
- adopted a wider mix of tools: global platforms such as ChatGPT, Grammarly and Copilot, alongside region-specific tools such as Doubao, DeepSeek, Kimi and Youdao Writing (especially common among the Chinese students who faced VPN restrictions)
- leaned on AI for very practical, UK-specific tasks: understanding visa rules, finding accommodation, opening a bank account etc
- worried about losing language skills if they relied too much on AI: questioning whether it might undermine the very purpose of studying abroad
- distrusted AI detection tools even more strongly than other groups: seeing them as biased and prone to false positives, especially against non-native English writing styles
- looked for proactive skill-building policies from universities: preferring guidance and capability-building over punitive approaches
- often saw AI as a potential equaliser: breaking down language barriers and supporting access, but only if issues of cost, access, and bias are tackled early
These differences really matter. Pre-sessional students rely heavily on AI to get through language, academic and everyday challenges. Policies, teaching approaches and student services need to reflect that from the very start.
How students said they were using AI
Academic and study-related uses
Language support dominated. Students told me they used AI constantly for grammar correction, vocabulary refinement, sentence restructuring and translation of technical or academic vocabulary. Some used it to check the flow and tone of their writing. Others explored the nuance of English words.
Real-time translation came up a lot. Students explained how they translated live online lectures or recordings, especially for complex terminology.
AI was not just for language. They used it to summarise articles, recommend references, generate essay frameworks, debug code, solve equations, and storyboard creative projects. A few students described “AI to AI” verification: checking one system’s answers against another, such as ChatGPT vs DeepSeek.
The diversity of tools was striking. While many used ChatGPT, Grammarly, Copilot and Claude, the majority-Chinese group leaned heavily on DeepSeek, Doubao, Kimi and Youdao Writing. Multi-tool workflows were common to balance out the limits of any single system.
Practical and everyday uses
Students also treated AI as an everyday assistant. It was their go-to for housing, visas, banking, transport and consumer choices. Some checked currency exchange rates before sending money. Others used it for trip planning. A few said they now asked AI instead of Baidu or Google because it gave faster, clearer answers.
For some, AI was also an emotional companion: a safe, always-available space to share worries or test ideas without feeling judged. One student told me, “I use AI as a place to vent so my family doesn’t have to carry my stress. It’s like a pressure-release valve, always there and non-judgemental.”
Language learning and communication
Many saw AI as a language tutor in their pocket. They practised spoken English, rehearsed presentations, checked pronunciation, and prepared for English language proficiency tests (including IELTS) with AI feedback. While they valued this immediacy, several worried that over-reliance could hold back the independent skill growth that is a key reason for studying abroad.
Decision-making support
Decision-making was another big theme. Students compared accommodation options, checked lease clauses, weighed career paths, and even asked AI to recommend university societies. Many said they trusted AI’s objectivity over human advice. Still, they sought confirmation from people on high-stakes choices.
Non-use and selective use
A small number deliberately avoided AI or limited it tightly. They wanted to strengthen their own skills. They were wary of trust issues such as hallucinated references. Some were unsure about the rules. For them, universities needed to provide low-AI pathways so avoiding AI does not mean falling behind.
Concerns they raised
Much of what I heard echoed the wider 2025 student cohort, but these pre-sessional students experienced it more intensely.
Loss of skills: the biggest fear was language erosion, but also critical thinking and subject knowledge if AI “did the work for them.”
Accuracy and misinformation: they had direct experience with hallucinations, mistranslations, or China-based tools giving advice that did not apply in the UK.
Erosion of learning opportunities: getting correct answers too quickly could short-circuit intellectual development.
Employability: STEM and business students saw AI literacy as essential. Arts and journalism students feared it might replace core creative or investigative work. All wanted more clarity on how AI is used in UK workplaces.
Equity and access: the cost of paid AI tools worried everyone. VPN restrictions were a China-specific frustration.
Trust and transparency: strong scepticism of bias in AI outputs, but even more distrust of AI detection tools. Students saw them as punitive and unfair.
Overdependence and wellbeing: those using AI as an emotional outlet worried it might reduce real-world interaction, human connections, and cultural integration.
Cultural and subject-specific limits: AI struggled with non-Western references or discipline-specific terminology. This created misplaced confidence in wrong answers.
What they said they needed
Students were clear about the support that would make a difference:
Clear, accessible guidance: in plain English, with examples of what is acceptable or not, and clarity on how pre-sessional rules differ from main programme rules
Integration into language learning: guidance on when to use AI for practice and when not to, to safeguard immersion
Practical, hands-on training: short, interactive workshops on prompts, translation checking, fact-checking and subject-specific applications
Support for UK academic expectations: side-by-side modelling of compliant and non-compliant uses
Early employability orientation: discipline-specific insight into how AI is used (or not used) in UK workplaces
Equitable access: reassurance that essential AI platforms would be freely available, with alternatives for those facing VPN restrictions
AI-literate staff: students wanted staff who could model good practice and integrate AI, not rely on detection systems
What this means for institutions
From what I heard, there are some clear priorities for universities:
- safeguard language development and integrity with plain-English, culturally relevant guidance
- close the access gap with licensed tools and support for VPN workarounds
- build AI literacy with practical workshops tied to real academic tasks and professional contexts
- support safe AI use in daily UK life, not just in academic work
- balance AI with peer and cultural integration to strengthen wellbeing
- raise staff capability so teachers can model good AI use and reduce reliance on detection tools
Final reflections
Pre-sessional international students are approaching AI with creativity and adaptability, but also with very immediate needs. For many, AI is not just a study aid. It is a cultural and linguistic lifeline: helping them translate lectures, prepare for English language proficiency tests (including IELTS), find housing, or open a bank account.
The Chinese majority in this group relied heavily on local tools such as DeepSeek, Doubao, Kimi and Youdao Writing, alongside ChatGPT, Grammarly and Copilot. That mix of tools and experiences will not be identical for every international cohort, but the patterns are instructive.
Their concerns, from language skill erosion to distrust of AI detection tools, echo the wider student body but feel sharper in this context. Their needs, accessible guidance, early employability orientation and staff who can model good practice, are also urgent. These students are navigating everything at once: a new language, a new culture, and a new technology.
What struck me most is that these students do not want to avoid AI. They see it as essential. But they want to use it well, use it fairly, and use it in a way that supports the very skills they came to the UK to build.
AI Tools Glossary
Here are some of the tools students mentioned in our discussions.
-
ChatGPT: Conversational AI. Used for text generation, summarising, brainstorming and Q&A. From OpenAI (US).
-
Grammarly: Writing assistant. Checks grammar, spelling, style and tone. From the US.
-
Copilot: Microsoft’s AI assistant inside Office apps. Helps with drafting, summarising and data analysis.
-
Claude: Conversational AI known for handling long text and natural responses. From Anthropic (US).
-
Gemini: Google’s multimodal AI. Used for research, summarising and translation.
-
DeepSeek: Chinese AI chatbot and search assistant. Used for translation, summarising and local content.
-
Doubao: Chinese large language model. Used for translation, summarising and Q&A.
-
Kimi: Chinese AI chatbot. Popular for quick queries, conversational practice and content generation.
-
Youdao Writing: Chinese writing assistant. Supports grammar correction, essay improvement and vocabulary expansion.
-
Grok3: AI chatbot with a humorous, conversational style. From xAI (US).
-
Baidu: Chinese search engine with AI-assisted results. Used for fact-checking and quick information.
-
Image generation tools (such as DALL·E, Midjourney, Stable Diffusion): Create images from text prompts. Mostly US and European origins.
-
Speech-to-text tools (such as iFlytek, Whisper): Convert spoken language into text. iFlytek from China, Whisper from OpenAI (US).
-
Translation tools (such as DeepL, Google Translate, Papago): Translate text or speech between languages. DeepL is from Germany, Google from the US, Papago from South Korea.
With thanks to
The University Centre for Academic English, University of Manchester.
Find out more by visiting our Artificial Intelligence page to explore publications and resources, learn more about our communities and sign up for our AI Literacy training.
For regular updates from the team sign up to our mailing list.
Get in touch with the team directly at AI@jisc.ac.uk