Three Study Smart 3 workshops delivered to:
- Computer Science & Engineering students
- Education students
- Mixed cohort (including Nursing students)
AI Tool(s) Used:
Students selected their own preferred AI tools. The most frequently observed were ChatGPT and Microsoft Copilot, alongside Claude, Gemini, Grok, Kling AI, Padlet AI and Canva.
AI Mode(s) Used:
- Text → Text (story generation)
- Text → Image (movie poster/cinematic scene)
- Text → Video (movie trailer)
What was the challenge?
Rather than demonstrating what generative AI can do through a lecture, the aim was to understand how students already use multimodal AI within their studies and to encourage them to explore tools beyond those they normally rely on.
Most students were already familiar with text-based AI for coursework, but fewer had experience connecting multiple AI modalities into a single workflow. The workshop therefore explored whether students from different academic backgrounds approached multimodal AI differently and what practical or ethical challenges emerged during use.
The findings also contributed evidence towards the wider Study Smart 3 project.
What did you do?
Each workshop followed the same structure.
- Introduced the Study Smart 3 project and explained that anonymised findings would contribute to university research into student AI use.
- Discussed participants’ existing experience with generative AI, including:
- which AI tools they regularly used;
- how AI fitted into their coursework;
- what types of assignments they normally completed.
- Introduced The Prompetition, where students were challenged to:
- create a science-fiction story based on the theme;
- generate a cinematic poster or image from that story;
- create an AI-generated movie trailer where possible, with a poster serving as an alternative if video generation was unavailable.
- Encouraged students to use the AI tools they already knew while also experimenting with at least one unfamiliar tool during the session.
- Asked students to upload their prompts, workflows and final outputs to a shared Padlet board.
- Facilitated discussion around students’ experiences using different multimodal AI tools.
- Ran a Kahoot quiz exploring academic integrity and ethical AI use through realistic scenarios.
- The third workshop integrated an AI-versus-human image comparison activity to prompt discussion around realism, trust and AI-generated content.
Multimodal Use
The activity intentionally required students to combine multiple AI modalities.
Students first generated a written narrative before transforming it into visual content and, where possible, into an AI-generated movie trailer.
Each stage depended on the previous output, encouraging students to think about consistency between story, imagery and video.
Rather than remaining within one AI ecosystem, many participants naturally adopted a multimodal workflow by moving between different AI platforms depending on which task each tool performed best.
What did this look like in practice?
Padlet submissions showed considerable variation in students’ workflows:
Representative examples included:
- “We used ChatGPT to generate the story by providing it with an image of the slide containing the instructions. Based on that, it created the narrative. We then used Gemini to generate the image, using the description of the main character provided by ChatGPT as the prompt.”
- “We used Claude to create the story then used Grok to generate the video.”
- “This was created using Copilot. The prompt used was ‘create a 500-word story about a space cowboy looking for his lost love amongst the stars’. The image was also created using Copilot where the story was copied into the AI chat and asked to create an image from the description. Some edits were made after by having Copilot produce more details within the story.”
- “Claude to prompt the story creation and perplexity was used to generate the image”
- “I used Copilot and Canva – the comment I used was: Create a short 500 word sci-fi story about a society that has one…”
- “Used ChatGPT for the story, and Padlet AI for the pictures.”
- “We used ChatGPT for the images and Kling AI for the videos.”
These examples illustrate a consistent pattern of tool-hopping, where students combined multiple AI systems instead of relying on a single platform throughout the activity.
Supporting evidence included:
- Padlet submissions
- Student prompts
- Generated posters
- AI-generated trailers
- Kahoot results
- AI-versus-human comparison activity
What was the impact?
The workshops generated high levels of engagement, with students actively experimenting with AI tools and sharing their workflows. Several patterns emerged across the three cohorts.
Differences between disciplines
Computer Science and Engineering students appeared more familiar with recently released AI technologies than Education students.
Although their coursework generally did not require creative outputs such as movie trailers, a possible explanation is that programming-heavy modules expose them to more advanced AI-assisted development tools, many of which already include multimodal capabilities. As a result, applying these tools to creative tasks seemed like a natural extension of existing practice.
Education students most reported using Microsoft Copilot and often described it as an extension of web searching rather than a conversational AI assistant. Fewer participants reported using mainstream standalone chatbots such as Claude or Perplexity, although several did regularly use ChatGPT.
These differences represented overall trends rather than strict divisions, as some Computer Science students also relied exclusively on ChatGPT.
Awareness of specialist tools
Most participants across all cohorts were unfamiliar with specialist multimodal generation tools such as Grok and Kling AI, with only a small number experimenting with them during the activity.
Image generation preferences
Although only one participant used Canva, its inclusion suggested that some students perceive dedicated creative platforms as more suitable for image generation than chatbot-based image creation.
Prompting skills
Perhaps the strongest observation was not differences in tool access but differences in prompting ability.
Many students attempted to complete the entire task using a single prompt, asking AI to write the story, generate the poster and create the trailer simultaneously.
Relatively few participants adopted an iterative workflow where each output informed the next stage. This suggested that while students were comfortable accessing AI tools, many were less familiar with conversational prompting techniques and how prompt refinement influences output quality. Interestingly, as the activity progressed, students naturally began tool-hopping, moving between different AI systems as they discovered the strengths and limitations of each platform.
Ethical or practical considerations
Several important considerations emerged.
Academic integrity
The Kahoot ethics activity encouraged discussion around where AI functions as legitimate academic support and where it becomes unauthorised assistance.
Equity of access
An important finding emerged during the mixed workshop.
Two nursing students highlighted that some disciplines have access to specialist healthcare
AI systems unavailable to other students. Participants described this as creating an uneven playing field and felt access to discipline-specific AI tools represented an emerging issue of educational equity.
Accuracy and bias
Discussions around hallucinations, misinformation and AI bias were reinforced through the image comparison exercise.
Data privacy
Student submissions were collected anonymously through Padlet and contributed only to aggregated findings within the Study Smart 3 project.
Reflections and Advice
What worked well?
- The multimodal workflow encouraged students to experiment naturally rather than simply observing AI demonstrations.
- Collecting prompts and outputs through Padlet provided rich evidence of authentic student workflows and highlighted patterns that would have been difficult to identify through questionnaires alone.
- Running the workshops as a fellow student appeared to encourage open discussion, with participants speaking candidly about which AI tools they did—and did not—know.
What would you refine?
- Future workshops would benefit from a short introduction to effective prompting before the creative activity begins.
- The largest capability gap observed was not access to AI tools but understanding how to structure prompts, refine responses and build conversations iteratively.
Advice for colleagues
- Do not assume that greater exposure to AI tools necessarily translates into stronger AI literacy.
- Students who frequently use AI may still benefit significantly from guidance on prompting strategies and multimodal workflows.
- Similarly, expect students to move naturally between AI platforms rather than remaining within one ecosystem, and design activities that support this behaviour rather than restricting it.
Further reflections
- The workshop demonstrated that multimodal AI activities can generate valuable research insights as well as engaging learning experiences.
- Perhaps the most interesting finding was that students adapted organically to a tool-hopping workflow, selecting different AI systems for different stages of the creative process.
- This behaviour emerged without explicit instruction, suggesting that students increasingly view generative AI not as a single application but as an interconnected ecosystem of specialised tools.
Quick Start Guide for Colleagues
Estimated time required:
Approximately 3–4 hours to prepare slides, Padlet, Kahoot and workshop materials.
Digital skill level:
Moderate
Common pitfalls to avoid
- Not allowing enough time for AI video generation.
- Not providing a poster alternative if video tools fail.
- Assuming students already know effective prompting techniques.
- Forgetting to explain anonymity before collecting Padlet submissions.Best suited for
Best suited for
- AI literacy workshops
- Digital skills sessions
- Peer-led learning
- Cross-disciplinary cohorts
- Induction activities
- Creative AI or multimodal learning workshops
