AI Prompt Library

The quality of the Generative AI output depends on the prompt that it is provided with.  

We recommend that you spend some time thinking about a prompt before you enter it into Generative AI.  Good prompt design can also lessen the environmental impact of Generative AI - if you have a good prompt design to start with, they you don't need to keep asking questions. 

Below are seven tips and ideas on how to refine your prompt to make sure you get the most out of it: 

1. Be clear and specific 

Think about what it is that you are trying to do and be as specific as possible to make sure you get the best possible output. If you’re using Generative AI to generate ideas, be clear about what ideas you want and what you don’t want.  

2. Set the context 

Provide some context with your prompt – for example, give background to the problem that you’re trying to solve. Provide your prompt with some additional background. 

3. Define the format 

What output do you want from your prompt? Is it a list or a project plan? Is it a summary?  How many points do you want your output to have? How many words do you want it to be? If it’s an image, specify the format. 

4. Use role instructions  

Provide the prompt with a role. For example, as a press officer, provide me with a summary press release for the education and student experience conference. Make clear who the prompt is for to help generate the correct tone from your output. You might want to think about using generative ai to create a more inclusive and accessible output. 

5. Refine the suggestions 

Don’t settle for the first output that Generative AI provides you with; make use of the refine features at the bottom of the prompt output to further refine and produce an output you are looking for.  

6. Use examples 

Provide examples of good outputs or content structures that you want the output to align with.   

7. Stay ethical 

As generative AI outputs are based on large language models, biases are produced. Encourage challenge these biases in the prompts that you provide. For example, if you are discussing nurses, use the pronouns he / him. If you are discussing mechanics, use the pronouns she / her. 

Prompts in the Welsh Language 

Prompts translated from English into other languages do not produce the same outputs as the original English prompts.  Researchers have found that English prompts often produce more accurate and comprehensive outputs than the same prompt translated into another language.  This includes high-resource languages – languages with many speakers that have a large presence on the internet – and low resource languages – languages with a much smaller presence on the internet.  However, outputs in languages other than English are improving as AI models improve.  Some models perform better in low resource languages than others. 

It is possible to prompt AI models in English and to ask for the output to be in a different language.  However, one study found that this produced significantly less accurate outputs than writing the prompt in the same language as the output, even for high-resource languages (German, Italian and Spanish).  That is, writing a prompt in English and asking for the output in German produces a less accurate output than writing a prompt in German and receiving a German output automatically.  

Welsh is a low resource language.  There has not been a great deal of research on engineering prompts in low-resource languages to improve AI outputs.  However, in addition to the general tips above, the following strategies may be helpful: 

• If you want an output in Welsh, a prompt in Welsh may provide a more accurate output 

• Asking for outputs in simple language may improve linguistic accuracy – for example, avoiding hallucinated words which appear when the model cannot find the correct word and fills the gap by creating a word.  However, such outputs will have less depth. 

• If an output includes a word that is unfamiliar, do a follow-up prompt asking the model to search for dictionary definitions and/or evidence of its use.  If no dictionary definitions exist, it is likely that the word is a hallucination, although the model will continue to assert that it exists. 

• You may need to be more specific in Welsh in order to get what you want – for example, by noting if you would like the model to use both English language and Welsh language sources or by specifying the cultural context. 

• If you are generating outputs where it would be useful to have them in Welsh and English, such as keywords, specify that you would like the outputs in both languages.  (Some models and topics will produce bilingual outputs, but not all). 

• If you want to use AI assisted machine translation, do not use a chatbot such as ChatGPT or Microsoft Co-Pilot.  Specific AI translation tools such as Microsoft Translate or Google Translate are more accurate.   

A principle driven, flexible prompt library that works across most AI tools, supporting student learning, critical thinking, and academic integrity. 

1. Core Philosophy

Principle Driven Prompts 

Prompts should focus on what librarians and students need to achieve (e.g., evaluating sources, building search strategies), rather than how a specific AI tool operates. 
This approach ensures longevity, adaptability, and alignment with information literacy principles. 

Student Level Modifiers 

Prompts may require adaptation depending on the academic level: 

• UG – foundational explanation, guided examples, accessible language 
• PGT – deeper analysis and more focus on methods 
• PhD – advanced critical thinking and specialised subject knowledge 

Accessibility Modifiers 

Modifiers for students with accessibility needs should also be considered, for example:

• Providing plain language versions 
• Reducing cognitive load 
• Adding step by step structures for neurodivergent learners 

2. Example Use Cases for Prompts

Prompts may be used to help librarians or students:

• Evaluate resources 
• Generate search terms 
• Check for bias 
• Provide information in an accessible format 

3. Safeguards (To Include in Every Prompt)

To promote safe, responsible use of AI in academic contexts, prompts might include: 

• A disclaimer against fabricating citations 
• Instructions for cross checking with library resources 
• A bias awareness clause 

Some recommended phrases: 

• “Do not fabricate citations. If unsure, say so.” 
• “Cite only real, verifiable sources.” 
• “Suggest how the user can verify your claims using library resources.” 
• “Highlight any assumptions or limitations in your response.” 

4. Core Prompts

Below are some example prompts ready to cut and paste. 

Evaluating a Source Using Critical Frameworks 

“Evaluate the credibility of this source: [insert description or link] using frameworks such as CRAAP. Explain your reasoning step by step. Do not invent facts about the source; base your comments only on what is provided.” 

Comparing Scholarly vs Non Scholarly Sources 

“Explain the differences between scholarly, professional, and popular sources for students in [discipline]. Provide indicators they can use to evaluate each type. Avoid providing fabricated examples.” 

Generating Search Keywords & Synonyms 

“Generate a set of search terms, synonyms, and related keywords for the topic: [insert topic]. Organise them into themes or concept groups. Do not create citations; keep the focus on terminology.” 

Explaining Peer Review to Students 

“Explain the peer review process to a first year undergraduate in clear, accessible terms. Include why it matters, its limitations, and how to verify whether a source is peer reviewed. Avoid invented journal examples.” 

Providing Guidance on Academic Integrity 

“Explain how students can maintain academic integrity when using generative AI tools. Include good practice, risks, and strategies for verifying information. Do not give advice that would enable plagiarism.” 

Helping Students Understand ‘What Counts as a Source’ 

“Explain what types of sources are appropriate for academic work in [discipline]. Include primary, secondary, and tertiary sources. Do not invent specific works or authors.” 

Converting a Student Question Into Searchable Concepts 

“Help convert this student question into a set of searchable concepts: [insert question]. Break down the question, identify key ideas, and propose search term groups. Avoid producing citations.” 

Suggesting Ways to Verify AI Generated Content 

“Provide ways students can verify claims generated by AI tools when conducting academic research. Include strategies using library resources, critical evaluation, and cross checking with reputable sources.” 

Understanding Bias & Perspective in Sources 

“Explain how to identify bias or perspective in academic and non academic sources. Give general examples of indicators without inventing specific articles.” 

5. Prompts for Students with Accessibility Needs 

Plain Language Rewriting 

“Rewrite the following text in plain, easy to understand language while keeping the original meaning. Break complex sentences into shorter steps and remove unnecessary jargon. Highlight any terms that may still require explanation.” 

Neurodiversity Friendly Step by Step Guide 

“Turn this task into a step by step guide suitable for neurodivergent learners. Include short, clear steps, optional visuals or metaphors, and a summary of key points. Avoid overwhelming detail. Do not add content unrelated to the task.” 

Increasing Clarity and Reducing Cognitive Load 

“Rewrite this content to minimise cognitive load. Use short paragraphs, bullet points, clear headings, and consistent terminology. Identify areas where the original text may cause confusion. Do not introduce new information.”

General Activities 

Feedback 

Ask Microsoft Copilot for feedback on something that you’ve created. 

This might be an email that you’ve drafted, communications, or a plan for your service.  

1. “Provide me with feedback on [name of item]. I am looking specifically for feedback on [insert what you want feedback on].” 
2. "Provide clear, constructive feedback on this document. Highlight strengths, identify areas for improvement, and offer specific, actionable suggestions 
3. "Review this document as an editor. Comment on clarity, structure, coherence, tone, and flow. Provide detailed recommendations." 

Summary 

Ask Microsoft Copilot to provide a summary of a document.  

1. “Provide me with a summary of [the attached document]. What are the main points of the document? What actions do I need to take?” 
2. "Provide a bullet‑point summary of the essential information in this [the attached document]." 
3. "Summarise this [attach document] with a focus on implications for decision‑making, risks, opportunities, and recommended actions 

Accessibility and inclusion 

Ask Microsoft Copilot to create something that’s accessible.  

1. “Make suggestions for how I can make this [insert item] more accessible.”  
2. “Rewrite this [insert item] to make it more readable for people with dyslexia.” 
3. “Add meaningful headings and subheadings to this [insert item] to improve navigation for screen reader users.” 
4. “How inclusive is the language in this document [insert item]? Can you make any suggestions to make it more inclusive?".  

Idea Generation 

Ask Microsoft Copilot to generate some ideas.  

1. "Generate ideas without constraints around budget, practicality, or existing norms about [insert project]. Think big and imaginative." 
2. "Give me 20 different ideas for this [insert topic], all deliberately varied in approach, style, and complexity." 
3. "Generate a range of possible solutions to this [insert issue], from simple fixes to ambitious redesigns." 

Communications 

Ask Microsoft Co-pilot to generate some communications for your service, research, or education and student experience activities.  

1. "Create a message about [insert topic] with separate tailored versions for students, academic staff, and professional services staff." 
2. "Rewrite this [insert item] in plain language suitable for a wide higher‑education audience while keeping all key information intact." 
3. "Write a friendly and supportive update for students about [insert topic], highlighting what they need to know, any actions required, and where to get help.” 

Data Analysis 

Ask Microsoft Co-pilot to generate some analyses based on data sets that you provide it with. When uploading datasets, be sure not to include personal information and follow our guidance using Generative AI safely. Be aware of the University’s data protection information.  

1. "Provide an exploratory analysis of this [insert dataset]. Summarise its structure, key variables, overall patterns, and anything interesting or unexpected." 
2. "Generate descriptive statistics for this [insert dataset] (mean, median, range, variance, quartiles, etc.) and explain what they tell us." 
3. "Review this [insert dataset] for missing values, duplicates, inconsistencies, or formatting issues. Suggest or apply cleaning steps." 

Process reviews 

Ask Microsoft Co-pilot to review processes. 

1. "Turn this [insert process description] into a clear, structured process map with steps, decision points, and responsibilities." 
2. "Identify steps in this [insert process] that could be automated or supported by digital tools, and explain the benefits." 
3. "Analyse the strengths and weaknesses of this [insert process], including efficiency, clarity, reliability, and user experience." 

Infographic creation 

Ask Microsoft Co-pilot to create visual representations of information.  

1. "Transform this text into an infographic outline with headings, visuals, and a suggested layout: [paste text]." 
2. "Turn this [insert process] into a step‑by‑step infographic with clear stages, arrows, decision points, and concise labels." 
3. "Turn this [insert dataset] into an infographic highlighting the most important stats, trends, and comparisons.” 

Learning and teaching activities 

Lesson planning 

Ask Microsoft Co-pilot to assist with creating lesson plans and ideas for new topics.  

1. "Create a detailed university‑level lesson plan on [topic], including learning outcomes, key concepts, session structure, activities, timings, and materials." 
2. "Develop a lesson plan aligned with the following learning outcomes: [insert outcomes]. Include activity suggestions that support each outcome." 
3. "Design a flipped classroom lesson on [topic], including pre‑class materials, in‑class activities, and follow‑up tasks." 

Assessment reviews 

Ask Microsoft Co-pilot to review assessment practices. 

1. "Check how well this assessment [insert assessment brief] aligns with the module’s learning outcomes [insert learning outcomes] and suggest improvements." 
2. "Review the clarity of instructions in this assessment [insert assessment brief]. Identify confusing wording, missing details, or potential misinterpretations." 
3. “"Evaluate how authentic and relevant this [insert assessment brief] is to the discipline and real‑world practice. Suggest enhancements." 

Communications 

Ask Microsoft Co-pilot to create and review communications related to learning and teaching.  

1. "Draft a clear and concise communication explaining [learning/teaching topic], focusing on what students/staff need to know, why it matters, and what actions they should take." 
2. "Create a step‑by‑step instructional message that explains how students should complete this [insert task or activity brief], including tips, pitfalls, and examples." 
3. "Draft a supportive assessment reminder [insert assessment information] for students that includes key dates, expectations, support resources, and a reassuring tone." 

Module reviews 

Ask Microsoft Co-pilot to review modules. 

1. "Conduct a comprehensive review of this module [insert module outline and structure]. Evaluate its learning outcomes, structure, assessments, teaching approach, student experience, and data indicators. Provide strengths, weaknesses, and actionable recommendations." 
2. "Review the weekly structure and learning activities [insert module breakdown and overview]. Suggest improvements to engagement, pacing, and sequencing." 
3. "Evaluate this module for inclusivity and accessibility [insert module breakdown and overview. Identify potential barriers for diverse learners and recommend improvements." 

Scenarios and Case Studies 

Ask Microsoft Copilot to provide case studies or scenarios for essay questions, classroom activities, or exam questions.  

1. “Provide me with some sample prompts for case studies for student nurses at university to understand diabetes. The prompts should ask the nurses what actions they should take.” 
2. “I’m testing the students knowledge of [insert topic]. I’m looking for a case study that will test this” 
3. “I’m writing an exam question on [insert question or topic]. Provide me with a scenario to test students” 

The following prompts are given as a guide for researchers who want to use AI to support, but not replace, the research process. They are written to encourage critical appraisal and transparency, but please be advised by your departmental guidance on AI use. Also see the guidance on ethical use of AI.

1.     Refining research questions:

1. “Compare the methodologies of ______ and list the advantages and limitations.”
2. “Based on the current research in _______, suggest two potential research gaps that could be explored.”
   
   

2.   Reviewing the literature for systematic reviews:

1. See the prompt guidance on conducting literature reviews for students. 
2. “I am conducting a Systematic Review on [topic]:
   1. Identify key concepts using the PICO framework
   2. Generate a list of synonyms and related terms for each concept
   3. Base suggestions on terminology used in academic research”
3. “Based on the concepts and synonyms below, create draft search strings for the databases: Scopus, CINAHL, Medline, PubMed [include any others].
   1. Use Boolean operators correctly
   2. Respect the database’s syntax (e.g. field codes, truncation, adjacency operators)
   3. Show each database string in a separate section
   4. Do not invent adjacency operators for platforms that don’t support them
   5. Include notes on terms or limits that need checking by a Subject Librarian
4. “Analyse the following search concepts and suggest additional synonyms, older terminology, and spelling variations that might appear in academic literature [add concepts]
5. “Identify controlled vocabulary terms for the following concepts: [insert concepts].
   1. Do not invent terms. If unsure, say so.
   2. Provide MeSH terms and CINAHL subject headings.
6. “Produce a description of how AI contributed to development of the search strategy, for inclusion in a Systematic Review’s methodology section. Include:
   1. What prompts were used
   2. What AI generated
   3. What was refined
   4. The limitations of using AI

3.     Data collection:

1. brainstorming around data collection or methodological design:

“I am designing a research study on [topic]. My research question(s): [insert questions]. Can you suggest appropriate data collection methods and research designs? Please include:

• Pros and cons of each approach
• When each method is most suitable
• Any potential limitations or biases

1. b. generating questions for interviews / focus groups / questionnaires:

“I am conducting a study on [topic] using [interviews / focus groups / surveys]. My research objectives are: [insert objectives] Please generate a set of high-quality questions that:

• Align with my objectives
• Avoid leading or biased phrasing
• Are suitable for [target population]

Group the questions by theme if possible.

4. Data analysis

1. “Explain which statistical tests are most appropriate for analysing [type of data] and why.”

1. “I am using [statistical test, e.g. t-test, ANOVA, regression] for my data. Can you explain:

• • • When this test is appropriate
    • Its assumptions
    • How to check those assumptions
    • Common pitfalls

Explain in a way suitable for a researcher with basic statistical knowledge.”

1. Getting step by step guidance for statistical tests:

“I have a dataset with the following characteristics: [describe variables, sample size, type of data] (never include the original data, that is not ethical practice). I want to test [research hypothesis]. Can you guide me step-by-step on:

• Which statistical test to use
• How to perform it (in software like R/SPSS/Python)
• How to interpret the results

5. Drafting research proposals:

“Draft a 400 word grant proposal summary on_[Insert your research topic]_research. Include objectives, methodology, and expected impact.”

6. LaTeX code assistance:

“Write LaTeX code for a table summarising survey results from three countries with columns: Country, sample size, key findings.”

2. Publication:

1. Finding possible journals for submission

“Here is my paper abstract: [insert abstract]. Can you suggest suitable academic journals for submission, if these are the publishers we have transformative agreements with in this area [insert relevant ones from the uni's webpage)?.  Include:

• • • Journal scope and audience
    • Impact level (approximate)
    • Whether my topic is a good fit”

1. To check suitability of a paper for journal:

“I am considering submitting to [journal name]. Here is my abstract: [insert abstract]. Based on the journal’s scope and typical publications, how well does my paper fit? Suggest any changes to improve alignment.”

1. Framing work for better target journal suitability:

“I am preparing a paper for submission to journals in the field of [field]. Here is my abstract: [insert abstract]. Can you suggest how I could better frame or position this work to align with high-impact journals in this field? Comment on:

• • • Novelty
    • Contribution to the field
    • Target audience
    • Suggested improvements to framing

 For further guidance please see: Embracing AI with integrity - UK Research Integrity Office