Finding and using knowledge

Take responsibility

AI tools are not human - they generate responses based on patterns not expertise or lived experience - so you are responsible for checking the information it gives you and what you do with that information.

Many AI tools will include a disclaimer advising you that the information it provides may be incorrect, even though it sounds convincing. The companies who own these tools take no responsibility for incorrect answers. If you use an AI tool to write an email and don't check it, you can't blame the AI tool when someone challenges you.

Entering sensitive, confidential or personal information into AI tools is not a good idea as this data may not be kept private and may even be used in responses to other users. It is possible to limit the information these tools have access to, for example some versions of Microsoft Copilot can be restricted to searching a company's internal data rather than the world wide web. This can offer more control over the data it has access to, but it is up to you to check this.

If you're not sure if you should be using AI to do something, refer to your organisation's policies and/or consult your manager. You can also check whether professional bodies in your field have created any guidance as many Royal Colleges have. It also helps to be transparent about when and how you are using AI.

What you can do: 

• Only use AI tools authorised by your organisation that are designed to keep data safe
• Refer to organisational guidance or guidance from professional bodies in your field
• Check disclaimers, read about what the tools do and how they use information
• Be aware that this technology is still emerging, it will make mistakes
• Ask it to explain it's reasoning - question the conclusions and how they were reached. See below: Writing effective prompts
• Cross-reference information with authoritative, human-authored sources

Be realistic

The answer AI tools give you is rarely the whole thing or a complete solution on the first try. Sometimes it's a good start or end, requiring significant human revision and refinement to prompts and outputs.

Don't underestimate the time needed. You may think it will save you time asking for a list of references - but you can't trust the references are correct or even real so it will take you more time to check them than it would have to collect them yourself in the first place or to ask a human to do this.

What you can do:

• Use AI to explore ideas and generate initial drafts
• Experiment, iterate, and adjust your prompts to refine the results. See below: Writing effective prompts
• Ask yourself whether using AI will save you time or whether you will end up spending more time checking outputs
• Ask an expert - human Librarians can provide in-depth and trustworthy literature searches and summaries for health and social care staff. See below: Trustworthy sources

Check other sources

AI should not be used in isolation. Most AI tools have a cut-off point on the information they are searching, so it can be out of date.

They also can't access content that you have to pay for, such as information and data from research articles and journals, meaning you may be missing narratives from a significant body of evidence.

Use other sources - The Knowledge Network, organisational websites, NHS websites, government websites, academic journals or reputable news outlets - to verify facts, identify gaps, and find alternative perspectives.

What you can do:

• Ask AI tools to provide links to trusted sources as part of your prompts - and check them because they can be made up
• Use AI customisation (like ‘personalisation’ or ‘agents’) to tailor the type of content it gives you back
• Break down the AI's claims into specific facts to search for and compare across multiple, diverse sources - The Knowledge Network can be a good starting point for trusted evidence and sources. See below: Trustworthy sources
• Use your own knowledge - AI is a useful support if you are knowledgeable in the area you are asking about and able to assess its responses

Engage ethically

AI is not an objective tool - outputs are generated through a process based on vast datasets which can contain biases and be impacted by geopolitics, meaning it can discriminate against or exclude some groups of people. For example, in searching the content of the PubMed database you would need to be aware this was impacted by a United States government shutdown in October 2025.

It may also inadvertently reproduce copyrighted text, images, or code - content that belongs to other people without their consent and from which they would usually make a living.

What you can do:

• Understand how different prompts and underlying data influence the output
• Understand and respect intellectual property rights and academic integrity standards when using AI-generated content
• Check the rules regarding using AI to summarise copyrighted material. See: AI and copyright | Copyright information for NHS Scotland staff
• Attribute original creators where possible, avoid unauthorised reuse, and ensure that any work submitted represents your own effort and critical analysis

Be open

AI is not the whole conversation but it is a tool that can be useful and it is attractive because it is quick and seemingly easy. You are allowed to use it but should tell people when you use it for work tasks. 

The general public will use AI for all kinds of things and it will sometimes give them incorrect information, which can present a challenge for health and social care staff - how can you argue with the results while being compassionate, understanding, and not coming across as patronising. If you dismiss their concerns, you risk them doubling down on the misinformation because it is what they want to hear and is easier to digest than the often complex information they are being given by you. You need to know how to build trust and find trustworthy alternative sources to engage in these conversations.

What you can do:

• Engage deeply with broader information on a topic to build a nuanced understanding
• Critically evaluate the strengths and weaknesses of the AI's information
• Support your arguments by citing a wide range of credible sources as evidence
• Use health literacy techniques to communicate compassionately. See: Health literacy | Finding and using knowledge

When to use AI

What is it good for

• Starting from a blank page
• Streamlining administrative tasks
• Woking in areas you are knowledgeable about so you can confidently assess outputs
• Summarising long, technical and/or publicly available documents - though you should always review the summary for accuracy
• Pulling together lots of different elements for presentations, promotional materials, etc

What it is NOT good for

• Clinical diagnosis - they are not a substitute for a healthcare professional's expertise and cannot be relied upon to inform treatment decisions or determine patient care pathways. They also cannot be held accountable if they are wrong.
• Sensitive or confidential information or personal data - AI tools acknowledge that information you put in may be used to train the model. Once you put the information in, you lose control over where it goes and how it is used. This information could potentially be used in responses to other users.
• Summarising copyrighted material - there are limitations to whether copyrighted material such as journal articles can be entered into AI tools for the purpose of summarisation. See: AI and copyright | Copyright information for NHS Scotland staff

Writing effective prompts

Prompting is a skill that takes time to learn. The response you get to a prompt is often inconsistent - AI learns from you so the answer may be different every time you ask and will be different if someone else asks the exact same thing. This makes it difficult to reproduce. 

This is also a field where we expect things to change and advice to vary. However, currently it seems the more detail you can give - whether in the initial prompt or in follow up prompts - the more likely AI tools are to give you what you want. Whatever approach you use, you should still always check the answer as hallucinations can still happen no matter how rigorous you are.

Prompt tips

• Tell it to use UK English
• Tell it who you are if relevant e.g. job title and sector
• Specify an audience and a tone - formal vs playful, etc
• Specify a format - present the information in a table, presentation, infographic, etc
• Give context
• Tell it if you want it to only use particular sources
• Tell it if you're not happy with the first response
• Provide follow up prompts

Customisation - personalisation or agents

Personalisation settings in Copilot (and similar agent tools) let you define your preferred tone, format, and style of responses once, so the system consistently tailors its answers to you without needing to repeat those instructions in every prompt.

They essentially act as “default instructions,” shaping how the assistant behaves and responds across conversations.

COMING SOON - an example of detailed instructions that could be used to give you ideas.

Prompt frameworks

Frameworks can help you think about prompts in a more structured way, especially if you want to be transparent about and record your use of AI. One example is the CLEAR framework. The acronym stands for:

• Concise: focused and brief, while providing enough detail to result in useful outputs

• Logical: organise your instructions in a sensible order

• Explicit: be clear about what you want

• Adaptive: adjust your prompt to suit the context or task

• Reflective: Review and refine your prompt for better results

Lo, L. S. (2023). The CLEAR path: A framework for enhancing information literacy through prompt engineering. The Journal of Academic Librarianship, 49(4).

Using AI with browsers and search engines

Almost all browsers and search engines are now using AI tools in some way, particularly to generate summaries of search results. How much they tell you about how these tools are being used and what they are doing varies. There are a couple of things to be aware of so you can decide how you want to use them.

Pros

• Can provide a quick and easy answer for basic searches
• Can highlight sources from results you may otherwise not have noticed or thought to explore
• When used in a single system with a limited evidence base, can provide a useful starting point

Cons

• AI summaries are on by default, but it's not always obvious that they're AI generated
• Can put casual searchers off of delving deeper and checking sources
• High energy usage which has an impact on the environment
• For wide reaching search engines like Google, does not have access to subscription content so can miss out important data

Turning it off

You can turn off the AI summary and there are various routes to do this, depending on the search engine and browser you are using and how they interact, so check the settings or help pages. You may wish to do this because:

• You don't need a particularly wide search
• Saves energy - searching more widely takes more energy and has an impact on the environment
• Takes away the summary at the top of results and turns them back to the standard list of search results

Useful links

Introduction to generative artificial intelligence (AI) | Technology enhanced learning support hub

Artificial intelligence guidance for Library and Knowledge Services | The Knowledge Network

Safe use of artificial intelligence (AI) | Scottish Social Services Council

AI and copyright | Copyright information for NHS Scotland staff

Digital sustainability - impact of AI | Knowledge Services Sway

Principles for the use of AI in Further Education | JISC

Find out more about JISC's programme of work on AI in education: Artificial intelligence hub | JISC

Check with professional bodies in your field or discipline (such as Royal Colleges and Chartered Institutes) for guidance on using AI

AI literacy reading list

Archambault, S. G. (2023) 'Expanding on the Frames: Making a Case for Algorithmic Literacy.' Communications in Information Literacy 17(2), 530–553.

Proposes updating information literacy frameworks (such as the ACRL Framework for Information Literacy for Higher Education) to include algorithmic literacy, with practical teaching approaches for understanding how algorithms shape information.

Archambault, S. G., et al. (2025) 'Fostering AI Literacy in Undergraduates: A ChatGPT Workshop Case Study.' Library Trends 73(4), 443–475.

Evaluates a ChatGPT workshop, showing increased student confidence but gaps between perceived and actual understanding.

Archambault, S. G., et al. (2024) 'Ethical Dimensions of Algorithmic Literacy for College Students: Case Studies and Cross-Disciplinary Connections.' Journal of Academic Librarianship 50(3), N.PAG.

Identifies key ethical principles of algorithmic literacy and demonstrates how teaching interventions can build students’ critical awareness.

Baer, A. (2025) 'Investigating the "Feeling Rules" of Generative AI and Imagining Alternative Futures.' In the Library with the Lead Pipe, N.PAG.

Critically examines librarians’ emotional and professional tensions around generative AI and calls for more reflective, collective engagement.

Cox, A. (2024) 'Algorithmic Literacy, AI Literacy and Responsible Generative AI Literacy.' Journal of Web Librarianship 18(3), 93–110.

Reviews competing definitions of AI literacy and argues for a broader, responsibility-focused approach beyond technical skills.

Hicks, A. (2024) 'Already Invested Or Absolutely Incandescent? AI, Scholarly Publishing and Information Literacy.' Journal of Information Literacy 18(2), 1–5.

Explores how AI is affecting scholarly publishing and highlights the need for clearer guidance on authorship and transparency.

Hutchinson, E. (2024) 'Navigating Tomorrow's Classroom: The Future of Information Literacy and Inquiry-Based Learning in the Age of AI.' Journal of Information Literacy 18(1), 118–122.

Argues that AI should be integrated into inquiry-based learning, with librarians guiding ethical and effective use in schools.

Lo, L. S. (2024) 'Evaluating AI Literacy in Academic Libraries: A Survey Study with a Focus on U.S. Employees.' College & Research Libraries 85(5), 635–668.

Survey findings reveal limited AI literacy among library staff, alongside a clear need for training and ethical frameworks.