Generative AI

Generic application documents

Whichever generative AI tool you use, it does not know you well enough to write an insightful, authentic and tailored application document. What it does produce is likely to be generic, and inauthentic and may also look very similar to a document written for someone else who asked the same thing of the AI tool. Employers, recruiters and opportunity providers are likely to be aware of this too – nothing is stopping a recruiter from using generative AI to mock up a cover letter and using that to compare it against the ones being sent in.

Make sure you check over any documents written using generative AI as a drafting tool. Ask yourself, does this sound like me? Is this accurate? Is there anything else I'd like to add?

You should also look at our advice on our pages on CVs and resumes, cover letters and application forms.

Garbage in, garbage out

"Garbage in, garbage out" comes from computer science and roughly means that the quality of output from a computer system is directly related to the quality of input it receives.

This principle applies to writing prompts for generative AI, so if you're planning to use these tools it is important to learn how to interact with them effectively and be prepared to review and revise anything you produce with them.

When using generative AI models to generate text based on prompts, the quality, clarity, and specificity of the prompt strongly influence the quality of the AI's response. If the prompt is vague, ambiguous, or poorly constructed, it can lead to inaccurate or nonsensical outputs. Conversely, a well-crafted prompt that provides clear context, guidelines, and relevant information tends to yield more coherent and relevant generated content.

To get meaningful and accurate results from generative AI, it's important to put effort into constructing effective prompts that convey the desired context and expectations.

AI can "hallucinate"

If generative AI "hallucinates," it means that the AI system produces outputs that are not accurate representations of the input data. Hallucinations in the context of generative AI refer to instances where the AI generates content that is speculative beyond the information it has been trained on, sometimes leading to fictional outputs.

For example, if a language model is given an incomplete CV and is asked to write a cover letter, it could generate a detailed but incorrect statement about an internship or experience you haven't undertaken. This could be considered a form of hallucination.

Hallucinations can occur when the AI extrapolates from patterns it has learned and creates content that goes beyond the provided input.

Researchers and developers are working to reduce the chances of this happening in future products and updates, but in the meantime, it is crucial to ensure you thoroughly check any documents produced with generative AI to ensure it is accurate to your experience.

Not all AI is equal

ChatGPT is a hugely popular generative AI tool, and ChatGPT-3.5 is currently freely available to use. However, this model's knowledge and information are based on limited data and events.

As a result, ChatGPT-3.5 may not have information about up todate events, developments, or changes. If you ask questions about recent events or information, the model might not be able to provide accurate or up-to-date responses. It may tell you this, or it may generate an incorrect response.

Different generative AI tools may also have limitations. It is important to do your research about any tool you might use to ensure it is helping you produce good-quality documents. This means you should not rely solely on generative AI to do research or to produce documents.

Employer distrust

Some employers actively dislike the use of AI in application documents, and if they detect it might disregard your application altogether.

It is difficult to predict which employers use AI detection tools, but some organisations have been very clear about this, for example the UK government webpages for people applying to teacher training.

Use effective prompts

Generative AI can only generate content based on the information given to it, and therefore it is important to feed the programme prompts that will effectively generate content that meets your requirements. Tips on how to do this include:

Give the programme clear and specific tasks. Define your request clearly and be specific about what you want the AI to generate. Vague or ambiguous prompts can lead to unexpected outputs.

Vague: Generate some interview questions.

Clear: If I upload a role description, could you suggest some interview questions I could be asked at an interview for this opportunity?

Contextual information helps the AI generate more relevant content. For example, feed in your CV if you want to ask the programme to generate a tailored cover letter because otherwise, the AI has no context about your skills and experiences.

Desired tone: If you have a specific style, tone, or voice in mind, mention that in the prompt. This guides the AI's output to match your expectations. For example, "suggest a short opening of no more than three lines for a personal statement in a professional tone based on the CV and opportunity description I have uploaded."

Constraints: If there are any constraints, restrictions, or guidelines to follow, communicate them. For instance, if you want a response within a certain word limit or you want the AI to avoid certain topics, make sure this is part of your prompt.

For example, "I have uploaded information about an opportunity. Suggest some interview questions based on this role, but avoid highly technical questions and focus mostly on motivation questions."

Background information: If the topic is niche or specialised, provide some basic background information. This helps the AI understand the context and generate more accurate content.

This could be relevant if you are applying for an opportunity in a highly specialised area. There may be some text on the company's "about us" page that could provide this background information.

Prompt length: While being clear and specific is important, overly long prompts might confuse the AI. Keep the prompt as concise as you can while conveying all the necessary information – you might decide to refine the output by asking questions in stages if your prompt is getting too long – see "progressive prompts" in the Review and Revise tab for an example of this.

Review and revise

The content generated by the AI will likely need refining and editing before you can use it. Below are some tips to help with this:

Progressive prompts: If the initial response isn't what you're looking for, consider using progressive prompts, which means building upon the previous responses to guide the AI in the right direction. For example, once some content has been produced, you could ask the programme to "rewrite this in a formal tone," or "rewrite this and take out the academic grades."

Iterative refinement: Don't hesitate to refine your prompt if the initial output isn't satisfactory. Tweaking the prompt can often lead to better results. This is different to progressive prompts because you will change the initial prompt rather than asking follow-up questions.

Try both options and see what works best for you!

As well as refining the output directly in the AI programme, you should also edit any content produced to ensure it is accurate to you and your experience, and that it isn't generic and non-specific. Think of content produced by generative AI programmes as a first draft – you will still need to finalise the content and style to suit your purposes.

Here are some ideas from some recent "Career Chat" discussions via LinkedIn Learning and Coursera.

1. You can use a GenAI (like ChatGPT) as a 'thought partner' to help discuss ideas. Prompts you could try include:

• • You are my thought partner. Help me figure out my goals
  • You are a life coach. Help me figure out what areas of my life I should focus on bettering and how to start changing them
  • You are a career coach. Ask me questions to help me create a personalised and specific plan for improving my career and professional development

You can always add more specific information and parameters to your prompt about what your current situation is, your past experience or your long-term goals. You can always adjust your prompt once you see the tool's output.

2. You could use GenAI to research a company. Make sure that your prompt is appropriately detailed:

• • I am applying for a position at [company name] as a [role title]. Please find information about [company name]'s values and mission as well as recent news about their work, avoiding news stories older than 2018

3. You can use GenAI to assist in screening a CV to a job role. Here is an example prompt to do so:

• • You are an Applicant Tracking System programmed by a professional recruiter to identify candidates for an open job role with the following listing [paste your job description]. Identify the relevant CV keywords. Then, assess the strengths and weaknesses of a candidate with the following CV [paste your CV]

Be sure to use the CV keywords as a checklist of words to include in your CV. Your strengths will reflect the CV items you should lead with, helping you to determine how to best organise your experience bullets or what to include in your CV summary. Your weaknesses will tell you the new experiences you should add to your CV or potential growth areas if you don't currently have those skills.

4. You could use the chatbot to edit your individual CV sections. One way to do this is with the simple prompt:

• • Edit these CV bullet points for clarity, according to  UK CV best practices, and to better cater to the skills in the job description [paste your work experience bullets]

Or, for a CV summary try:

• • Edit this CV summary according to UK CV best practices and to better cater to the skills in the job description [paste your CV summary]

5. You can use GenAI to write an example cover letter for a role. It is a good idea to ask this of the system once it has already looked into the company and seen the job advert (see earlier examples). Your prompt here could be:

• • Generate a cover letter for this job

PLEASE NOTE: This is likely to be generic and will have only limited relevance to you as an individual. Remember to edit any content produced to ensure it is accurate to you and your experience and that it isn't generic and non-specific.

---

In short, generative AI is one tool to help you during the application process and works best when used as a starting rather than a finishing point for your documents. Always edit and tailor your resulting CV and/or cover letter until you are confident it demonstrates your enthusiasm and skills relevant to the role.

• Imperial College statement on generative AI [webpage]
• Charlotte Whitehead – Can ChatGPT help your career? [LinkedIn blog post]
• Digital Waffle – The ultimate guide to using ChatGPT for CVs, cover letters and job interviews [Article]
• Forbes – Prompt engineering boosted via are-you-sure AI self-reflective self-improvement techniques that greatly improve generative AI answers [Article]
• Raj Sidhu – What ChatGPT changes recruitment FOREVER [YouTube video]
• Silei Cheng et al – Prompting GPT-3 to be reliable [links to PDF download]
• The Forage – We asked ChatGPT to write cover letters. Here's what it got right and wrong. [Article]
• Unlocking Success: the best AI tools for your career [LinkedIn article]