*/
Dr Thomas Lancaster on the risks of using GenAI in your pupillage application
Pupillage places are scarce and the competition is fierce. It is important to present your skills, expertise and abilities in a positive light, something that generative artificial intelligence (GenAI) systems like ChatGPT can help with. But if you’re tempted to simply use ChatGPT to write your pupillage application for you, do think again.
My background is that of a computer scientist, and I’ve seen AI systems develop drastically increased capabilities during my lifetime. The single biggest AI advance is one you’ve likely encountered during your time as a student. It is the easy interface to GenAI through systems like ChatGPT that has begun to cause concern in academia and in the workplace.
At face value, these GenAI systems are useful and powerful. They can be used to automate roles humans find tedious, such as collecting data systematically from websites, or searching for information hidden in obscure research papers. They can output text that is well written and grammatically correct.
Despite the benefits, there are dangers to be aware of when using these systems. Generated text can contain hallucinations or be entirely fictional. Information may be grammatically correct but lack factual accuracy. In legal areas, a law that is correct in one jurisdiction may not exist in another, but the GenAI system may not differentiate between the two.
These dangers are not that surprising if you consider how GenAI systems for text creation work. They are mathematical models, not intelligent beings. The models output text that complies with known rules for structure. They match words to patterns based on previous input. Without getting too technical, there is also randomness in the responses, so two people putting in identical prompts will get two different responses. This randomness is dangerous in legal work, for which accuracy and attribution matters.
There is ongoing debate regarding how reliably GenAI written content can be detected. As an educator, signs of unoriginality can stand out to me when I read completed student assessments. For example, some wording choices appear regularly in GenAI text that would rarely be spotted in traditional student writing. There can be factual errors. An argument that seems convincingly written may not make sense to someone who knows the field. The written work could describe a situation or physical location that a student like you is unlikely to know about.
Many companies offer AI detection services, playing one mathematical model against another. Often, the detection models do correctly identify AI generated text. But often might only mean 90% of the time. From an AI research perspective, 90% accuracy would be considered a good result. From an educational perspective of determining if a student has unfairly used AI, relying solely on the detection service would mean falsely accusing one student in every ten of breaching academic integrity. That is just not fair to anyone. Based on that, I always urge caution when trying to decide if GenAI misuse has taken place or not.
Many people, including legal professionals, also feel that GenAI use is easy to spot. The reaction of a pupillage application reviewer to spotting suspected GenAI use is likely to be different to the typical educational one. For pupillage applications, where there are more suitable applicants than available places, reviewers may possibly err on the side of caution. To some individuals, caution might mean rejecting a suspect pupillage application. Therefore, it is best that your application is human written, with all its expected quirks.
You might read this and think I am discouraging the use of GenAI. That’s not the case. I’m a big supporter of the benefits of using GenAI to work efficiently, while being mindful of its potential limitations. But I also believe that it’s essential to follow the rules governing GenAI use in different situations, including in your pupillage application.
For those AETOs that do allow some use of GenAI, how could you use it ethically on any preparatory material relating to your pupillage application? My expertise relates to academic integrity and below are steers for sensible use of AI in preparation for your human written application.
First, you could ask the system to suggest ideas for points you should make in your application. Feed in details of your background, interests and career aspirations. Ask it to help you to develop examples to include in the application. You still need to make sure these are accurate and written in your own words. Don’t directly copy them from the GenAI system.
Next, you could use GenAI as a proofreading partner. Ask the system to check your draft application and point out typos, wording choices that do not make sense, and the like. Once again, it’s important that you don’t ask the system to make the changes for you. Just ask it where the problems are, then you make the adjustments so that everything remains in your unique writing style.
Finally, you could use GenAI to critically evaluate your application. Ask the system to provide points that are missing from the statement, areas where you could better present your argument, or suggestions for how to reduce the word count if the statement has become too long. Make the changes yourself to ensure they accurately reflect your own abilities. Sometimes the system will suggest standard examples that could appear in many statements, but your examples need to relate to your own experiences and stand out to the people who are reviewing your application.
You can, and should, treat the review process as iterative. Perhaps ask the GenAI system to think about this from the point of view of different personas involved with the application review process. You could also use multiple different GenAI systems for this review. You don’t have to just stick with ChatGPT. Consider also asking for reviews from Gemini, Grok and the like. You may get different responses and ideas that you wouldn’t otherwise have had access to.
Once again, I emphasise that you must study what each AETO will or will not allow around AI usage. Even those that do allow it, may still be specific in terms of what they will and won’t permit within that usage. So the above suggestions made here need to be read within the context of each individual AETO policy. Ultimately it is critical you adhere to whatever is stipulated there.
GenAI systems continue to evolve rapidly, as do views about how they can be best used in the legal profession and beyond. Right now, though, your application should be personally written by you. It should provide an authentic reflection of your abilities and experiences and demonstrate that you can adhere to requirements.
As an AI enthusiast, I feel that any job applicant should show they are committed to continual learning about emerging technologies and are willing to adapt working practices to include GenAI when this is expected and in an ethical way. This matches the way in which the legal profession is increasingly embracing technological solutions that uphold professional standards, and you are helping to position yourself as the ideal barrister who is prepared for the future.
Dr Thomas Lancaster is a Principal Teaching Fellow in Computing at Imperial College London. He is a popular keynote speaker and media contributor, known for research into contract cheating and academic integrity.
Pupillage places are scarce and the competition is fierce. It is important to present your skills, expertise and abilities in a positive light, something that generative artificial intelligence (GenAI) systems like ChatGPT can help with. But if you’re tempted to simply use ChatGPT to write your pupillage application for you, do think again.
My background is that of a computer scientist, and I’ve seen AI systems develop drastically increased capabilities during my lifetime. The single biggest AI advance is one you’ve likely encountered during your time as a student. It is the easy interface to GenAI through systems like ChatGPT that has begun to cause concern in academia and in the workplace.
At face value, these GenAI systems are useful and powerful. They can be used to automate roles humans find tedious, such as collecting data systematically from websites, or searching for information hidden in obscure research papers. They can output text that is well written and grammatically correct.
Despite the benefits, there are dangers to be aware of when using these systems. Generated text can contain hallucinations or be entirely fictional. Information may be grammatically correct but lack factual accuracy. In legal areas, a law that is correct in one jurisdiction may not exist in another, but the GenAI system may not differentiate between the two.
These dangers are not that surprising if you consider how GenAI systems for text creation work. They are mathematical models, not intelligent beings. The models output text that complies with known rules for structure. They match words to patterns based on previous input. Without getting too technical, there is also randomness in the responses, so two people putting in identical prompts will get two different responses. This randomness is dangerous in legal work, for which accuracy and attribution matters.
There is ongoing debate regarding how reliably GenAI written content can be detected. As an educator, signs of unoriginality can stand out to me when I read completed student assessments. For example, some wording choices appear regularly in GenAI text that would rarely be spotted in traditional student writing. There can be factual errors. An argument that seems convincingly written may not make sense to someone who knows the field. The written work could describe a situation or physical location that a student like you is unlikely to know about.
Many companies offer AI detection services, playing one mathematical model against another. Often, the detection models do correctly identify AI generated text. But often might only mean 90% of the time. From an AI research perspective, 90% accuracy would be considered a good result. From an educational perspective of determining if a student has unfairly used AI, relying solely on the detection service would mean falsely accusing one student in every ten of breaching academic integrity. That is just not fair to anyone. Based on that, I always urge caution when trying to decide if GenAI misuse has taken place or not.
Many people, including legal professionals, also feel that GenAI use is easy to spot. The reaction of a pupillage application reviewer to spotting suspected GenAI use is likely to be different to the typical educational one. For pupillage applications, where there are more suitable applicants than available places, reviewers may possibly err on the side of caution. To some individuals, caution might mean rejecting a suspect pupillage application. Therefore, it is best that your application is human written, with all its expected quirks.
You might read this and think I am discouraging the use of GenAI. That’s not the case. I’m a big supporter of the benefits of using GenAI to work efficiently, while being mindful of its potential limitations. But I also believe that it’s essential to follow the rules governing GenAI use in different situations, including in your pupillage application.
For those AETOs that do allow some use of GenAI, how could you use it ethically on any preparatory material relating to your pupillage application? My expertise relates to academic integrity and below are steers for sensible use of AI in preparation for your human written application.
First, you could ask the system to suggest ideas for points you should make in your application. Feed in details of your background, interests and career aspirations. Ask it to help you to develop examples to include in the application. You still need to make sure these are accurate and written in your own words. Don’t directly copy them from the GenAI system.
Next, you could use GenAI as a proofreading partner. Ask the system to check your draft application and point out typos, wording choices that do not make sense, and the like. Once again, it’s important that you don’t ask the system to make the changes for you. Just ask it where the problems are, then you make the adjustments so that everything remains in your unique writing style.
Finally, you could use GenAI to critically evaluate your application. Ask the system to provide points that are missing from the statement, areas where you could better present your argument, or suggestions for how to reduce the word count if the statement has become too long. Make the changes yourself to ensure they accurately reflect your own abilities. Sometimes the system will suggest standard examples that could appear in many statements, but your examples need to relate to your own experiences and stand out to the people who are reviewing your application.
You can, and should, treat the review process as iterative. Perhaps ask the GenAI system to think about this from the point of view of different personas involved with the application review process. You could also use multiple different GenAI systems for this review. You don’t have to just stick with ChatGPT. Consider also asking for reviews from Gemini, Grok and the like. You may get different responses and ideas that you wouldn’t otherwise have had access to.
Once again, I emphasise that you must study what each AETO will or will not allow around AI usage. Even those that do allow it, may still be specific in terms of what they will and won’t permit within that usage. So the above suggestions made here need to be read within the context of each individual AETO policy. Ultimately it is critical you adhere to whatever is stipulated there.
GenAI systems continue to evolve rapidly, as do views about how they can be best used in the legal profession and beyond. Right now, though, your application should be personally written by you. It should provide an authentic reflection of your abilities and experiences and demonstrate that you can adhere to requirements.
As an AI enthusiast, I feel that any job applicant should show they are committed to continual learning about emerging technologies and are willing to adapt working practices to include GenAI when this is expected and in an ethical way. This matches the way in which the legal profession is increasingly embracing technological solutions that uphold professional standards, and you are helping to position yourself as the ideal barrister who is prepared for the future.
Dr Thomas Lancaster on the risks of using GenAI in your pupillage application
The Chair of the Bar reports back
Clement Cowley, Partner at The Penny Group, explains how global events can influence the performance of pensions and ISAs, and why taking a long-term approach to investing is often the most effective strategy
Mário Barroso, Head of R&D and Method Development at AlphaBiolabs, discusses what family law professionals need to know about hallucinogens, from the main hallucinogenic drugs encountered in family proceedings, to their effects and the testing options available
Harmony Christian Ministries is the latest charity to benefit from a £500 donation from AlphaBiolabs via the company’s Giving Back initiative
By David Green
Mário Barroso, Head of R&D and Method Development at AlphaBiolabs, examines the forensic science underpinning hair drug testing, its evidential scope and limitations, and why it remains the gold standard for evidencing patterns of drug use in family proceedings
Developing a tech start-up demands a different definition of excellence to the Bar, says Saara Idelbi. The co-creator of Advocatr on entrepreneurship, AI advocacy training, and matching the thrill of a killer cross-examination
At the start of her term, Chair of the Bar 2026 Kirsty Brimelow KC set out a series of priorities for the year ahead. One of those priorities was children in the justice system
As the world becomes more digital, Barbara Vinagre Mota examines international research into the mental health consequences of judges’ exposure to graphic and immersive evidence, highlighting the importance of support systems
Reviewed by Daniel Barnett
James McNeill Whistler at Tate Britain until 27 September Rothko in Florence at Palazzo Strozzi until 23 August Reviewed by Stephen Cragg KC
