Recently, many people have been discussing this question: will AI replace lawyers?
As someone who has actually used AI in legal work, my view is that AI is very useful, but it is still a long way from “replacing lawyers”.
This is not to say that AI has no value. It can help you organise your thoughts, refine your writing, summarise materials and conduct preliminary research. However, if you are dealing with a specific legal issue involving a visa, contract, litigation, family matter or company dispute, you should never treat AI’s answer directly as legal advice.
The reason is simple.
1. AI is often inaccurate when researching law
AI can easily produce an answer that “looks professional”.
It may cite legislation, discuss cases and analyse the logic, all in a particularly confident tone. The problem is that sounding plausible does not mean it is correct.
Law is not ordinary general knowledge. A single word, provision, policy or case’s scope of application may affect the ultimate conclusion.
This is especially true in areas such as migration law, tax law, company law and litigation procedure, where rules change quickly and details matter. AI may combine old and new rules, rules from different countries, and similar but inapplicable cases.
Its answer may look comprehensive, but a lawyer can immediately see that:
the logic is wrong, the source is wrong and the circumstances in which it is being applied are also wrong.
2. AI does not automatically understand law; a person must tell it where to learn
Many people assume that if they ask AI a question, it will find the most accurate legal authority by itself.
The reality is that AI may not know which legal source it should consult.
Proper legal analysis is not a matter of running a casual Google search or reading a few online articles.
A lawyer needs to know:
should the source be legislation or a regulation?
Should it be a policy or a direction?
Should it be an explanatory statement or a tribunal/court decision?
Should it be the current version or the historical version in force at the relevant time?
Which department’s official guidance should be consulted?
Is a particular case binding, or merely persuasive?
Making these distinctions is itself part of legal training.
AI can assist with reading materials, but only if you know which materials to give it.
3. AI has no genuine capacity for self-correction, so a lawyer must still verify the result
One of AI’s greatest problems is that it may not know when it is wrong.
If you ask, it will answer.
If you ask a follow-up question, it will explain.
If you challenge it, it may even change its wording and continue answering.
However, it has no genuine professional judgement or awareness of legal responsibility. It will not repeatedly ask itself, as a lawyer would:
Is this provision the latest version?
Have any facts been omitted?
Are there any exceptions to this conclusion?
Is there a conflict with a higher-ranking law?
Is there a new tribunal or court case that affects the assessment?
What are the risks if the client acts this way?
What is the worst possible outcome?
Legal work is not about writing an article that “looks reasonable”; it requires constant error checking, verification, rejection and reconstruction.
AI can assist, but a lawyer must ultimately determine:
whether the answer can be used at all, where the risks lie and whether the client can rely on it.
4. AI cannot empathise with clients
Legal services are not simply about providing answers.
Often, a client approaches a lawyer not only because they want to know “what the law says”, but because they are anxious, uncertain and afraid of making the wrong decision.
A person whose visa has been refused worries about whether they can remain in the future.
A person facing a contract problem worries about whether they can recover their money.
A person involved in a family dispute worries about children, property, dignity and a sense of security.
A person whose business is in trouble worries about whether years of effort will be lost.
A lawyer’s role is not only to analyse the law, but also to understand the person’s circumstances.
Sometimes what a client truly needs is not a cold answer, but someone who can explain:
what options are now available;
the risks of each option;
which course is realistically viable;
which course merely appears hopeful;
when to contest a matter and when to settle;
when to continue and when to cut their losses.
AI cannot replace this merely by producing a few passages of text.
5. Most importantly, AI is not accountable to the client
This is the most important point.
Lawyers are licensed, bear professional responsibilities, are regulated, carry insurance and have professional ethical obligations.
If a lawyer gives advice recklessly, they may face a complaint, a penalty or liability, and their right to practise may even be affected.
But what about AI?
If AI gives a wrong answer, it will not bear liability.
If AI misleads you, it will not compensate you.
If AI causes you to miss a deadline, it will not help you remedy the problem.
If AI gives you the wrong strategy, it will not face the tribunal, court or government department on your behalf.
Ultimately, you are the person who bears the consequences.
The most dangerous aspect of a legal issue is therefore not “having no answer”, but receiving a wrong answer that looks convincing and then believing it.
AI can be used, but it must not be trusted blindly
I am not opposed to AI.
On the contrary, I believe AI is a very useful tool. It can make lawyers more efficient, help organise documents, assist with research and make complex content clearer.
However, it should be a tool, not a lawyer.
Genuine legal services require legal knowledge, experienced judgement, factual analysis, risk management, client communication and, most importantly, accountability.
My view of AI at this stage is therefore:
AI can assist your understanding, but it should not directly replace legal advice.
AI can help you organise your questions, but it should not decide your legal strategy.
You can consider what AI says, but a professional lawyer must ultimately help you decide whether it can be trusted.
Law is not a chat game.
One wrong step may affect a visa, property, a business, a family or even the direction of a person’s life.
When you encounter a genuine legal issue, therefore, the same point applies:
AI can be a reference; a lawyer is accountable.