Can I Use AI for My New Zealand Visa Application?

There’s an old tale about two men who go to a rabbi to settle a dispute. The first man explains his side of the story. The rabbi listens carefully, nods, and says “You’re right.” The second man is outraged and explains his side of the story. Again, the rabbi listens carefully, nods, and says “You’re right.” A third person, who has been listening, protests: “Rabbi, they can’t both be right!” The rabbi thinks for a moment, turns to him and says, “You’re right too.”

I’m often reminded of this story when using generative AI. AI tools such as ChatGPT, Claude and Gemini can be extraordinarily useful. Increasingly, people are using them to research New Zealand immigration requirements, assess potential visa eligibility, prepare documents and help with their own New Zealand visa applications.

There are real benefits to doing so but equally there are also important limitations; particularly when AI moves from helping someone organise and communicate information to effectively providing legal advice.

Can I use ChatGPT or other AI tools for a New Zealand visa application?

AI tools can certainly be useful when preparing a visa application but that doesn’t mean they should be relied upon to determine your eligibility or how your application should be approached. Much depends on what you are using AI for.

Generative AI can explain complicated concepts in plain English, organise information, and turn out a polished looking document in seconds. But it also has an important characteristic: it is very good at responding to the way a question is framed.

Tell an AI tool why you think you qualify for a visa and ask whether your reasoning is correct, and you may receive a compelling explanation of why that is. Change the emphasis, introduce a new fact, focus on the weaknesses or ask it to explain why you don’t qualify, and you may receive an equally compelling answer going the other way. Like the rabbi, it can sometimes seem as though everyone is right. The difficulty is knowing which answer, if any, you should actually rely on.

What are the risks of relying on AI for immigration advice?

The usual warning about AI is that it can “hallucinate”, in that it can confidently provide information that is inaccurate or even entirely invented. But I think the more subtle risk for someone preparing their own visa application is judgment. Immigration cases frequently aren’t resolved simply by locating an immigration instruction and applying it mechanically.

The difficult questions are often things such as, what information is actually relevant? Is an explanation likely to resolve a concern or inadvertently create another one? Is a weakness sufficiently important that it should be addressed proactively? How might a particular statement be interpreted by an immigration officer

Experienced lawyers develop a feel for which issues are routine and which should make them stop and dig further. While AI may assist with analysis, it can’t replace professional judgment or assume professional responsibility for getting the advice right.

There is also a difference between having access to the published immigration instructions and understanding how they operate in practice. The instructions can involve significant nuance and interpretation. Immigration New Zealand also provides guidance and clarification to the immigration industry, not all of which is necessarily available publicly. Lawyers and advisers may also seek clarification directly from Immigration New Zealand where an instruction or its application is unclear.

Then there is the less tangible knowledge that comes from dealing with the same types of applications repeatedly: knowing which issues tend to matter in practice, which usually don't, when something unusual warrants further investigation, and when it may be appropriate to seek clarification from Immigration New Zealand. An AI tool working primarily from publicly available information doesn't necessarily have the benefit of that wider context.

The other problem with using AI to assess your own visa eligibility is that you are supplying the facts. If you inadvertently leave out something important, the AI generally cannot know that it is missing. If your understanding of a fact is wrong, the answer may nevertheless proceed on the basis that it is correct. And if your prompt contains an assumption about how New Zealand immigration law works, the resulting answer can inherit that assumption. This can create a feedback loop. For example, you explain why you think you qualify for a particular visa, AI analyses the case you have presented, and its answer reinforces your original view.

It can work the other way too. An applicant who has become overly concerned about a particular issue can frame a series of questions around that concern and receive answers that reinforce the impression the issue is more significant than it really is. In other words, AI can both underestimate and overestimate risk. A lawyer reviewing the same circumstances may ask completely different questions. Sometimes the most valuable question is the one the client didn’t know needed to be asked.

What can AI be useful for when preparing a visa application?

At this point I should acknowledge the obvious: of course an immigration lawyer would write an article warning people about replacing immigration lawyers with AI. I appreciate the irony but I don’t think the sensible response is to avoid AI altogether. Used appropriately, I think it can be genuinely useful for people dealing with lawyers and the immigration system.

For example, AI can help with things like turning a messy immigration or relationship history into a clear timeline, tidying up grammar (especially where English is not your first language) and helping with clearer communications, explaining an unfamiliar term you’ve never heard before, or knocking out a first draft of a supporting document. In these ways, AI can make interactions between a client and their lawyer much more efficient – and often produce better quality work as a result.

But there are clear limitations. For example, I would be more cautious about using AI to decide which legal questions you should ask your lawyer. One of the purposes of obtaining legal advice is for the lawyer to work out which questions actually need to be asked. A client – or an AI tool working from the client’s understanding of the situation – may focus on something that turns out to be relatively unimportant, overlook something that matters considerably more or frame a question around an assumption that’s incorrect. The value of good legal advice isn’t simply receiving answers to the questions you already have but is also identifying the issues and questions you didn’t know were relevant.

Can I use AI to write a visa support letter?

Potentially yes but considerable care is required. AI may help to structure a support letter or prepare an initial draft but the final letter should remain their evidence. It needs to be truthful and accurate and should reflect their actual experiences, knowledge and way of expressing themselves. Facts should never be invented or embellished simply because an AI tool thinks they would make the letter more persuasive.

Ten beautifully written support letters that all sound suspiciously as though they were written by the same robot are considerably less useful than ten imperfect but authentic letters from real people. AI should assist the person telling their story, rather than manufacture the story for them.

The same principle applies when using AI to communicate with your lawyer. It can be useful for improving the clarity of an email, particularly for someone who struggles with written English. But make sure you understand and agree with what it has written. A beautifully drafted explanation is not helpful if it doesn’t accurately reflect what you actually mean.

Be careful about uploading immigration documents to AI

Visa applications can contain extremely sensitive information: passports, medical histories, criminal records, financial information, relationship evidence and information about third parties. Before uploading this material to any AI service, users should understand what happens to their data and the privacy settings and terms applying to the particular tool.

Similar considerations apply to lawyers. The New Zealand Law Society, which regulates lawyers in New Zealand, has identified privacy, confidentiality, data protection, accuracy and human oversight as important issues when generative AI is used in legal work.

AI can assist with legal work but judgment still matters

In my view, both “AI can do my visa application” and “AI has no place in immigration work” are overly simplistic. AI is becoming an extremely capable assistant. I use AI tools myself, subject to appropriate safeguards. The legal profession is increasingly doing the same.

And, yes, AI will undoubtedly reduce the need for lawyers to perform some work that has traditionally been done manually. In my view, that’s not necessarily a bad thing. As mentioned above, the important distinction is between assistance and judgment.

As mentioned, AI may help an applicant become better organised, better informed and better able to communicate. But it doesn't necessarily have access to all of the information or practical context that informs professional advice. Furthermore, even in an apparently straightforward New Zealand visa application it can create unnecessary problems; for example, by oversimplifying an issue, giving undue significance to something relatively unimportant or underestimating something that actually matters.

Also, of course, an applicant may not always know whether their application really is straightforward in the first place. For a complicated or high-stakes application, those risks become even more significant.

The question shouldn’t be whether AI can produce an impressive answer. It should be whether you have enough knowledge to recognise when the impressive answer is wrong.

Because sometimes AI will tell you you’re right, and sometimes, when Immigration New Zealand takes the opposite view, it may effectively tell them they’re right too.