Immigration is one of the categories where AI assistants get used most heavily, and the reason is straightforward. The questions are complicated, the official sources are written in dense bureaucratic language, and the person asking is usually anxious and does not know the right vocabulary yet. An assistant that explains a pathway in plain English is genuinely useful to them.
Those conversations often end with a recommendation. Someone asks which consultancy handles a particular kind of case in their city, and the assistant names two or three. If your firm is not among them, you never entered the process. There is no ranking report telling you this happened, which is why most firms have no idea whether it is happening at all.
How to run the check
The method is simple and you do not need any tools. Open ChatGPT, Gemini and Perplexity, and ask each of them the questions your clients would ask. Record what comes back. That is the whole exercise.
The important part is asking the way a client would, not the way you would. A consultant might ask about subclass 190 nomination requirements. A client asks whether they can move to Australia as a nurse from Manila. Use their words.
A reasonable starting set looks something like this. Who are the best immigration consultants in your city for the pathway you specialise in. What should someone do after a visa refusal in your jurisdiction. Which firms handle a particular category of application. And then, most directly, ask about your own firm by name and see what it says.
Ask each question in a fresh conversation. Assistants carry context within a conversation, so asking follow-up questions in the same thread will skew what you get. Turn off any personalisation or memory features first, otherwise you are seeing results shaped by your own history rather than a stranger's.
Reading the result honestly
There are four outcomes and they need different responses.
The assistant names your firm accurately. This is the good case, and it usually means you have consistent information across your site, directories and third party mentions. The work here is maintenance rather than repair.
The assistant names competitors but not you. This is the most common outcome and the most useful one, because you can look directly at what those firms have that you do not. Usually it is a combination of directory listings, clear service descriptions and content that answers the question being asked.
The assistant describes your firm inaccurately. This happens more than people expect, often because of an outdated listing or a stale profile somewhere. It is fixable and worth fixing quickly, because a wrong description repeated confidently is worse than no description.
The assistant declines to recommend anyone. Some models are cautious about recommending regulated professional services, and this varies by how the question is phrased. Try a few variations before concluding anything, because a single refusal is not a pattern.
Run the same questions once a month and keep the answers in a document. A single check tells you where you stand today. A run of six tells you whether anything you did made a difference, which is the only honest way to measure this.
Why some firms get named and others do not
Assistants are not ranking firms by quality. They are assembling an answer from sources they can read and trust, which means visibility depends on being present, clear and consistent across those sources.
Being present means appearing somewhere the model draws on. Directories, professional association listings, roundups, local business listings and news mentions all count. Your own website matters less here than most firms assume, because a site can only tell a model what you say about yourself.
Being clear means having sentences a model can lift without interpreting them. A page stating that a firm handles Australian skilled migration for applicants in the Philippines gives an assistant something usable. A page describing comprehensive global mobility solutions does not.
Being consistent means describing yourself the same way everywhere. If your website, your association listing and your Google profile give three different descriptions of what you do, an assistant has to decide which to believe, and hedging or omitting you is the safer choice from its point of view.
What to do about it
Start with the inconsistencies, because they are quick and they compound. Write one paragraph describing your firm, covering what you handle, who you serve and where. Then make every profile you control match it. This is unglamorous and it is usually the highest return work available.
Then look at the questions that produced competitor names and check whether you have a page answering each one directly. Not a service page mentioning the topic. A page that answers the question, in plain language, near the top.
Then get listed where you are missing. Professional directories carry disproportionate weight in this context, partly because assistants treat them as structured and reliable.
All of this overlaps heavily with ordinary search optimisation, which is a good sign rather than a coincidence. Work that makes you easier for a machine to quote generally makes you easier for a person to understand. We covered the underlying mechanics in our guide to GEO and AEO, and the writing side of the problem in why immigration firm websites sound the same.
A caution about expectations
Nobody can guarantee an AI citation. There is no submission process, no dashboard and no lever to pull. What you can do is remove every reason a model would skip you, then measure whether anything changes. Any agency offering guaranteed placement in AI answers is selling something that does not currently exist.