Who actually reads your Australian visa application?

When you press submit on a visa application, it is natural to picture a person on the other side. Someone in an office, reading your bank statements, weighing your travel history, deciding your fate.

For some visas, no such person exists. A computer checks your details against a fixed set of rules and grants the visa in minutes. For others, a case officer really does read every page, over months. But there is one thing no computer is allowed to do to any Australian visa: refuse it.

We know this because the Department has been made to release the documents: the instruments that let computers decide visas, the grant data behind the Electronic Travel Authority, the student-visa operating procedures, the auditor's report on family migration. Read together, they show a system built around one rule and a lot of triage. This post is about which lane your visa lands in, and what puts you there.

Who actually reads your Australian visa application?

For a visa a computer can grant, often no one reads it. The computer checks your details against fixed rules and issues the grant. For everything else, a person reads it, sometimes closely. But whichever path your visa takes, one rule holds across all of them: a refusal is always a human decision.

The authority for a computer to decide a visa sits in one section of the Migration Act 1958: section 495A, added in 2001. It lets the Minister "arrange for" computer programs to make decisions, and says the Minister "is taken to have" made whatever the program decides. Those programs run under written arrangements that were never published in the ordinary way. They became public only in September 2022, when the Department released 93 pages covering the section 495A arrangements in effect from 2015 to 2018, under FOI reference FA 22/02/00278.

Every one of those arrangements carries the same sentence: it "does not authorise the use of computer programs to make decisions to refuse to grant a visa." A computer can say yes. Only a person can say no. Home Affairs confirmed as much publicly in October 2025, stating that refusing a visa "is never undertaken through computer-assisted processing."

Automation is not there to judge you. It is there to clear the easy cases fast. The moment your file stops being easy, it moves to a person, and a person is the only one who can refuse it. So the whole game is to stay easy.

Which Australian visas can a computer grant on its own?

A specific, named list of visas can be granted by a computer, and it is shorter than most people assume. The released arrangements name each one. If your visa is not on the list, a person grants it.

Skilled points-tested visas are the clearest example of where the computer deliberately stops short. For General Skilled Migration, the application runs through a rules engine, but the engine only produces "regulatory assessment recommendations." The arrangement spells it out: "All decisions to grant a GSM visa are made by a Visa Processing Officer" (IMMI 08/026). The monthly SkillSelect invitation rounds are a "fully automated process" (IMMI 12/109), but an invitation is not a grant. It is only permission to apply.

How many ETAs are granted by a computer with no human involved?

In the 2024-25 financial year, 93.5% of Electronic Travel Authorities were granted by computer with no human involved. That is 2,066,558 of 2,209,379 applications, according to a data release the Department published under FOI reference DA 25/07/00308. Another 5.2% (115,419) were granted manually by an officer. A small remainder, roughly 1.2%, were not granted in that period.

The ETA (Subclass 601) is the visa most travellers from eligible passport countries use for a short trip. It is applied for through a phone app that scans your passport chip. For most nationalities, the system checks you against its rules and issues the grant in minutes. For a fuller walkthrough, see our guide to the Australian ETA.

Here is how automatic the process is for the biggest passport groups, all figures for the 2024-25 financial year to 30 June 2025:

The United Kingdom number is striking: of 107,039 UK ETAs, just 14 all year were granted manually. Everyone else was auto-granted. (The UK figure is also smaller than you might expect, because most British travellers use the free eVisitor visa (Subclass 651) instead of the ETA. If that is you, our eVisitor guide covers the difference.)

The ETA is not one of the arrangements in the 2022 release. Its automation predates the section 495A framework, since the ETA has existed since 1996, so it does not appear as a named instrument in the bundle. But the grant data shows the same pattern: computers grant, people refuse.

A note on the figures: the released table gives clean counts for computer-assisted grants and for manual grants. The "not granted" remainder is what is left over after those two, so we treat it as approximate rather than an exact count of refusals.

Why is Malaysia the exception?

Malaysia is the one large country the automated ETA system does not wave through. Only about 27% of Malaysian ETAs were computer-granted in 2024-25. About 59% (108,476) needed a human officer to grant them, and the rest were not granted in that period. Put another way, Malaysians accounted for roughly 94% of every manually-granted ETA in the world that year.

The data tells us that this happens, not why, and we will not guess at reasons the document does not give. But the shape is clear: even in a stream that is 93.5% automated, a whole nationality can sit almost entirely in the manual lane, set there before an officer looks at anything personal about you. So do not assume "everyone gets auto-granted" applies to you. If your passport puts you in the manual lane, your file needs to hold up to the person who will read it.

Are partner visas decided by a computer?

No. No computer program is authorised to grant a partner visa, and none can refuse one either. Every partner visa decision, at both the temporary and permanent stage, is made by a case officer reading your evidence. But your application is still sorted by algorithm before a human ever opens it.

The auditor documented exactly how, in ANAO Report No. 16 2022-23, "Management of Migration to Australia: Family Migration Program" (March 2023, the full report). Three things happen before an officer assesses your file:

The line that matters is in a footnote: "Risk ratings cannot be used as the basis for making a decision on an application. Visa decisions must be based on evidence that directly relates to the legal criteria for the specific visa type." Triage decides how much scrutiny your file gets and how fast it moves. It cannot decide the outcome. That is still a person, reading your evidence against the law.

The Department's own step-by-step instructions for granting a partner visa show the same thing from the officer's desk. The 820/801 Standard Operating Procedures, used from 2014 to 2020 and released in 2024, tell the case officer to "Complete MAL, Movement and Risk checks" for each application, and where a risk alert appears, to make sure "all alerts are actioned and relevant notes are made" (FOI FA 24/04/00324). Even the risk checks are run and cleared by a person.

The only computer-issued grant a partner applicant actually receives is the bridging visa, back at the start. The partner visa itself takes real time precisely because a person reads all of it. The median for a permanent Partner visa was 271 days in the 2024-25 financial year. For how that queue is ordered, and the internal guideline that decides which partner files get moved up it, see our guide to partner visa priority processing, and for wider processing-time patterns our guide to Australian visa processing times.

Not every family visa is even triaged this way. The same auditor noted that the Parent, Child and Other Family programs "do not use prioritisation tools to triage their caseloads" at all.

How were student visas triaged?

Student visa scrutiny starts with a risk score, and that score is set by a computer program of its own. Before an officer looks at anything, a system called CVOR combines the risk rating of your country with the risk rating of your education provider to decide how much evidence you even have to supply.

This is itself a section 495A arrangement (IMMI 16/030). Each country and each registered education provider is rated one, two or three, "based on a statistical analysis of student visa refusal, cancellation and non-compliance rates." A low-risk combination gets "streamlined" evidence requirements. A higher-risk one gets "regular" requirements, which means you must produce documentary proof of your English and your finances up front, before a human is involved.

Once your application is with an officer, the same trigger logic continues. This comes from the Department's Student Visa Standard Operating Procedures and its Quick Reference Guide, both dated 6 June 2018 and released under FOI reference FA 22/11/00808 (the Quick Reference Guide is part 2).

Date this one clearly. In 2018, the genuineness test for a student was the Genuine Temporary Entrant (GTE) requirement. GTE was replaced by the Genuine Student (GS) requirement in 2024, so the specific test in these documents is historical. What carries over is the triage architecture: scrutiny happens on triggers, not on everyone.

There was one group the system flagged automatically: people already inside Australia on a student visa applying for another one. "A broad GTE assessment must be undertaken for all onshore applications by holders or previous holders of student visas." Repeat study, assessed against your immigration and study history, was structurally suspicious. If you were hopping from course to course, the system was built to notice.

How are visitor visas risk-tiered?

Visitor visa applications are sorted into statistical risk classes built from overstay data before an officer weighs your personal circumstances. The machinery for this is a public interest criterion called PIC 4011, reproduced in full inside the Department's New Delhi visitor visa processing instructions, released under FOI reference FA 25/10/00377.

PIC 4011 marks you as a "risk factor" case in two situations. The first is personal history: if you applied for any permanent-residence visa in the five years before your visitor application, you are automatically flagged. That is worth knowing if you are, say, an offshore partner-visa applicant who also wants to visit on a tourist visa.

The second is statistical. The Minister can specify an entire class of people as higher risk, and the law says that class must be built from overstay statistics. Specifically, the Department looks at movement records of people who stayed past their visa, and can define risk classes using any of these characteristics:

Two people with identical documents can start from different baselines because the statistics attached to their profile differ. This is the visitor-visa mechanism sitting underneath the broader country-risk framework we cover in our post on how your passport affects your visa.

The New Delhi release also confirms something applicants rarely realise: the Department maintains office-specific processing instructions. New Delhi has its own visitor visa manual, and so do other posts. The operational detail is redacted, but the surviving PIC appendix shows the risk-classification logic it applies. Different offices assessing the same visa differently is not a rumour. It is documented.

If you want to see what this looks like at the sharp end, our breakdown of real visitor visa refusal reasons shows the patterns officers cite once your file reaches a human.

What does this mean for your application?

Whether a computer grants your visa in minutes or an officer reads it for months, one thing never changes: a refusal is always a human decision, and it is based on what that human sees. So the two things most within your control, completeness and consistency, are exactly the two that decide whether a human ever has to get involved, and what they find if they do.

This is the same lesson our own refusal data keeps teaching. A human reading your file is where risk enters the process, because a person can doubt, question, and refuse in a way the fast lane never does. So the aim is not to impress the officer. It is to be so clear and so consistent that the file never needs one.

A few things follow from that:

You cannot change your passport, your age, or your travel history. But you decide how complete and how consistent your application is. That is the part the system rewards, and it is entirely yours to control.

At Tern, this is the whole idea behind how we prepare an application. Before anything is submitted, we check it for the inconsistencies a case officer would catch, so the gaps get closed on our side rather than found on theirs. If you would rather not navigate the triage system alone, that is where we can help.

The documents behind this post

These are the primary source documents this analysis draws on. All are published on Australian Government websites.