AI in hiring 2027, in Australia, means LinkedIn's recruiter agent, applicant systems and some voice screens can read you before a person does. Which ones an employer uses varies. Three things help most. Put your level, industry and city in your LinkedIn headline. Reply to every recruiter message within a day. Write each application yourself.
Last updated 5 October 2026. I checked every source below that month.
A few weeks ago I typed a search into one of the AI sourcing tools recruiters are being sold. I wrote it the way I'd say it to a colleague: find me an AI engineer with RAG experience in Sydney, one to two years in. It came back with four people. It hadn't searched LinkedIn's own database. It had found them through Google and GitHub, and pulled their contact details from public pages. The tool also wrote a one-line summary of each person for the recruiter, built from what their profile said, so the recruiter read the tool's version of them first.
I've run that kind of test more than once, and the thing I keep noticing is that in the sessions I run, candidates are rarely told it's happening. You're not applying for a job in that moment. A tool is deciding whether you exist. I call it vibe sourcing: the recruiter describes the person in plain English, and the tool hands back a pack. In one of my tests, several people in the pack hadn't filled in their Experience section at all. One had been picked up because his profile mentioned "financial" and "data" lower down. Words anywhere on your profile can be the reason you appear.
Here's the whole picture in one table, then the nine steps.
| The read | What does the reading | What it looks at | Your move |
|---|---|---|---|
| 1. Finding you | LinkedIn's recruiter agent, AI sourcing tools | Your profile, your public work, your Open to Work settings, your InMail replies | Plain words for level, industry and city. Reply fast. Publish one piece of work. |
| 2. Scoring you | The employer's applicant tracking system | Your CV against the ad, plus signs of automation or fraud | A CV that matches your LinkedIn. Apply by hand. |
| 3. Screening you | A voice or phone pre-screen, at some employers | Your spoken answer | One opening answer, said out loud once. |
| 4. Deciding | A person | What the first three put in front of them | Evidence they can check. |
If you want what happens to a single application after you send it, that's in the five things changing in Australian and New Zealand hiring before 2027. This post is about the other half: how you get found, and what's reading you before anyone calls.
They take a plain-English search from a recruiter and match it against your LinkedIn profile, your public work and your Open to Work settings.
Put your level, industry and city in your headline and your About section, for example "Senior Product Manager | Retail and marketplaces | Sydney". LinkedIn announced Hiring Assistant 2 on 29 September 2026. It's the AI agent recruiters use to find and shortlist people. LinkedIn says customers who already use it in English will update automatically, for free, beginning in November (LinkedIn newsroom, 29 September 2026).
LinkedIn says it now searches on what a recruiter means, not only what they typed (LinkedIn Talent Solutions, 29 September 2026). Its example is a product manager close to the office, in retail, at mid level, because those are the people that recruiter hired before. That's one recruiter's hiring history. I'd read it as a hint, not a rule. I'd expect a search to lean on your level, your industry and your city, if your profile says them.
An InMail is a message a recruiter sends you on LinkedIn. On your profile, open Open to, then Finding a new job, and choose Recruiters only. LinkedIn says it takes steps to hide this from recruiters at your current company, but it can't guarantee complete privacy (LinkedIn Help). Check your InMails once a day, and reply to every one within a day, even to say no. Leave Open to Work on, except when you do the two-week off and on reset in the 30-day job search plan.
Why? LinkedIn says the shortlist now carries a "candidate interest" signal, an indicator of how receptive you're likely to be. It's based on signals like platform engagement, Open to Work details and InMail responsiveness (LinkedIn Talent Solutions). "Thanks, not the right fit, keep me in mind for other roles" takes a minute and counts as a reply.
In September, from my own reading, I told clients to reply to a recruiter even when they don't want the role, because the platform treats you as more responsive. LinkedIn said it out loud on 29 September. I rarely hear clients mention it. It's the cheapest way I know to look responsive.
On your profile, click Resources, then About this member, then Verify now. For Australia and New Zealand, LinkedIn's Help page lists a passport with a chip and a selfie. You do it in the LinkedIn mobile app. The page lists no driver licence (LinkedIn Help, checked October 2026). LinkedIn said on 29 September 2026 that 115 million members are verified. It also said verification will now show up in a recruiter's own system, with partners like iCIMS, JazzHR, Jobvite, Lever, SmartRecruiters and Workable (LinkedIn Talent Solutions).
In research LinkedIn commissioned, 39% of talent acquisition professionals said their top challenge is not knowing if a candidate or their skills are real. The survey covered 500 US recruiters and 500 US executives at companies with 200 or more employees, in August 2026, so it isn't Australian data (LinkedIn newsroom). Verification is optional, and it means uploading an ID, so it's your call.
Publish a public repo, a write-up or a LinkedIn article, and link it from your profile. LinkedIn is adding connected apps with partners like Behance and Replit, and including Medium, so your profile shows work you did (LinkedIn Talent Solutions). They only help if you use them for your work. Most of the partners are creative or AI tools, so a finance or operations profile may have nothing to connect.
I think the work you publish outside LinkedIn matters more, because it's what the tool I tested found. AI sourcing tools say they pull from LinkedIn, GitHub and other public sites. hireEZ, one of them, publishes a guide to sourcing candidates on GitHub (hireEZ, July 2024). I've seen recruiters ask ChatGPT or Gemini the same question directly. How to be found by an AI recruiter search shows what to fix on the profile itself.
An applicant system doesn't throw your CV out on its own, at least according to Greenhouse and Lever, who say fraud flags never auto-reject. Even so, scores and flags reach the person before your CV does.
Put your CV and your LinkedIn profile side by side, and compare every title, date and skill. Fix any that differ. A company's applicant tracking system is the software that stores its applications. Once a company connects it to LinkedIn, Hiring Assistant 2 can evaluate people who applied outside LinkedIn too. Today that works with Greenhouse, Lever, SmartRecruiters, Zoho Recruit and Tracker, and iCIMS and Workday are coming soon, according to LinkedIn (LinkedIn Talent Solutions).
LinkedIn's newsroom also says signals including verification, applicant tracking system insights, pre-screening responses and relevant work experience are now brought together in one place (LinkedIn newsroom). I'd expect an old application to get looked at again for a new role, but I haven't seen that confirmed for Australian employers.
Write or tailor each application yourself, and stop using auto-apply tools. Apply only when your CV says the ad's main requirements in your own words, and you've written to someone inside first, as in the 30-day plan.
I once told a client who'd sent 180 applications for two interviews, "You've destroyed the market for yourself already. You are everywhere." Each employer sees its own record of you, and 180 scattered applications reads as volume, not fit.
Automation is the part the systems are starting to measure. Employers using Workday can switch on reporting. It includes an automation score, the IP address and country, how long the application took, and how many of that employer's roles you applied to. That comes from a Workday partner's summary of the 2026R1 release, not Workday's own release notes. It doesn't say any of it is used to reject (Coreteam Group, February 2026). Lever describes fraud signals too, each one labelled verified, suspicious or flagged for review, and says no one is ever automatically rejected (Lever, May 2026).
Greenhouse said in October 2025 that Real Talent, which helps businesses fight spam and fraud in applications, would be available in Australia in early 2026 (AAP, October 2025). I haven't found a page that confirms it's live here. Greenhouse's own team wrote in February 2026 that one signal is a prompt to look closer, and several together are a strong indicator of fraud. In one pilot for a machine-learning role, more than 35% of applicants were flagged as having high-risk signals. Its security team reviewed some of those further, and 91% of the ones reviewed were confirmed as fraudulent (Greenhouse, February 2026). Those are Greenhouse's own numbers from one role. Greenhouse also says the decision to reject a candidate should be a human one, not made automatically by a system.
I suspect sending a hundred applications through a tool is the pattern these checks are built to notice, though no source says a hundred triggers anything.
Here's what the applicant systems are doing to the person who opens your file:
| System | What it says it does | Source |
|---|---|---|
| Workday HiredScore | Grades candidates, which Workday calls AI-driven candidate grading | Workday |
| SmartRecruiters Winston Match | Lets a recruiter show sub-scores for skills, experience and education (March 2026 release) | SmartRecruiters |
| SEEK Assist | Scores and ranks applicants against the hirer's criteria and delivers a shortlist, with an explanation of why each is a strong fit | SEEK |
Greenhouse says the reject decision should be a human one, and Lever says no one is ever automatically rejected. Workday's page doesn't say either way, and neither does SEEK's. What has changed is what the person sees first: scores, summaries and flags written by a program. Which checks an employer has can depend on the plan they pay for. Greenhouse's identity verification, for example, is only on its Plus and Pro tiers (Greenhouse support), and Workday sells its AI through a credits model (Workday, September 2025). You can't tell from outside, so assume you're scored.
Some employers now screen by voice or phone before a person speaks to you, so prepare one short opening answer.
Write your opening answer to "tell me about yourself", and say it out loud once. Give your current scope and what you specialise in, in a few sentences, with one line of colour. Skip the tour of your history, because the history comes up later in the process. LinkedIn says it's adding voice to the pre-screening that Hiring Assistant runs for recruiters (LinkedIn Talent Solutions). SEEK says SEEK Assist runs AI voice interviews with an employer's top candidates (SEEK). Greenhouse's June 2026 release notes list Puck, a partner, whose integration offers AI phone interviews (Greenhouse release notes).
Which of these a given employer uses isn't visible from outside. So treat any unexpected call or voice prompt as a real screen.
From 10 December 2026, some Australian organisations must say in their privacy policy when a computer program helps make a decision that significantly affects someone (OAIC, September 2026).
If it says a program screens or ranks applicants, write your CV for a tool and a person: use the ad's own words, with evidence next to each one. Put a reminder in your calendar for 10 December.
The rule applies when an organisation uses a computer program to make a decision, or to do something substantially and directly related to making it. The decision has to be one that could significantly affect someone, and it has to use personal information (OAIC, September 2026). The rule covers organisations under the Privacy Act, generally those with turnover over A$3 million (OAIC). They must list the kinds of personal information used and the kinds of decisions involved.
The OAIC's guidelines give an example that fits this post. An organisation buys another company's AI system to screen and rank job applications (OAIC APP 1 guidelines, updated September 2026). The same guidelines say a decision can be in scope even when a person reviews the program's output.
It's a disclosure rule. The rule requires a statement in the privacy policy. On the guidance I read, it doesn't create a right to an explanation of your result, and the policy only lists kinds of decisions. It won't show how a tool scored you. It comes from the Privacy and Other Legislation Amendment Act 2024, and the OAIC is the Office of the Australian Information Commissioner. It's an Australian law. New Zealand's Privacy Act has no matching rule, though a New Zealand company doing business in Australia can be caught by it (Information & Data Manager, April 2026). That last point comes from a trade report rather than the regulator, and I haven't covered New Zealand's own rules.
Ask ChatGPT, Claude and Gemini to find you on the first of each month, and fix one gap at a time.
Type "find me a [your title] in [your city]" into ChatGPT, Claude and Gemini. Write down whether you appear. AI sourcing tools say they pull from LinkedIn, GitHub and other public sites. So a miss tells you something is missing from what they can read, such as your level, your city or a piece of public work. It doesn't prove a recruiter would miss you.
Fix one gap, then run it again next month. Step 4 is the usual fix.
A paper first posted in 2025 and revised in June 2026, by researchers at the University of Maryland, the National University of Singapore and Ohio State, tested this (arXiv). In a controlled experiment, the larger AI models were 67% to 82% more likely to pick a CV they'd written themselves over an equivalent human-written one. The researchers also ran a simulation across 24 occupations. Candidates who used the same model as the screener were 23% to 60% more likely to be shortlisted. The comparison was with equally qualified candidates who submitted human-written CVs.
It tested 2025-era AI models on controlled CVs. It didn't test any named employer's tool, and nothing Australian. You can't know which model an employer's tool uses. So I'd still write your CV from your own evidence, because that's what an interview tests.
"Referrals are five times more likely to get hired" has no source I could trace. The Glassdoor figures in the 30-day plan are the ones I'd use. LinkedIn's headline multipliers for its recruiter agent, and its line about connected apps and recruiter messages, have no published method behind them, so I haven't used them.
It's working when your headline names your level, industry and city, your recruiter messages get a same-day reply, and the monthly find-me test turns you up. Check each one:
It works least when your work is private, you're early in your career, or you're applying outside Australia.
Two tools in Careersy AI line up with this. AI Discoverability scores how visible you are to AI-driven recruiter sourcing. ATS Score shows your match percentage against an ad and the keywords still missing. Every account starts with 10 free messages on us, no card needed, at app.careersy.ai/signup.
If you'd rather do it by hand, that works too. If you only have ten minutes, do the step 9 test and a Ctrl+F on your CV for the ad's main keywords.
AI now sits in three places before a person decides. LinkedIn's recruiter agent finds and shortlists people, applicant tracking systems score applications and check for fraud, and some employers run voice or phone pre-screens. Each one reads your profile or CV, so plain words, matching documents and public work matter more.
They take a plain-English description of the person a recruiter wants, then search LinkedIn, GitHub and other public sites. In September 2026, LinkedIn said its Hiring Assistant 2 searches on what a recruiter means. Its example was a product manager close to the office, in retail, at mid level. My reading is that a profile stating level, industry and city in plain words gives the agent something to match.
Yes, as one input. LinkedIn said on 29 September 2026 that its recruiter agent shows a candidate-interest signal, based on signals like platform engagement, Open to Work details and InMail responsiveness. The wording is "signals like", so those are examples, and LinkedIn doesn't say how much each one counts. Keeping Open to Work on for recruiters only, and answering every InMail within a day, are the two things in your control.
Yes. SEEK Assist, Workday HiredScore and SmartRecruiters Winston Match all say they score, grade or rank applicants. From 10 December 2026, some Australian organisations must say in their privacy policy when a program helps make decisions that significantly affect people (OAIC, September 2026). It's a disclosure rule, and on the guidance I read it doesn't give you a right to ask why you were rejected.
One paper first posted in 2025 found that the larger AI models were 67% to 82% more likely to pick a CV they'd written themselves (arXiv). It tested 2025-era models. It didn't test any named employer's tool, and nothing Australian. Write your CV from your own evidence, because an interview tests that.
The panel decides from their notes, so every answer has to give them something to write down. Here are the six steps I run with clients before an interview, with the exact Careersy AI prompts that grill each story before the panel does.
20 min read
A 30-day plan with the order that works: get recruiters finding you, get people referring you, and apply last. Includes the numbers to check and when to reset.
18 min read
