A skilled engineer arrives in Toronto with a decade of experience, permanent residency in hand, and a CV that would clear the bar in most labour markets. She applies to sixty roles in her field and hears back from none. Somewhere between her application and a recruiter's inbox, an algorithm decided her international experience did not count. Simon Trevarthen, CEO of the Toronto Region Immigrant Employment Council (TRIEC), argues this scenario is not an edge case: it is the predictable output of AI-powered resume screening tools that automate a longstanding bias against foreign credentials, filtering skilled immigrants out of hiring pipelines before a human being ever opens the file.

The numbers around that gap are stark. Newcomers with permanent residency face unemployment at nearly double Canada's roughly 6% national rate, and underemployment, the mismatch between qualifications and the work someone actually does, affects an estimated 25 to 30% of newcomers compared with about 12% of Canadian-born workers. For Employment and Social Development Canada (ESDC), which administers the programmes meant to connect skilled immigrants to the labour market, those figures represent a structural return-on-investment problem: Canada's points-based immigration system selects for education and experience, then the private-sector hiring funnel quietly discards both.

How the filter actually works

Most large employers in Canada no longer read every CV that arrives in their applicant tracking system. At sufficient volume, that is a physical impossibility. Trevarthen cited one major financial institution, consistent with Canada's Big Five banking sector, that receives over one million applications for 2,000 roles annually. At that scale, automated pre-screening is not a preference; it is the only way the funnel keeps moving.

The problem lies in what those systems are trained to reward. Resume screeners typically weight signals that hiring managers have historically valued: recognisable employer names, familiar university brands, local job titles, and a track record of work inside Canada. For a newcomer whose degree, employers, and job titles all sit outside that reference frame, the algorithm's confidence score drops. The application is deprioritised or filtered out before any recruiter reviews it.

Trevarthen is direct about the structural flaw: there are implicit biases within AI systems, and two slightly different prompts fed into the same model can produce quite different results and quite different recommendations.

That fragility has real consequences. If the same candidate can be ranked strong or weak depending on how a prompt is phrased on a given afternoon, the screening decision is less a measurement of fit than a reflection of the model's training data and the recruiter's wording choices. Canada's Canadian Human Rights Commission (CHRC) has noted, in its guidance on technology-facilitated discrimination, that automated employment decisions that produce disparate outcomes on protected grounds, including national or ethnic origin, may engage the Canadian Human Rights Act, R.S.C. 1985, c. H-6, s. 7. Provincial human rights codes, including Ontario's Human Rights Code, R.S.O. 1990, c. H.19, s. 5(1), extend equivalent protection at the provincial level, which covers the roughly 94% of workers not in federally-regulated sectors.

The AI-versus-AI arms race

Candidates have adapted. Faced with screening systems they cannot inspect, many now use generative AI tools to write, rewrite, and tailor applications at scale. The result is what Trevarthen calls a blizzard of AI-generated applications that recruiters are trying to wade through. The irony is layered: a process originally designed to handle volume has triggered a further surge in volume, and the signal-to-noise problem it was meant to solve has deepened.

The blizzard ultimately disadvantages the candidates the process was meant to surface. When every application arrives keyword-optimised and structurally polished, a screener's differentiating signals default back to the features it was trained on: familiar employer names, familiar geographies, familiar career trajectories. Newcomers lose ground again, not because their skills are weaker but because their career shapes are less legible to a model trained on a narrow slice of the labour market.

For employees and job seekers

If you have moved to Canada recently, or are in the process of doing so, the screening pipeline is a real structural obstacle, not a matter of presentation. Statistics Canada / Statistique Canada survey data cited by Trevarthen shows that about 40% of newcomers report being told they lacked sufficient Canadian experience, while roughly a third said their international experience was not accepted by employers during hiring. Trevarthen attributes this pattern to a combination of lengthy credential recognition processes, employer expectations of local work experience, and limited professional networks in the Canadian market.

Several concrete moves can help you get past the filter rather than simply hoping the system will recognise your value:

• Translate your CV for the Canadian context, not just the language. Match your job titles to their Canadian equivalents, spell out the scale of previous employers using revenue, headcount, or market position, and adopt the vocabulary the target market uses. A title like Senior Analyst, Retail Banking travels across borders; an internal designation that only makes sense inside one specific organisation does not.
• Front-load transferable achievements. Screeners weight the top third of a document heavily. Lead with quantified outcomes using percentages, CAD figures, or headcount, not with a chronological employer list the system will not recognise.
• Route around the automated funnel where you can. Referrals, professional associations such as CPHR Canada's provincial member bodies (HRPA Ontario, CPHR BC/Yukon), and immigrant employment councils like TRIEC bypass the initial screen entirely. When your application reaches a recruiter through a person, the algorithm's confidence score matters far less.
• Pursue the interview, not just the application. Trevarthen is clear on this point: an interview still remains one of the best ways to assess a candidate. Anything that moves you from a document to a conversation shifts the odds in your favour.

For HR leaders and recruiters

For those running recruitment functions, the newcomer gap is not solely an equity story. It is a talent-loss story with measurable cost. Trevarthen frames the underlying problem plainly: when you look at underemployment, you find a mismatch between people who have foreign qualifications and the level of work they are actually doing. In recruitment terms, a segment of your pipeline carrying the precise skills your organisation reports difficulty finding is being systematically deprioritised by the tool you are using to find them.

The logic of Canada's immigration system reinforces the point. As Trevarthen puts it, the reason for an immigration system is to attract talent to Canada. When the screening infrastructure works against that logic, the loss sits on the employer's side of the ledger, in missed capability, lower diversity outcomes, and a narrowed talent pool for hard-to-fill roles.

A structured audit of your AI screening set-up is worth initiating now, before regulatory pressure makes it compulsory. Useful questions to put to your HR technology vendor and internal team include:

• What training data underpins the screener? If the model learned from your company's historical hires, it has absorbed the preferences of past hiring managers, including any bias toward local experience or recognisable employer brands.
• What does the confidence score actually reward? Ask for the top ten features driving the ranking output. If Canadian experience, local employer name, or any proxy for either appears near the top, the model will not correct that on its own.
• Where does human judgment enter the funnel? Trevarthen's framing is worth adopting: understanding that AI is a tool and balancing it with human judgment is what produces the right hiring decision. If a recruiter never sees any application below the machine-set threshold, the tool is making the final decision, not supporting one.
• Do you sample the rejected pool? A periodic review of a random sample of screened-out applications, assessed by an experienced human recruiter, is the most direct way to catch systematic bias before it hardens into unexamined policy.
• What is your regulatory exposure? Canada's proposed Artificial Intelligence and Data Act (AIDA), currently progressing through Parliament, would impose transparency and accountability obligations on high-impact automated systems, including those used in employment decisions. Vendors that cannot produce an audit trail today represent a compliance liability as that legislation advances. Separately, CHRC guidance already applies the Canadian Human Rights Act, R.S.C. 1985, c. H-6, s. 7 to automated hiring tools that produce discriminatory outcomes on protected grounds.

The trade-off Canada HR leaders need to name

Pure AI throughput and genuine talent evaluation are, at the current state of available tools, in tension. A screener tuned for volume rewards pattern-matching against historical data. A hiring process tuned for genuine capability requires human review at the shortlist stage, and that review takes time.

That trade-off is where the newcomer story stops being a Canada-specific challenge and becomes a general question about what hiring technology is actually optimising for. Canada's immigration system selects for skills and credentials that the private sector says it needs. When the same private sector deploys screening tools that discard those skills before a recruiter reads the file, the gap is not a newcomer problem. It is a system-design problem, and the cost of leaving it unaddressed falls on employers, on the newcomers who carry real capability into a market that cannot see it, and ultimately on the labour force productivity that Statistics Canada / Statistique Canada tracks in its quarterly Labour Force Survey releases.

For candidates, the practical response is to fight for the interview. For HR leaders, it is to audit the screener before the screener quietly audits your talent pipeline on your behalf, and long before CHRC or a provincial human rights tribunal does it for you.