AI Candidate Sourcing on AcademicJobs: Tapping THE Network for Top Academic Talent

A provost asked me recently why a computational phonology search had been open for seven months with five applicants, none of them a real match. I asked what the hiring team did in the first week. Posted the ad, he said. Then the waiting began.

That is the structural problem AI candidate sourcing was built to fix. Most strong academic candidates never see an advertisement. They are presenting at conferences, publishing on Google Scholar and ResearchGate, and busy in roles they never planned to leave. Posting an ad and waiting catches the people already looking. AI candidate matching reaches the people a department actually wants, whether or not they have opened a job board this year.

The platform doing this work in higher education is AcademicJobs.com, which launched in 2023 and built its AI candidate sourcing product, A.I. Candidate Connect, on top of THE Network, short for The Higher Education Network. The operating team is not new to the sector. It has run academic job boards for two decades, and that history matters, because the product assumes something important: a shortlist of ten credible names beats two hundred unqualified applications every time.

What A.I. Candidate Connect does

A.I. Candidate Connect uses natural language processing, usually called NLP, to read research profiles across Google Scholar, ResearchGate, Scopus and LinkedIn. It does not stop at keyword matching. It looks at research topics, publication context and the terms scholars use to describe their own expertise, then compares that picture against the requirements in the job advertisement. The company's framing is that it identifies the top 20% of people with the specialties you seek, in the area you seek them from.

The speed is the most quoted number. Within roughly 90 seconds of a posting going live, the system notifies up to 100 of the closest specialists and invites each one to consider the role or pass it on through their professional network. These are often passive candidates, the people who would show up late or never in a traditional applicant pool.

Victoria University of Wellington used the tool when hiring a Chair in Generative AI. The role was posted a second time with keyword optimisation tuned for a Google Jobs boost, and an AI podcast was produced so prospective candidates could hear what the university was doing in the field. That combination, an active shortlist plus content aimed at the specific discipline, is different in kind from a classified ad.

Pricing is fixed, which matters in a sector that works on annual approval cycles. Institutions in the United States can post a single listing for US$345, a month of unlimited postings for US$999, or an annual unlimited plan for US$9,000. In Australia the tiers are A$315, A$999 monthly and A$12,000 annually. A.I. Candidate Connect is included in the unlimited plans and was priced at A$200 per role when sold separately in Australia. First-time users get an A.I. Candidate Connect boost free with their first job posting.

Why slow searches cost real money

The wait behind that provost's frustration is measured, not anecdotal. The US Bureau of Labor Statistics projects about 118,900 openings a year for postsecondary teachers through 2033, with employment growing 8 per cent over the decade. Hiring in the sector is slow everywhere. CUPA-HR benchmark data puts the median tenure-track search in the United States at nine to twelve months from requisition to acceptance. A process-mapping study at the City University of New York found searches ran about 386 days on average, with a range of 308 to 461. The University of Central Florida only recently adopted a 45-to-50-day time-to-fill benchmark because it had no standard before.

Every month of vacancy has a price. Institutions in the United States collectively spend between US$55 million and US$100 million a year on external search firm fees, and individual leadership searches routinely run from US$50,000 to US$180,000 plus expenses. Six-figure retainers make sense for a presidency. They are harder to justify for a specialist lecturer or a niche postdoc, which is precisely where fast, targeted sourcing changes the arithmetic.

The people building the tool know those rhythms because they have run boards inside them. Jarrod Kanizay started UniJobs in Melbourne in 2006 and grew it to work with more than 700 institutions worldwide before selling it to TES Global, the parent company of Times Higher Education, in 2015. The acquisition was framed at the time as a step toward matching global talent with global universities. He stayed through the renamed THEunijobs until 2020 and launched AcademicJobs.com in 2023. That lineage is part of why the platform treats a hiring process as a calendar, not a feed.

Different systems run different calendars, and a timeline that reads as thorough in one country reads as indifference in another. An Australian university may run a five-week panel process with an external assessor, though labour-market-testing rules can stretch an academic search past 20 weeks and one documented example ran more than eight months. A Singaporean institution can close in a fortnight with a delegated committee. A German faculty takes a semester and a senate vote before the state ministry signs off. None of these is wrong, but teams advertising internationally need to know the rhythm before the role goes live.

Where THE Network fits

THE Network is the ad-free professional network underneath the job board. Members join specialty rooms, post milestones such as an accepted paper or a grant submitted, and colleagues give kudos, a function borrowed from the fitness app Strava. Research consortia can form around a shared question. The network routes four kinds of opportunity: jobs, speaking and media invitations, research funding leads, and direct contact from hiring committees. The structural choice to carry no ads removes the incentive to keep people scrolling, which shifts the system toward whether replies are useful rather than how long the feed stays busy.

The competitive set explains the bet. LinkedIn reports more than one billion members, but a general feed buries a computational phonologist under corporate recruiters and sales content. HigherEdJobs says it draws about 1.5 million monthly visitors and served 3,500 institutions across 416,000 openings last year. THE's Unijobs board claims 65 million-plus annual visitors. AcademicJobs.com, by company figures, reaches 500,000-plus monthly academic visitors and 57,000-plus live listings across 13,900 institutions, with a network of 15 million academic professionals. Independent Semrush estimates put the domain closer to 246,000 monthly visits. That gap between self-reported and third-party traffic is normal and worth checking at the time you evaluate the platform, not a reason to dismiss it.

Adoption is moving quickly

AI candidate matching has become the most common AI application in recruiting, and the adoption curve in hiring is steep. SHRM's State of AI in HR 2026 survey, fielded in December 2025 with 1,908 HR professionals, found 39% of organisations already using AI in HR and another 7% intending to launch it during 2026. Recruiting is the most frequent use, ahead of HR technology, learning and development, and employee experience. A year earlier, SHRM put AI use for HR tasks at 26%; by 2025 it had climbed to 43%.

Higher education follows the same direction, deliberately. The Institute of Student Employers found 28% of student employers using AI in recruitment in 2025, up from 9% in 2022, while applications hit about 140 per vacancy in 2024. Ninety-four per cent of employers in that survey said AI increased speed and efficiency, yet 71% still prefer a human-centric approach and 67% worry about bias. Those two findings are not contradictory. They describe the sensible operating rule: let the machine widen and sort the pool, keep people in charge of the decision.

The efficiency numbers hold up across sources. SHRM found 89% of AI recruiting users report time savings or efficiency gains, 36% report lower hiring costs, and 24% report a better ability to identify top candidates. That last figure is smaller than the speed gains, and it is the one search committees should watch most closely, because speed only matters if the shortlist contains the right people. The full detail of how student employers are using the technology is covered in the ISE's recruitment survey analysis.

Rules to know before you rely on a score

AI sourcing is legal, but it is regulated, and universities are not exempt from the rules that govern everyone else. In New York City, Local Law 144 has required employers using automated employment decision tools to commission annual independent bias audits, publish a summary of the results, and give candidates at least ten business days' notice with an option to request an alternative process since 5 July 2023. A December 2025 audit by the New York State Comptroller found the city's enforcement of that law ineffective: 75% of test calls to the 311 hotline about AI hiring were misrouted, and the Department of Consumer and Worker Protection flagged one compliance issue across 32 bias audits while the Comptroller identified at least 17. The practical lesson is that a posted audit is not a shield, and employers should expect closer scrutiny, not less.

The European Union's AI Act entered into force on 1 August 2024 and treats AI used for recruitment and candidate evaluation as high risk. After a May 2026 agreement, the full high-risk obligations for standalone systems are due to apply from 2 December 2027, but two deadlines are already live: AI literacy measures have applied since February 2025, and transparency duties for candidates from August 2026. Deployers must ensure meaningful human oversight, keep logs for at least six months, and inform workers and their representatives before deployment. Penalties for deployer breaches can reach EUR 15 million or 3% of global turnover. The Illinois Artificial Intelligence Video Interview Act has required notice and consent for AI analysis of recorded interviews since 2020, with later amendments adding reporting duties.

None of this forbids AI sourcing. It forbids treating a model's output as a decision. The operating rules that keep an institution safe are simple enough to write down:

• commission an independent bias audit before deployment where the law requires one, and repeat it annually
• give candidates notice when an automated tool will be used, including how to request an alternative
• keep a documented human override so no application is rejected on an automated rank alone, and a person records a substantive reason for every removal
• retain audit logs and model version records for at least the period your jurisdiction prescribes

The part only people can do

An algorithm can surface a shortlist in 90 seconds. It cannot book campus visits, chase a reference who is on sabbatical, write a clear offer letter, or tell a candidate what the teaching load really is. The universities that fill hard roles fastest are not always the richest ones. They are the ones where somebody owns the candidate's week-to-week experience by name.

That somebody is usually an administrator. The coordinator who confirms interview dates across four time zones, formats the scoring rubric so the committee actually uses it, and answers a finalist's question about spousal hire policy the same day. AI candidate sourcing hands that person a better starting list. It does not replace the work of managing the relationship afterward, and institutions that forget the second half will lose the candidates the algorithm just found. Employer branding in higher education works differently from corporate branding, and research support, teaching loads and a credible promotion path do more than any perks page ever will.

Candidates notice the difference immediately. Research on recognition finds that employees who do not feel adequately recognised are roughly twice as likely to say they will leave within a year. A prompt, specific reply after an interview is recognition in its simplest form, and it costs nothing. When an AI tool shortens the time to a credible shortlist but the committee then sits on it for five weeks, the speed gain is forfeited.

One step to take this month

Do not pilot this on a safe, easy-to-fill role. Pick the search you are afraid will stay unfilled, a niche chair, a specialist postdoc, a lecturer in a field where good people are rare, and run it as a 60-day test. Post it with A.I. Candidate Connect enabled, the first posting carries the boost free, and keep a simple log: minutes from posting to the first notifications, how many contacts reply, how many replies become applications, and when the hiring manager first sees a real name.

At the end of the 60 days, compare the log against the same search from last year, line by line. Time to first credible names, not total applications, is the measure that matters. That comparison will tell you, with your own data, whether the tool changed the search or just modernised the advertisement.

Photo by Zulfugar Karimov on Unsplash