At a glance
Boolean strings match keywords, not people. How academic hiring teams can widen the search across more than 5,000 specialties on THE Network, and how to test the claim before spending a listing budget.
The string finds the words, not the person
The question came from a faculty HR partner in Melbourne and it was narrower than most: how do we find the person we need when the person we need doesn't use our words?
She had a lectureship in computational linguistics open. Her saved search read: ("computational linguistics" OR "natural language processing") AND PhD AND (teaching OR lecturer). It returned a workable pile. That is the difficulty. A pile that looks workable stops the search.
Boolean search is the query grammar most recruitment databases still run on: keywords joined by AND, OR and NOT, quotation marks holding a phrase together, brackets setting the order of operations. It descends from George Boole's nineteenth-century logic and it is genuinely useful. AND narrows. OR widens, which is how you catch the person who wrote "NLP" instead of "natural language processing". NOT removes. Capitalisation matters, because LinkedIn requires AND, OR and NOT in upper case or the string quietly does something else. Google doesn't recognise NOT at all, so you use a minus sign with no space before it.
None of that solves my partner's problem. Her string asks a database whether a profile contains certain characters. It does not ask whether a person can teach phonology to second-year undergraduates.
Four names for the same job
A lecturer in Melbourne, an assistant professor in Ohio, a maître de conférences in Lyon and a W2 professor in Bavaria can carry broadly equivalent standing and share no vocabulary at all. Add the DPhil, the habilitation, the "senior research associate" that means a permanent post in one system and a soft-money post in another. Titles drift between sectors and between countries, and a keyword search inherits every mismatch in reverse: it pulls every discipline that uses the word and misses every discipline that calls the same job something else.
Universities know this well enough to have built job titles around it. Monash University recently advertised a talent sourcing position that names Boolean search, LinkedIn Recruiter and talent mapping as the daily work, alongside a named applicant tracking system. The skill has become a formal post rather than something a committee chair does at night.
The mismatch gets worse as the field gets smaller. A department hunting a computational phonologist doesn't need two hundred profiles; it needs ten credible names and a way to reach them. A general search returns the ten people who described themselves in the committee's dialect, which is not the same as the ten best people.
What keyword filtering quietly costs a faculty
Harvard Business School and Accenture put a number on this in 2021 and nobody has really improved on it. Their report, Hidden Workers: Untapped Talent, estimated that 27 million people in the United States were being screened out of jobs they could do, because automated systems sorted them on what their application lacked rather than what they could bring. Three quarters of employers were using applicant tracking software to do the sorting, and 88 per cent of the executives surveyed admitted that qualified, high-skilled candidates were being vetted out for failing to match the exact criteria in the job description. The Harvard Gazette interview with the authors is worth an hour of any search committee's time.
Joseph Fuller, who co-chaired the research, made the point about veterans in a way that transfers almost directly to academia: their skills "don't match with the skill descriptions employers are seeking." Swap veterans for a postdoc returning from two years of parental leave, a scholar whose last title was research fellow rather than postdoctoral researcher, or anyone whose doctorate was conferred as a DPhil. Half the companies in that study used a filter to exclude applicants with a work gap longer than six months.
Faculty hiring adds its own trap. Job descriptions get drafted from the last appointment, then amended each cycle until the person you want must match a list nobody has read end to end since 2019. Filtering for all of it is how a department ends up with a shortlist of people who are already employed somewhere similar, which is to say a shortlist of people who were never going to move.
Photo by Markus Spiske on Unsplash
The case for keeping a string you can defend
Boolean is not obsolete, and the loudest people arguing for its retirement usually have a product to sell. A string is deterministic. You can print it, read it, hand it to a search committee, and answer a candidate who asks why they weren't approached. That matters more in academic hiring than in most commercial recruiting, because appointments are auditable by nature and committees are expected to explain their reasoning.
It matters legally too. If you source candidates in Europe rather than waiting for applications, the General Data Protection Regulation requires you to tell those people how their data is being processed, generally within a month of first contact, and regulator guidance points to holding unsuccessful candidate data for no more than six to twelve months without explicit consent. Recruitment AI is classified high-risk under the EU AI Act, with those obligations applying from 2 December 2027. A readable search string is evidence. A relevance score you can't inspect is not.
The mechanical limits are real regardless. Reed.co.uk, which holds around 15.5 million searchable CVs, caps the keyword field at 3,000 characters, so a string thorough enough to be useful is often long enough to be unusable. Its recruiter guidance is a plain-English primer on the operators. And no string can report what it missed. Benchmarks in this market measure precision, the share of returned profiles that fit. Recall, the share of fitting people who never came back, is the number that decides whether your search worked, and no vendor publishes it.
What THE Network is actually selling
This is where AcademicJobs.com has placed its bet. THE Network, the company's academic networking layer, organises members into specialty rooms instead of a feed, and the company describes coverage across more than 5,000 specialties. Membership is free for verified faculty, researchers, leaders and professional staff, and there is no advertising. That changes the incentive: an ad-funded network profits from keeping you scrolling, a room-based one from whether the replies are useful.
The hiring layer underneath is called A.I. Candidate Connect. When a job is posted, it reads profiles on Google Scholar, ResearchGate, Scopus and LinkedIn alongside the platform's own data, compares research topics and publication context against the advertisement, and notifies up to 100 of the closest matching specialists, which the company describes as roughly the top fifth for the role. Most outreach goes out within about 90 seconds of a posting going live. Pricing is published: US$345 for a single U.S. listing, US$999 for a month of unlimited postings, US$9,000 for a year. Candidate Connect comes with the unlimited plans and has been sold separately in Australia at A$200 per role.
Two things deserve saying plainly. Speed of notification isn't a shortlist, and no algorithm sits on an interview panel. The second is density. A specialty room is only useful if the people in your field have joined it, and you can't audit that before signing up because the directory sits behind a login. The platform's own launch material concedes that it won't find anyone a job by itself and won't make a search committee move faster. Take that at face value and you have a discovery tool. Sell it as a pipeline and it will disappoint you.
Worth holding alongside the marketing: the self-reported scale figures, around 72,000 listings from 13,000 institutions, are not audited. Founder Jarrod Kanizay built UniJobs in Melbourne from 2006, sold it to TES Global in 2015 and left in 2020 before launching AcademicJobs.com in 2023. Whatever the company ships next, the question for a committee stays the same: are the ten people you actually need in the room?
The competition isn't really LinkedIn either. A PLOS ONE case study of researchers at the University of Caen Normandy found 75.1 per cent were not interested in using author identifier services and 40.2 per cent were not interested in academic social networks, citing the friction of maintaining several profiles. The rival is apathy, plus the cost of one more login, which is why the profile-building step has to be almost effortless. It's also why a well-built academic CV still does more for a candidate than any profile summary.
The diversity claim, after the statements came out
AcademicJobs.com now positions itself as the job portal for diversity in higher education, and that phrase needs translating, because the ground beneath it moved. Heterodox Academy's analysis of more than 16,500 full-time faculty advertisements from the autumn 2025 cycle found explicit requests for applicants to address diversity, equity and inclusion had fallen from about 25 per cent to 11 per cent in a year. Standalone diversity statements collapsed from 13.4 per cent of ads to 2.5 per cent. Seventeen states had laws restricting them by December 2025, according to the full Heterodox Academy report.
The variation is the interesting part. Thirty-four per cent of faculty ads in California carried an explicit diversity requirement, against less than 1 per cent in Kansas and Alabama. Community colleges went the other way, rising from 4.8 per cent to 8.3 per cent. And 37 per cent of ads didn't ask for a statement at all while still signalling that a commitment would be valued, which puts roughly 48 per cent of postings carrying some expectation, down from about two-thirds the year before.
Read that the way a candidate reads it. A job advertisement no longer tells you whether the institution means it. So the diversity question has moved upstream, from what you ask applicants to write to whom your search actually reaches.
The University of Sydney's engineering faculty shows what upstream looks like. Fewer than one in five academics there are women, and in five schools running a dedicated recruitment campaign, between 12 and 17 per cent of continuing academic staff are female. The faculty reserved academic roles for women only, one of the largest such campaigns in the discipline's history, directed by biomedical engineer Renae Ryan, as the Guardian reported. Diversity Council Australia found 14 per cent of qualified engineers identify as female, only 4 per cent of those were born in Australia, and the gender pay gap in engineering runs at 24 per cent against 14 per cent across all industries.
Now run a keyword search for "engineering lecturer" and see how much of that pool it finds. Almost none of it, because the pool isn't applying, and much of it isn't described in the words a committee would type.
Photo by Rahul Mishra on Unsplash
A worked search, and the person who should run it
Say a department needs someone in coastal sediment modelling who can also teach field methods. The string most committees write is ("coastal" OR "sedimentology") AND ("modelling" OR "modeling") AND PhD. It works, roughly, and it is full of holes. Modelling with one l and with two is two different searches. A candidate whose work is described as morphodynamics may never use the word sediment. Someone in the Netherlands or Germany may list a habilitation instead of a PhD.
A better string, and a shorter one: ("coastal" OR nearshore OR estuarine) AND ("sediment transport" OR morphodynamics OR "morphodynamic modelling") AND (fieldwork OR "field methods") NOT (undergraduate OR "PhD candidate"). Fewer OR clauses, each earning its place, and a NOT group that removes the two categories which always flood the results. Then run it, read the first twenty profiles, adjust. Boolean is a conversation, not a formula, and the first string is never right.
Who runs it matters more than what's in it. In most faculties the person who actually owns the shortlist is an administrator or an HR partner, not the chair. They are the ones who notice the same eleven names appearing across three separate searches, who know that a candidate's file went missing for a fortnight, and who can tell you whether the specialty room is full yet. Give that person the time, the title and the standing to do the work properly, and the process stops depending on a chair's evenings. I've seen searches turn around on that single change more often than on any software purchase.
For candidates on the other side of the same transaction, the arithmetic is unforgiving. Collaborators with complementary methods are needed before a tenure clock starts, not after, which is why the pipeline work described in our guide to thriving in a postdoctoral role belongs in year one rather than year three. The postdoctoral labour market itself keeps expanding at the base: the U.S. Bureau of Labor Statistics projects postsecondary teaching employment to grow about 8 per cent between 2023 and 2033, with roughly 118,900 openings a year, and most of those are replacements rather than new seats, per the Occupational Outlook Handbook. A replacement market rewards being findable long before you need to be.
Which brings me to the one thing worth doing before the next round opens. Take the string your faculty normally uses and give it to the administrator who runs the shortlist. Ask them to run it, then to run the same search in plain language with no operators at all, and to put the two lists side by side. The names on the second list and not the first are the people your process was about to lose.
That comparison costs an afternoon. It tells you more about your reach than any coverage figure a vendor will quote you, and it gives the person who already does the invisible work a reason to be in the room when the decision is made.
- academic-recruitment
- faculty-hiring
- higher-education-recruitment
- academicjobs.com
- diversity-hiring

