Everlasting jobstoppers: How an AI bot-war destroyed the online job market

It was the best of times, it was the … best of times.

Or was it?

According to a wide variety of institutions and publications, the past two years have featured the strongest labor environment in decades. The Commerce Department announced in February of 2023 that “Unemployment is at its lowest level in 54 years.” When this April’s official numbers showed that the U.S. recorded its 27th straight month of sub-4% unemployment, tying the second-longest streak since World War II, the Center for Economic and Policy Research was but one of a multitude of sources celebrating: “This matches the streak from November 1967 to January 1970, often viewed as one of the most prosperous stretches in US history.” In June, Investopedia practically gushed that “U.S. workers are in the midst of one of the best job markets in history. They haven’t had this much job security since the 1960s, and haven’t seen a longer stretch of low unemployment since the early 1950s.”

Arguments about statistical methodology aside, there’s nothing to suggest that those headline numbers were incorrect to any significant extent. But raw unemployment is considered a lagging economic indicator, and there is quite a bit of evidence supporting the premise that, below the surface, the biggest drivers of new employment — online job listings — have become elaborate façades destined to cause more problems than they solve for those seeking work. 

While commentators were singing the praises of America’s labor resiliency, the first stage of the job-hunting meltdown was already showing. That stage came in the form of “ghost jobs,” posts by employers soliciting applications for positions that had already been filled, were never truly intended to be filled or had never really existed at all. While this practice had been expanding for years, its true severity was not well understood until Clarify Capital released a September 2022 survey of 1,045 hiring managers that was the first to focus specifically on the topic of ghost jobs.

Half the managers in question said that one emphatically ambiguous reason they would keep such job listings open indefinitely was because “The company was always open to new people.” That was actually one of the better answers on a list of very bad ones. A tie, at 43%, went to the next most-common responses, “To give the impression that the company is growing” and “To keep current employees motivated.” Perhaps the most infuriating replies came in at 39% and 33%, respectively: “The job was filled” (but the post was left online anyway to keep gathering résumés), and “No reason in particular.”

That’s right, all you go-getters out there: When you scream your 87th cover letter into the ghost-job void, there’s a one in three chance that your time was wasted for “no reason in particular.”

One kind of question that Clarify did not ask, however, was about how many hiring managers were automating their activity. In other words, were they using early forms of what could now be considered management AI to automatically re-submit ghost listings every month, so they would look new? Were the managers using AI to help create the imaginary job descriptions in the first place, or perhaps to identify where the fake ads for fake work would receive the most real attention?

Luckily, there are other sources that not only provide such information, but openly advertise it. Workable, for example, offers employers the service of automatically “cloning/copying” aging job ads — complete with a reminder to delete the words “Copy of” in the titles of any clones/copies (imagine the embarrassment) — in roughly the same sort of bland and straightforward manner that one would expect from a talking pair of pleated khakis. Another company, Propellum, provides job boards with the technology necessary to perform what’s called “job post scraping,” which we’ll let Propellum itself define:

Job post scraping is the process of automatically gathering job posts from a variety of online sources, including career websites, job boards, and company portals. Using this automated method, web pages are accessed programmatically, their content is parsed, and relevant data—like job names, descriptions, locations, and requirements—is extracted… Recruiters can save time and costs by using advanced web scraping AI algorithms to quickly and thoroughly gather information about automated job postings without requiring personal interaction.

What does this mean in plain English? Well, Propellum may claim that this is merely a tool for “recruiters” to use to judge the market for certain positions, but just one click leads to a different page entitled, "Why automate job listings on your job site?” Conveniently, Propellum has the answer, and as you might guess, it involves a profound preponderance of Propellum: “Job automation guarantees the listing of quality and customized jobs on your board. Take Propellum, one of the leading job data automation service providers, for instance. Propellum deploys a host of tools, such as job crawlers and feeds, to scrape the most vaunted jobs present on the internet before posting them onto your job board’s web platform… Contact us to know more irresistible reasons to opt for our automation tools and services.”

Irresistible! Yes, if you run your own listing site, there’s nothing quite so hard to resist as copying job openings from other sites and presenting those openings as your own, regardless of their quality or ghost-iness. Is this for the good of providing another place for employers and future employees to meet? Possibly. Is it because the more raw volume you have, the more traffic is likely to stumble upon your site, and that generates ad revenue? Absolutely. Just ask the second half of that same Propellum page: “As a direct result of this [traffic], you could boost your revenues through advertisements on your website. Websites can earn good money from Google ads and keep their visibility high with intelligent Search Engine Optimization (SEO) and the automation of their job listing process.”

But enough about Propellum, as the editor of this article surely concluded about two paragraphs ago. The big boys in the industry don’t need to outsource their scraping; they can easily do it themselves. Indeed uses the fact that it does so much scraping — ahem, “aggregating” — to make an incredibly bold claim: “Indeed… aggregates job listings from thousands of sources across the internet, including job boards, company career sites, local news sites, staffing agencies and recruiter listings. Since Indeed collects job listings from such a wide range of sources… Indeed delivers 47% of all hires in the US.”

Author Paul Fuhr has a slightly different way of describing Indeed: “[Y]ou’re usually applying to a listing that’s likely dated or dead. That’s the nature of web crawling and job cloning… I personally know a hiring manager who saw one of her own job listings on the site — a middle-management job she’d filled four months earlier. Even after contacting Indeed to take it down, her listing remained there for another five days. Now, think of all those hopeful applicants firing off applications toward a job that wasn’t even there.”

Competitor LinkedIn would like the world to know, according to one of its FAQs, that “Basic job posts are free job posts gathered by LinkedIn from job boards and aggregators across the web to create a comprehensive job-seeking experience for LinkedIn members.” But LinkedIn would also like the world to know, judging by the heavy push in marketing materials, that those creating job ads receive “AI-assisted recommendations to auto-fill and optimize job post attributes.” Additionally, once a listing has garnered some applicants, “Using AI-assisted messages is easy. Once you’ve selected a candidate in Recruiter, all you have to do is click the button that says ‘draft personalized message’ and a unique and personalized message will be crafted for you to review, edit, and send.” Hiring managers can even choose to automate follow-up messages, in case the AI’s first “unique and personalized message” didn’t really hit home.

Then there are the scammers. With so much automation available, it’s become easier than ever for identity thieves to flood the employment market with their own versions of ghost jobs — not to make a real company seem like it’s growing or to make real employees feel like they’re under constant threat of being replaced, but to get practically all the personal information a victim could ever provide. Returning, for a moment, to Paul Fuhr:

Like so many others, I’d missed the news that online thieves have not only leveled up their game, but they’ve put The Unemployed directly in their crosshairs. Armed with AI-driven interactive voice emulators, domain name spoofing, Python-powered web crawlers, the ability to post a fake job while posing as a real company, and some basic 3rd-grade distraction skills, scammers are winning in ways that would’ve seemed like science fiction just a few years ago.

In less than two weeks, I’d applied to, interviewed for, and succeeded in landing a job that didn’t actually exist.

In less than two seconds, I gave scammers everything they needed to secure loans, open utilities, get credit cards, score a Florida driver’s license, and gain access to my bank account.

This is the latest reality for those searching high and low on the internet for work: not only are plenty of companies tricking you into applying, but so are the people who used to pose as Nigerian princes or strangely-incompetent tech support workers. According to the FTC, there were more than five times as many fake job and “business opportunity” scams in 2023 as there were in 2018, costing victims nearly half a billion dollars in total. Technology is expanding the variety of possible con jobs with every passing year; today, with the rapid advancement and proliferation of AI-fueled deepfakes, not even video calls can provide reliable confirmation of who exactly is on the other end.

To say that all the above information is upsetting for job seekers would be akin to saying that the meteor from 66 million years ago was bothersome for the dinosaurs. Social media is awash with those who have been living one of the harsher realities of unemployment for months or even years on end: looking for a job is, itself, a job, and a relatively soul-crushing one at that. The real pain can be felt by listening to some of the people who have been slamming their heads against brick walls that they now believe were often full of bots, clones, ghosts and scams:

Politicians are just talking about how the economy’s so great. I just wanna scream from the rooftops, "Then how come no one can find a job?"

It’s over sixty applications within fourteen days. You know how many companies got back to me? [Three.] You know how many interviews I’ve done so far? Just [one] … I do my interview, they said they had a couple more interviews to conduct and then they’ll call Monday or Tuesday and let me know. That was almost two weeks ago… I’ve been ghosted. I can’t tell you how many times I’ve been ghosted.

The final step is the almost total lack of response you get to job applications. Even just a semi-automated "we will not be going forward with your application" is too much to expect, I guess.

You know what? The self-doubt gets really real, at least it got really real for me around day 30… 40… 50… it’s day 80, okay. And now it’s a pretty regular struggle to be combating these feelings that I have and these thoughts that I have in my brain, these very intrusive thoughts that are saying pretty harsh things to me.

Granted, these are purely anecdotal examples, and any one or even any given series of anecdotes can be (and often are) dismissed by frequently-self-appointed Online Employment Gurus, who will happily explain what’s wrong with an individual person’s résumé or their presentation or their approach to interviews or their attitude or their need to network more aggressively or their lack of willingness to strap on a job helmet, squeeze into a job cannon and fire themselves off into Jobland.

But here’s the thing: While those criticisms may well be accurate for a certain number of anecdotal examples, anybody who searches for more such examples will never, ever run out. They can’t all have jumbled résumés. They can’t all need interview coaching.

Moreover, broad-based statistics confirm the anecdotal messages: Finding work is becoming much more difficult, a trend that started at least as early as 2023, when the average “time-to-hire” across all sectors reached a record high of 44 days. LinkedIn reported in March that hiring on its platform was down almost 10% over the previous year. Pantheon Macroeconomics economist Oliver Allen says that shifts in small business payrolls imply overall private sector job growth “dropping to zero over the next few months.”

The internet, as a collective, has responded to increasing job-hunting difficulty over time in ways that only the internet can. Memes were created. Viral TikToks were recorded. Viral YouTube videos were, too. “Ghost jobs” as a topic got its own Wikipedia page in August of 2023. The Marketplace podcast, ranked #40 in the country in terms of audience reach by Chartable, did an entire episode on ghost jobs this past January. Then there were the Reddit posts. Oh, the Reddit posts. So many Reddit posts.

Far more important, however, the “between jobs” army began to fight fire with fire. The virtual ink on Clarify Capital’s ghost jobs report was barely dry before the world was introduced to ChatGPT in November of 2022. Job applicants could now adjust résumés, customize cover letters and generate answers to any pre-interview questions or tests at speeds that were previously impossible. Optimism was suddenly in fresh supply, even on Reddit. “It felt very meta to have AI writing bullet points that some HR AI would evaluate,” wrote Reddit user Low_Cartographer2944. And that was only the beginning. 

The quaint rudimentary uses of ChatGPT and competing programs in the early days of public AI quickly gave way to software that was more and more specialized to the task of finding and applying for jobs. Sonara, Jobscan, LazyApply, SimplifyJobs, Massive and so many other types of job-hunting AIs now exist that it’s impossible to keep track of all of them. For the relatively tech-savvy, it’s not even difficult to use AgentGPT or similar bot-creating software to conjure up entirely new programs designed specifically to the user’s liking. In other words, anyone can use AI to make more AI; in fact, that first generation of AI could be coded to learn from its operations when constructing an improved second generation, which could go on to build a third generation and so on. The previously hopeless humans, the ones attempting to break through layer after layer of HR software specifically designed to prevent most résumés from reaching a pair of real live eyeballs, suddenly had a technological arsenal of their own.

Want a daily wrap-up of all the news and commentary Salon has to offer? Subscribe to our morning newsletter, Crash Course.

Rather than solving the problems raised by employers’ methods, however, the use of automated job-hunting only served to set off an AI arms race that has no obvious conclusion. ZipRecruiter’s quarterly New Hires Survey reported that in Q1 of this year, more than half of all applicants admitted using AI to assist their efforts. Hiring managers, flooded with more applications than ever before, took the next logical step of seeking out AI that can detect submissions forged by AI. Naturally, prospective employees responded by turning to AI that could defeat AI detectors. Employers moved on to AI that can conduct entire interviews. The applicants can cruise past this hurdle by using specialized AI assistants that provide souped-up answers to an interviewer’s questions in real time. Around and around we go, with no end in sight.

Movie buffs might be reminded of one of the final lines Heath Ledger’s Joker delivered to Batman: “I think we’re destined to do this forever.” Techno-geeks, on the other hand, could be prompted to think of Dead Internet Theory, which posits that the vast majority of public content on the web will soon be generated by bots instead of humans. This, instead, would be the Dead Indeed Theory:

Job listings written by AI and kept “evergreen” by AI cloning/copying will receive résumés and cover letters written by AI, which will then be passed through multiple levels of AI, first the now-archaic software designed to reject applications that lack the requisite amount of keywords and other unique indicators matching the job ad (a system so adorably antiquated that it didn’t merit its own space in this article), then some version of the aforementioned AI-applicant detector. If the detector fails to properly identify the artificial nature of the candidate’s digital paperwork, an AI messenger will send a “personalized” message to that candidate, with potential follow-up communication already written if there isn’t a response within a certain amount of time – although there most certainly will be, as job-hunting AIs will always answer at what they have determined to be the optimal moment. Seeing this reply, an AI scheduler will set up an interview with an AI voice caller or AI deepfake video caller, either of which the potential employee will talk to using optimized statements instantly provided by AI support. If that conversation is satisfactory, then the lucky applicant may be offered a salary that AI has estimated he or she is likely to accept, be hired and onboarded by AI, and, once that process is complete, could even oversee the work of AI interns. Further down the line, however, the news may not be so rosy — another HR AI might determine that this new hire should be included in a round of layoffs.

This is the Singularity of the online job market, the point at which AI growth has become so exponential that humans can’t compete. It is a war against and between the machines, not in the streets and skies but on our desks and in our pockets. And it may kill off the very notion of finding jobs via the internet — permanently.

Thankfully, certain things can still be trusted. Take, for example, what you’re reading right now. A story about AI tricking people, even tricking other AIs, into believing that original content was created by a genuine, authentic, flesh-and-blood individual. At least this was written by a real human being … right?