Question
In my LinkedIn profile, I have listed that I manage Windows servers, file servers, mail servers, database servers, and web servers, specializing in Microsoft Server software and technology.
So why would LinkedIn send me the profile of a server at a local restaurant when we’re looking for people with IT skills and experience? Is this just bad keyword matching? Like you have said, they really are not that useful in a job search.
The person’s profile is very clear: “Experienced server with a demonstrated history of working in the restaurant industry. Skilled in pleasing customers through great customer service, a positive attitude, proven multi-tasking abilities, and a never-quit-until-it’s-done perspective.”
Experience includes: “Talking with the guests; bringing around food, beverage and sauce samples; running the TV tower and changing channels so the guests can watch the games they like near them, celebrating birthdays, and other fun things.”
It seems LinkedIn’s A.I.-based algorithms “saw” and matched on nothing but the keyword “server!” How do they get away with this?
Nick’s Reply
If LinkedIn can’t distinguish a restaurant server from a computer network server, WTF is LinkedIn selling to corporate HR departments?
Stories about job boards and A.I. failing to deliver are so abundant that users have become numb to the marketing campaigns telling them that No, no, the technology really, really is intelligent! — even if it’s apparently doing nothing but trivial database character-string matching.
Why do you think they call it artificial intelligence? (For a stunning expose of A.I. in recruiting, read about Hilke Schellmann’s excellent book, Algorithm, here.)
Is it all just LinkedIn marketing?
We frequently discuss the backdrop of phony claims about recruiting technology, but the marketing is evolving and becoming more complex than what it’s advertising! So I’ve lost interest in what the technology is. What I’m interested in is the state-of-the-art marketing of LinkedIn recruiting snake oil!
So before we get into this, my goal with this column is to ask you, dear Readers:
3 LINKEDIN QUESTIONS
- How often do you get bad matches from LinkedIn?
- What wild new promises have you encountered about how LinkedIn A.I. technology is going to match people to jobs?
- And, what are the latest and most shocking experiences you’ve had with this paragon of A.I. — LinkedIn?
Is it all really just marketing?
Wait, wait! How stupid is this?
LinkedIn has been at the networking business since 2003. It claims to use A.I. to “connect” people and to match people and jobs. It claims to use “semantic processing algorithms” and “context” to “understand” your professional background, industry, skills, and network to suggest relevant job openings, connections, and content.
Gee, they’ve been at it 20 years and LinkedIn’s technology still cannot tell the difference between a waitperson that serves diners and an IT person that manages servers — even when LinkedIn turns on the “understanding” feature of its A.I. How stupid is this?
New LinkedIn A.I. or old, old database technology?
Please stop and think about it. Your example is a very, very simple case of a humiliating matching error. No A.I. is required for such errors. If it were, it would “understand” that, in context, you and the waiter are no match at all. (That’s why I printed all the “context” details of the person you received as a match. The context is clear!) To me, this error reveals LinkedIn is merely matching character strings — old, old database technology.
Please take no offense, but the mistake LinkedIn made with you is nothing compared to the shocking numbers and kinds of mistakes LinkedIn makes while collecting billions of dollars from HR departments every day.
That’s why recruiters that stare at LinkedIn all day contact you about so many jobs that are so laughably wrong for you.
This is after LinkedIn has been working at it for 20 years.
The price of artificial recruiting
The cost of a standard LinkedIn Recruiter “seat” for a single recruiter that hopes to find the right candidates is approximately $12,960 per year. A typical larger organization with extensive hiring demands could have dozens or even hundreds of recruiters. Do the math. A big company with 200 recruiters sends LinkedIn over $2.5 million every year to find waiters and waitresses to work in their computer server rooms.
LinkedIn’s estimated revenue last year from Premium subscriptions including recruiter seats was $6.44 billion. Is that the price to find servers or servers?
Vulse reports that “The [LinkedIn Premium] platform’s search now includes semantic matching, meaning it understands the context and intent behind queries, not just the keywords. Semantic search is an advanced search technology that enables search engines to understand the meaning behind words and phrases. Instead of simply matching keywords, semantic search returns content that aligns with the overall intent of the query, leading to more accurate and relevant results.” [All emphases added.]
What would your servers like (to swallow) for lunch?
You asked a question that haunts hiring managers and job seekers every time they dream about matching jobs and applicants: Why all the errors? Why the simplest, most obvious, most embarrassing errors?
How much can everybody swallow? Is somebody lying big-time about whether LinkedIn really, really works? Why did it match you with a server?
So what’s the point? If all of LinkedIn’s “A.I. technology,” “semantic processing,” “understanding context” and “understanding meaning” can’t tell the difference between a restaurant server and a computer server — what’s everybody paying for?
Somebody could make out like a bandit.
Please read the 3 QUESTIONS I posed above. I’d really like an update on your experiences with LinkedIn, especially its new, improved A.I.! I’m sure everyone would be interested! How often do you get bad matches from LinkedIn? Thanks!
: :
Hi Nick,
Thanks for raising these three questions — I’ve often thought similar things but never saw them articulated so clearly (or entertainingly). Your example is on point but also profoundly frustrating for anyone who expects even a basic level of contextual understanding from a platform as massive and resourced as LinkedIn. Allow me to answer your three questions.
1. How often do I get bad matches from LinkedIn?
Far too often. I’d say 70-80% of the job “matches” LinkedIn sends me are misaligned — and that’s being generous. Some are off by industry, others by seniority, location, or core skillset. It often feels like LinkedIn is grabbing any keyword overlap and ignoring all nuance. I’ve been matched to entry-level helpdesk roles despite 15+ years of experience managing enterprise infrastructure. I’ve also been suggested roles in restaurant chains, presumably because of the word “server.”
2. What wild new promises have I seen about LinkedIn’s matching AI?
LinkedIn constantly peddles the “semantic matching” and “context-aware AI” narrative, primarily through paid ads and LinkedIn Recruiter’s product descriptions. I’ve seen claims like:
• “Smarter job alerts that learn your preferences over time.”
• “AI-powered insights to suggest the best candidate in your pipeline.”
• “Advanced algorithms that understand your intent, not just your keywords.”
But these “smart” suggestions often look like a glorified Ctrl+F. If this is AI, it feels more like Artificial Ineptitude.
3. Most shocking recent experience with LinkedIn’s matching?
Recently, LinkedIn recommended I apply for a “People Greeter” role at a local department store. Apparently, “people” and “HR” were close enough in its database to warrant a match. Another gem was flagged for an event hostess role because I’d listed “employee engagement” and “culture-building” on my profile. It’s laughable — until you remember, this is a paid service sold as a recruitment solution for global enterprises, startups, small businesses, and everyone else in between.
The real problem? This isn’t just an annoyance for job seekers — it’s a business model failure being passed off as cutting-edge tech. Companies spend millions on LinkedIn Recruiter seats and “Premium AI tools,” only to be served mismatched profiles and inaccurate insights.
If this is what 20 years of “connecting professionals” look like, I’m not sure we’re advancing—we’re just repackaging the same keyword search engine with more marketing spin.
Thank you for continuing to spotlight these issues. I would be interested to hear what other experiences your readers share!
Appreciate everyone reading this far!
Demina
@Demina: I read that far because it was good reading!
I am not arguing with your take on LinkedIn.
For what it is worth (and I do not recommend it), four years ago I landed my current job off of LinkedIn Easy Apply.
Gives new meaning for AI (Artificial Ignorance).
I stopped using LinkedIn several years ago. Between the recruiters contacting me under the guise of a non existent job opportunity and seeing self congratulatory posts for the dumbest things (often copying someone else’s idea in their self congratulating post) and users promoting their reinvention of the wheel as if it were something new or useful, I could not handle the dumbed down data giveaway packaged as networking. The only valid job opportunities I learned about on there were from my human contacts.
Hi Nick- I get more bad matches than good. I’m retired now so I don’t pay too much attention to LinkedIn. I do help people I know who reach out to me with questions about companies and people that I’ve worked for during my 40 year career in IT.
My favorite LinkedIn mistake occurred a few years ago. One of my job titles was “Service Delivery Manager” in IT. I received a job opportunity for a “Labor and Delivery Nurse” at a local hospital. I was thankful for a good laugh!
@Gary: LinkedIn keeps touting its ability to “understand.” Then reveals it makes the stupidest mistakes.
Yes, Linked In is crap, as are virtually all job boards. Here is my recommendation for using it:
1. Keep your profile up to date — it’s easy and not time-consuming and can’t hurt.
2. If you have the time and inclination, respond to posts from your connections, such as interesting comments or congratulating them on work anniversaries. That keeps you in the algorithmic mix, could reinforce existing relationships, and I suppose is a nice thing to do.
3. If you have the time and inclination, feel free to post stuff or write a newsletter (which I do) — it’s minimal and again, can’t hurt.
4. Don’t bother to apply for jobs or gigs posted on the site, unless there is an “easy apply” option button — it’s worth the 30 seconds it takes to fill that out, but not much more.
5. Under no circumstances pay anyone a penny who claims they can get you jobs, sales or gigs by enhancing your profile and connecting you to prospects — it doesn’t work and it’s not worth it. A marketing friend of mine said that the only people who get any value from Linked In are people selling commodities, such as books or webinar series. Nobody seeking to build a meaningful relationship as a trusted advisor or employee gets any traction worth the cost and effort.
6. In my 15 years of consulting, I have gotten about a dozen inquiries off of Linked In. A handful progressed to actual online meetings that resulted in crafting proposals. Two became clients — both were lousy and lasted only a few months. People shop for stuff on the site the same way they would Google deals on laptops. Serious employers, clients, prospects use better and more reliable means to find people.
7. The best thing about Linked In is that it is a semi-reliable source of information on people. That’s all.
@Larry: To me, LinkedIn is a good digital people directory or “phone book.” It’s good for looking people up. I see nothing functionally useful.
Nick, I remember how long you’ve been saying LI wasn’t a good way to go. I still get notices that my profile has been seen, and teaser notices from recruiters – all of which make me wonder if anyone (or anything) is actually checking the relevance of my profile components. I haven’t updated in years, partly because I didn’t know what I was going to do going forward and partly because it didn’t seem like a major priority since I know of profiles belonging to dead people. So this AI stuff just seems like more of the same, and the server example is perfect as the illustration.
Hello Nick,
Last night I received an invitation to apply for the role of Director of Pharmacy at St. Anthony Hospital. Good old AI told me that I could be one of the first 25 applicants. I guess that it determined that my experience with Walgreens Pharmacy Information Technology qualifies me to run a pharmacy department.
Love your stuff.
My LinkedIn profile lists me as “Principal Retiree at Home”.
You can imagine the job suggestions I get…
Nick, I really thought you were being sarcastic in your response to the original inquiry and post. In reality, while these AI algorithms come with biases due to human input, their design is also based on computations. So if one keyword shows up multiple times, then it’s likely going to pick up things that closely match that keyword. This is how data works. If you don’t look at all of the data first and instead jump to filtering, then your results will be biased.
I don’t know what his LinkedIn profile looks like, but if he mentioned server that many times without being more specific about the types of servers he specializes in or even tried using another industry term to reflect his professional expertise, then that’s likely why he got paired with a restaurant server. The other thing you have to consider is whether their inquiry/post was just impulse after receiving one LinkedIn profile that didn’t match.
Just want to add that since he mentioned “Windows servers, file servers, mail servers, database servers, web servers, and Microsoft Server,” those are different variables linked to the word “server.” So it’s not surprising that restaurant server was considered a match.
@Christine: I understand what you’re saying, but that’s why this is such a problem. What you’re describing is artifacts of computations on the data that’s available. He provides plenty of context about what “server” means in his credentials. Enough that a human can quickly deduce he’s an IT guy.
The pairing is still incorrect.
@Nick, yes, but some of those descriptors are vague and won’t necessarily be placed with a technical role. Corporations are spending lots of money to match with the right candidates, but what is the alternative if most of these AI algorithms function the same way?
If there was a site offering something comparable to LinkedIn, you might find that their systems are similar. You also have to remember how quickly companies implemented AI into their systems. It became popular really fast. They’ve only just begun to make adjustments within these specific algorithms to make it more intuitive. So the models are still being tested throughout.
What is the alternative? When my team was hiring software engineers, we had a radical process.
For university job fairs, we sent an HR person and a developer. They *spoke* to candidates. Anyone who seemed reasonably intelligent got an interview.
For resumes sent in cold, an HR person took a brief look at them, and if they seemed reasonable, passed them on to my team (which meant me, until I had a couple senior people who I could share hiring with). The bias was towards “give a person a chance”.
(I had joined the company on a personal recommendation – I talked with my friend’s boss for an hour and then got hired.)
By the way, one of the best hires I made was someone who completely crashed-and-burned her interview. She tried to salvage things by saying she was a quick learner. I thought it over, then sent her (and HR) an email: Go to this (free) C++ tutorial site, focus on these sections, then get in touch with HR for another interview. She did all that, got hired and was excellent.
This worked out pretty well. And it’s not like I had superior judgement or anything. Two other best hires were people I was ambivalent about, but my team told me I had to hire them.
re: a radical process
Bravo!
It has been a long time since I hired. This was my criteria:
Can they do the job?
Do they play nice with others?
I worked with no more than three candidates at a time (unless we were hiring for multiple roles).
Phone Screen | Interview | Second Interview | Offer/Background check
If none of the three resulted with a body in a seat, rinse and repeat.
All my hires were good. Some better, some worse. I did not end up with any dogs.
If I ever decide to p[pursue my doctorate, I will probably research is all the gyrations, games, and utter bulls- in the hiring process actually result in better hires.
@Christine: I think HR is so brainwashed by the HR tech companies and job boards (incl LinkedIn) that they don’t stop to consider that they are creating the problem themselves.
When they post a job on a job board that’s really just a huge database, they are asking for thousands or tens of thousands of applications. Of course, most of them are garbage simply because the boards allow it. (HR blames the applicants. But when you make it so easy to apply you’re encouraging people to game the system because you’re telling them to play the odds.)
But now, having created the problem of too many applicants drawn from a universe of literally hundreds of millions of people, HR needs more technology to sort and filter them. But that’s the wrong time to filter. The right time to filter is when HR chooses the “universe” to recruit in. LinkedIn and job boards are too big and lend themselves to massive percentages of false positives (and negatives, by the way – many good candidates are wrongly rejected – ask and I’ll give you examples).
I’ll repeat: HR must learn to narrow the pool of potential candidates by looking in the right places. Looking at ALL people on LinkedIn and trusting database matches is, as we all know, a joke! It just doesn’t work.
So, what to do?
Carefully figure out where the small population of people you might want to hire hangs out, and GO THERE. There’s an excellent example in this article:
https://www.asktheheadhunter.com/11884/recruit-competitive-advantage
In recruiting, less is better if you know where to start looking. This is what employers pay good headhunters for. (There are few really good headhunters.) HR can
buy 10,000 resumes on LinkedIn for about $100, or pay a headhunter $30,000 for 3 carefully vetted candidates. See if you can tell why employers use headhunters.
This may also help:
https://www.asktheheadhunter.com/12838/manager-diy-hire
And don’t miss Gregory’s excellent comment above! Recruiting is not a volume game.