The JavaScript Developer Job Search Report 2026

Two months ago I asked a question on LinkedIn that I did not expect anyone to answer honestly.

I run a JavaScript job board. Every week around seventy roles pass through my hands, and every week a handful of developers write to me directly. The messages follow a pattern I had started to find unbearable. A senior engineer with eight years of production experience apologising for taking my time. A developer in Eastern Europe asking whether it was normal that he had not heard back from anyone in three months. A frontend engineer who wanted to know, quietly, whether he should stop calling himself a developer.

I could see the job postings. I could see the messages. What I could not see was the scale. Was the developer writing to me an outlier having a bad month, or was he the median experience of an entire profession that had quietly stopped working the way it used to?

So I asked. Seven questions over eight weeks, posted publicly, answered anonymously. 918 votes came back.

This is not a vendor survey. Nobody paid for it, nobody had a product to sell at the end of it, and none of the questions were designed to make a conclusion look inevitable. I asked what I actually wanted to know, in the order I became curious about it, and I am publishing what came back including the parts that contradict what I expected.

The headline finding is not that AI is taking jobs. That story has been written a thousand times and it is mostly wrong in the way it gets told. The finding is stranger and more uncomfortable than that.

The developers doing everything right are the same developers who cannot find work.

Ninety-one percent of respondents are fully transparent with their employers about using AI. Eighty-seven percent review every line of AI-generated code before it ships. Fifty-two percent describe themselves as adapting rather than panicking. By any reasonable definition, this is a professional, responsible, self-aware cohort doing exactly what the industry says it should be doing.

Forty percent of them are still searching.

That gap is the story. Competence stopped being the filter, and almost nobody is talking about what replaced it.

How This Report Was Built and What It Is Not

I want to be straightforward about the limits of this data before I start drawing conclusions from it, because too much industry research is presented with a confidence it has not earned.

Seven single-question polls were posted on my LinkedIn account over roughly eight weeks in mid-2026. Vote counts per question ranged from 69 to 225, for a total of 918 responses. Every poll was anonymous. Every poll was open to anyone who saw it, which means the sample is self-selecting rather than randomised.

That self-selection matters and it cuts in a specific direction. My audience skews toward JavaScript and TypeScript developers who follow job market content, which means it over-represents people who are actively thinking about their careers and under-represents developers who are comfortably employed and not paying attention. If anything, this population should be more engaged and more prepared than the average developer, not less.

Keep that in mind when you read the numbers, because it makes several of them worse rather than better. When forty percent of a highly engaged, career-aware population is still searching, the figure for the general population is unlikely to be sunnier.

What this report does not have is causation. I can tell you that forty percent sent more than a hundred applications. I cannot tell you from this data whether those hundred applications were well targeted. I can tell you that eighty-seven percent review AI code. I cannot verify that they actually do it, only that they say they do. Self-reported data measures what people believe about themselves, which is genuinely useful and genuinely different from measuring behaviour.

I am publishing raw percentages, the exact question wording, and the vote counts, so anyone who wants to argue with my interpretation has the same material I do. The interpretation is mine. The numbers belong to everyone.

Part One. The Search Is Longer Than Anyone Admits Publicly

JS DEVELOPER JOB SEARCH REPORT 2026 · QUESTION 1 OF 7 How long did it take to find your last JS developer job?

40% Still searching

Still searching 40% 1–3 months 28% 3–6 months 17% Under 1 month 14% JS Developer Job Search Report 2026 jsgurujobs.com

The first question was the one I most wanted answered, because it was the one behind almost every message in my inbox.

How long did it take you to find your last JS developer job?

Under one month, fourteen percent. One to three months, twenty-eight percent. Three to six months, seventeen percent. Still searching, forty percent.

I stared at that fourth number for a while before I posted anything about it.

Forty percent. Not "searching but close." Not "between contracts." Still searching, as an ongoing present-tense condition, with no end date attached. That was the single largest response to any question in the entire report, and it came from an audience of working professionals who follow job market content closely enough to see a poll and vote in it.

The fourteen percent who found something in under a month deserve attention too, because their existence proves the market has not stopped. People are getting hired. Fast hires still happen. But fourteen percent is a small door, and pretending it is the normal experience is how developers end up believing something is wrong with them personally.

Here is what the shape of this distribution tells you. The market is not uniformly bad. It is bimodal. A minority moves quickly and a large plurality does not move at all, with a thinning middle in between. That is the signature of a filtered market rather than a shrinking one. Jobs exist. Access to them is unevenly distributed in a way that has very little to do with how good you are at writing code.

I wrote about this pattern before I had the data, in a piece about why so many JavaScript postings never get filled. The polls confirmed something I had only suspected. The bottleneck is not at the interview. It is much earlier, at the layer where a human being decides whether to look at you at all.

The Volume Problem

The second question was the natural follow-up.

How many jobs did you apply to before getting hired?

Under twenty, twenty-one percent. Twenty to fifty, twenty-two percent. Fifty to one hundred, eighteen percent. More than one hundred, forty percent.

Two in five developers sent over a hundred applications. Read that again with the actual labour attached to it. A hundred applications means a hundred job descriptions read, a hundred resumes tailored or not tailored, a hundred cover letters or forms, and then a hundred silences to absorb. At even fifteen minutes each that is twenty-five hours of work, and fifteen minutes is generous only if you are copying and pasting.

The twenty-one percent who got hired in under twenty applications are the interesting group here, and I want to be careful about what conclusion to draw from them. The tempting reading is that they were simply better. I do not think that is what the data shows, and I do not think that matches what I see on my board.

What separates fast hires in my experience is almost never raw technical ability. It is specificity and visibility. They knew exactly what kind of role they wanted, they applied to roles that genuinely matched rather than roles they could plausibly stretch into, and they had something a recruiter could find and point at before the first conversation. A live project with real users. A GitHub with recent activity. A profile that reads like a person with a point of view rather than a list of technologies.

The hundred-application group is not less skilled. In most cases they are running a volume strategy because the targeted strategy has not worked, or because nobody ever told them there was another way. Volume feels productive. It fills the day. It produces a number you can report to yourself at the end of the week. And it is the least efficient possible way to search in a market where three to six hundred people apply to every remote posting.

If you are in the hundred-plus group right now, the practical implication is not "try harder." It is the opposite. Cut the volume in half and put the recovered time into making fewer applications actually land, which starts with making your profile findable before you ever apply to anything.

What Actually Worked

The third question was the one I thought would produce the clearest answer. It did not, and the ambiguity turned out to be more useful than a clean result would have been.

What actually got you your last developer job?

Cold applying, twenty-eight percent. Referral from someone, twenty-five percent. Recruiter found me, eighteen percent. Still searching, twenty-nine percent.

Two things jump out.

First, "still searching" appears again, this time at twenty-nine percent. Across two separately posted polls with different vote counts, roughly a third of respondents describe themselves as being in an unresolved job search. That consistency is the strongest single signal in the whole report. It is not a fluke of one bad week or one unrepresentative audience. It is a stable feature of this population.

Second, and this genuinely surprised me, cold applying and referrals are nearly tied. Twenty-eight against twenty-five.

The standard career advice, repeated so often it has stopped being examined, is that cold applications are dead and networking is everything. This data does not support that. Cold applying is still the single most common path to a job among people who found one. It works. What it does not do is work efficiently, which is exactly what the volume data showed. Cold applying works at a hundred applications. Referrals work at three.

Both paths lead to a job. They cost wildly different amounts of effort to walk.

The eighteen percent who were found by recruiters is the number I would most want to increase if I were advising a single developer. That is the only path in this list where the search happens while you sleep. It is also the path that depends almost entirely on things you build before you need them, which is precisely why most people neglect it until the moment it would take six months to fix.

Part Two. The Fear Is Real but It Is Not Panic

Having mapped the search, I wanted to understand the emotional weather behind it. Almost every message I get carries some version of the same anxiety, and I wanted to know whether the anxiety was universal or whether I was hearing from a distressed minority.

How worried are you about AI replacing your job?

Not at all, seventeen percent. Slightly, but I am adapting, fifty-two percent. Very worried, it keeps me up at night, fourteen percent. I have already felt the impact, sixteen percent.

Eighty-two percent feel some level of AI pressure. That is the number the headlines would run with, and it is technically true and mostly useless, because it flattens four very different psychological states into one scary statistic.

The distribution is more interesting than the total. The dominant response by a wide margin is "slightly worried but adapting," at fifty-two percent. That is not panic. That is a working professional acknowledging a real change and doing something about it. It is arguably the healthiest possible response to what is actually happening in the industry.

Only fourteen percent describe losing sleep. Given the volume of apocalyptic content aimed at developers over the past two years, I expected that figure to be double. The catastrophic framing that dominates the discourse is not the dominant lived experience. Most developers are somewhere between concerned and pragmatic.

The number that deserves more attention than it will get is the sixteen percent who say they have already felt the impact. That is not a forecast or a worry. That is a report from people who have experienced a concrete consequence: a layoff, a role that changed underneath them, a hiring freeze that closed a door. One in six. That is a real cohort with real losses, and it is worth separating them from the people who are anxious about a hypothetical future.

I wrote earlier this year about the moment Sam Altman publicly said his predictions about AI job losses were wrong. This data sits comfortably alongside that. The mass replacement did not happen. Something narrower and more specific did, and sixteen percent of my respondents are living inside it while the other eighty-four percent read about it.

Part Three. The Ownership Finding Nobody Predicted

The next two questions were the ones I designed to be uncomfortable. I expected them to expose something unflattering about how developers actually behave with AI when nobody is watching. They exposed the opposite, and this is the section of the report I find most genuinely surprising.

Would you hide your AI tool usage from your employer if asked?

Yes, nobody needs to know, six percent. No, I am fully transparent, ninety-one percent. My company requires disclosure, two percent. I do not use AI tools, one percent.

Ninety-one percent.

I wrote that question expecting a much messier result. The premise seemed obvious to me at the time: AI tools are a competitive advantage, developers are under pressure to look productive, and admitting that a machine wrote a chunk of your output feels professionally risky. I assumed a meaningful minority would say they keep it quiet.

Six percent did. That is a rounding error away from nobody.

There are two readings and I think both are partly true. The generous reading is that AI use has been normalised so completely and so quickly that hiding it no longer occurs to anyone. It is a tool, like an IDE or a linter, and you do not hide your linter. The skeptical reading is that this is a public poll on a professional network and "I am fully transparent" is the socially correct answer, so some portion of that ninety-one percent is aspiration rather than fact.

I lean toward the generous reading, because the next question backs it up in a way that would be difficult to fake.

What do you actually do when AI generates bad code in your project?

Fix it myself and say nothing, six percent. Tell my team it was AI, zero percent. Delete and rewrite myself, seven percent. I always review AI code, eighty-seven percent.

Two numbers here, and the second one is the finding of this entire report.

Eighty-seven percent say they review AI output before it ships. That is a functioning professional norm, developed organically and rapidly, without any regulation or mandate forcing it. Nobody made this rule. Engineers arrived at it themselves because they understood the consequences of not doing it.

And then: zero percent. Not one respondent out of sixty-nine chose "tell my team it was AI."

Sit with that. When AI produces broken code inside their project, not a single developer in this sample points at the machine. They fix it, they rewrite it, they catch it in review. They do not blame the tool.

I have thought about this more than any other result in the report. What it describes is a profession that has quietly absorbed AI into its existing accountability structure without renegotiating any of the terms. The machine writes the code. The human owns the outcome. There is no "the AI did it" defence, because engineers apparently never considered it available.

This is the thing that should be reassuring to anyone worried about their career, and it is completely absent from the mainstream conversation about AI and software. Everyone is arguing about whether AI can write code. The engineers actually doing the work have already moved past that question and settled a much more important one, which is who is responsible when it goes wrong. They answered: we are. Every time.

That answer is why the job is not disappearing. Accountability cannot be automated. A model cannot be paged at three in the morning, cannot be held responsible in a post-mortem, cannot stand in front of a customer and explain what happened. I made this argument at length in a piece about why system design is the skill AI cannot automate, and this poll is the closest thing to empirical support for it that I have.

Part Four. The Verdict After Everything

The last question was the one I almost did not ask, because I was slightly afraid of the answer.

By this point in the series I had established that roughly a third of respondents cannot find work, that two in five sent over a hundred applications, that eighty-two percent feel AI pressure, and that one in six has already taken a concrete hit. If ever there were a population with grounds for regret, this was it.

If you could go back, would you still become a developer in 2026?

Yes, without hesitation, forty-two percent. Yes, but I would specialise differently, twenty-nine percent. No, I would choose another field, nineteen percent. Too late to wonder, nine percent.

Seventy-one percent would do it again.

Forty-two percent said "without hesitation," which is a strong phrase and they chose it anyway, in a market that has been actively unkind to a large fraction of them.

The twenty-nine percent who would choose a different specialisation are not a rejection of the profession. They are the most practically useful group in the report, because they are saying the field is right and their positioning within it was wrong. That is a solvable problem. Specialisation can be changed. It is changed all the time.

The nineteen percent who would choose another field entirely are the honest cost of the last two years, and I do not want to explain them away. Nearly one in five people who built a career in software would not do it again. That is a real number, it reflects real damage, and any report that glosses over it is doing propaganda rather than research.

But hold the whole picture together for a second. This is a population where forty percent cannot find a job, forty percent sent over a hundred applications into silence, and eighty-two percent feel the ground shifting under them. And seventy-one percent would still choose this.

That is not naivety. Naive people do not send a hundred applications. That is a group that has looked directly at the difficulty and decided the work is still worth doing.

What the Numbers Say When You Read Them Together

Individual polls are anecdotes with decimal points. The value is in the cross-references, and there are three that I keep returning to.

The competence paradox. Eighty-seven percent review AI code. Ninety-one percent are transparent with employers. Fifty-two percent are actively adapting rather than freezing. This is a diligent, honest, self-aware population. And forty percent of them cannot find work.

If the market were selecting for competence, these numbers would be incompatible. They are not incompatible, which means the market is selecting for something else. Based on what I see on my board, it is selecting for legibility. Whether a system can parse you. Whether a recruiter can categorise you in eight seconds. Whether your profile matches a template someone wrote before they met you.

Good engineers who are illegible to the filter lose to mediocre engineers who are legible to it. That is not a moral failure of the industry, it is a mechanical consequence of six hundred applications per posting. But it does mean the advice "just get better at your craft" is incomplete to the point of being misleading.

The fear-commitment gap. Eighty-two percent feel AI pressure and seventy-one percent would still choose this career. Those two numbers coexisting tell you the fear is not existential, it is situational. Developers are not afraid that software engineering is ending. They are afraid of a specific hiring market at a specific moment. Those are extremely different conditions, and the second one changes.

The ownership signal. Zero percent blame the machine. Eighty-seven percent review its output. This is the clearest evidence I have seen that the profession has already answered the question everyone else is still debating. The debate about whether AI will replace engineers is being conducted almost entirely by people who are not engineers. The engineers have quietly settled it in practice: the machine drafts, the human owns. That division is stable because it is not a technical arrangement, it is an accountability arrangement, and accountability has never been something you could hand to a tool.

What I Actually Think, Including the Part You Will Not Like

I have been sitting with this data for weeks, and I want to say the part that does not fit neatly into an encouraging conclusion.

The reassuring narrative available here is easy to write: developers are responsible, AI is not replacing them, the profession is healthy, keep going. Every number above supports some version of that, and it is the article most people would want me to publish.

I think it is half true and the missing half matters more.

The missing half is that being good is no longer sufficient, and a great deal of career advice is still written as though it is. The developers in this survey are, by their own account, doing the right things. They review the AI output. They tell their employers the truth. They adapt rather than deny. And forty percent of them are sending applications into a void.

The advice industry keeps responding to this with more craft advice. Learn another framework. Sharpen your algorithms. Build another project. That advice is not wrong exactly, but it is answering a question these people already answered. They are competent. Competence is not what is failing.

What is failing is the connection between being good and being seen. And almost nobody writes about that honestly, because "become more legible to a filtering system" is a depressing thing to tell a craftsperson and it does not sell courses.

Here is my actual position, and I hold it strongly enough to argue about it. In 2026, a JavaScript developer's career has two separate components that require separate investment. There is the engineering, which is the part everyone talks about and the part that keeps you employed once you are hired. And there is the legibility, which is the part almost nobody talks about and which determines whether you get hired at all. Most developers over-invest in the first and treat the second as either beneath them or as something that will sort itself out.

It does not sort itself out. The eighteen percent who were found by recruiters did not get lucky. They built something findable, months or years before they needed it, and then it worked while they slept.

I will also say this, because it comes up in every conversation I have about these numbers. The nineteen percent who would leave the field are not weak and they are not wrong. Some of them will leave and be happier. A career is allowed to end. What bothers me is not that some developers want out. It is that a chunk of that nineteen percent are people who concluded they were not good enough, when what actually happened is that they were invisible to a machine.

The Thing to Take Away From 918 Votes

If I compress this entire report into one sentence, it is this.

The developers who are struggling are not the ones who failed to keep up. They are the ones who kept up beautifully and were never seen doing it.

Eighty-seven percent review the AI's work before it ships. Zero percent blame the machine when it breaks. This is a profession that took a technology capable of destabilising it and, without anyone issuing an instruction, absorbed it into an existing culture of ownership. That is not a group of people about to be replaced. That is a group of people doing the exact thing that cannot be replaced, which is being accountable.

And seventy-one percent of them, after a hundred applications and months of silence and a genuinely brutal year, would choose this work again.

The market is filtered, not closed. Forty percent are still searching, but fourteen percent found something in under a month, and cold applying still put more people into jobs than any other single channel. The doors are narrow. They are not locked.

If you are in the forty percent right now, the number I want you to hold is not that one. It is the eighty-seven percent who review every line before it ships, and the zero percent who blame the machine when it breaks. That is who you are professionally, and that has not changed no matter how many applications went unanswered.

Being unhireable and being unseen look identical from the inside. They are not the same thing, and the difference is the only part of this you can actually change.

If you want more observations like this from inside the JavaScript job market, I publish weekly at jsgurujobs.com.

Full Data Tables

Question 1. How long did it take you to find your last JS developer job? (225 votes) Still searching 40%. One to three months 28%. Three to six months 17%. Under one month 14%.

Question 2. How many jobs did you apply to before getting hired? (187 votes) More than 100 applications 40%. Twenty to fifty 22%. Under twenty 21%. Fifty to one hundred 18%.

Question 3. What actually got you your last developer job? (113 votes) Still searching 29%. Cold applying 28%. Referral from someone 25%. Recruiter found me 18%.

Question 4. How worried are you about AI replacing your job? (128 votes) Slightly, but I am adapting 52%. Not at all, I feel safe 17%. I have already felt the impact 16%. Very worried, it keeps me up 14%.

Question 5. Would you hide your AI tool usage from your employer if asked? (87 votes) No, I am fully transparent 91%. Yes, nobody needs to know 6%. My company requires disclosure 2%. I do not use AI tools 1%.

Question 6. What do you actually do when AI generates bad code in your project? (69 votes) I always review AI code 87%. Delete and rewrite myself 7%. Fix it myself and say nothing 6%. Tell my team it was AI 0%.

Question 7. If you could go back, would you still become a developer in 2026? (109 votes) Yes, without hesitation 42%. Yes, but different specialty 29%. No, another field 19%. Too late to wonder 9%.

Total: 918 votes across seven anonymous LinkedIn polls, mid-2026. Self-selecting sample of JavaScript and TypeScript developers.

FAQ

What is the JavaScript Developer Job Search Report 2026 based on? It is based on seven anonymous single-question polls posted publicly on LinkedIn over roughly eight weeks in mid-2026, gathering 918 total votes from JavaScript and TypeScript developers. No vendor sponsored it and no product was being sold. The sample is self-selecting, which means it over-represents developers actively thinking about their careers rather than the general developer population.

What was the most surprising finding in the report? Zero percent of respondents said they would tell their team that bad code came from AI, while eighty-seven percent said they always review AI-generated code before it ships. Not a single developer blamed the machine. Taken together these suggest the profession has already absorbed AI into an existing culture of human accountability without anyone mandating it.

If most developers are doing the right things, why are so many still unemployed? Because the hiring filter is not selecting for competence. Ninety-one percent are transparent with employers and eighty-seven percent review AI output, yet forty percent are still searching. With hundreds of applicants per remote posting, the constraint has shifted from whether you are good to whether a system can parse and categorise you quickly, which is a separate skill from engineering.