Five signs of a dead-end job posting

Five signs of a dead-end job posting

Of the 5 986 jobs in our bot's feed over four months, 19% are repeats: the same job posted under different names across a dozen channels. There are fewer real jobs than there are rows in the feed. Below are five signs that make a dead end visible before you apply, and separately on unpaid test tasks: when they are normal and when they are a way of getting work for free.

Of the 5 986 jobs over 4 months in the Parselona feed, 19% are repeats: the same job posted under different names across a dozen channels. There are fewer real jobs than there are rows in the feed. Below are five signs that make a dead end visible before you apply, and separately on unpaid test tasks: when they are normal and when they are a way of getting work for free.

Why this matters more than it looks

An application costs time. A good one — with the task analysed and the tone adjusted to the client — costs between fifteen minutes and half an hour.

Ten of those into the void is a lost day. And worse, it is a blow to motivation: the person concludes there is no market, when in fact he was writing to the wrong places.

Figures from my scrape, for a sense of scale. 5 986 jobs over four months, seven categories, 49 jobs a day. But:

19% are repeats. One listing gets fanned out across channels, sometimes from different accounts. In the feed it looks like ten vacancies; in reality it is one.

An explicit budget appears in 17% of cases. In the other 83% the contractor names the price, and that is normal — but it is in that mass that the people with no intention of paying at all are hiding.

Sign 1. A share of the profits instead of payment

The most common and the most recognisable.

How it sounds: "payment as a percentage of sales", "you will come in as a partner", "there is no budget right now, but the project has prospects", "we will do it together and split it later".

Why this does not work. You give up a guaranteed resource — your time — for an unguaranteed outcome in someone else's business that you have no influence over. If the project does not take off, you will not be the one blamed, and there will be no money either.

The exception where this is legitimate: you want the stake yourself, you know that business, you have seen its figures and you are ready to take part as a partner. Then it is not a job but a joint venture, and it is set up differently.

What to do: answer in one line — I work for a fixed fee, and I can give you a range for the task. Half of these clients then name a budget.

Sign 2. The task is described through tools rather than a result

How it sounds: "we need someone proficient in Midjourney, Veo, ChatGPT, n8n, Figma and Photoshop".

A list of tools instead of a description of what should come out.

Why this is a warning. A client who knows what he wants describes the result: "we need listings for three products by Friday". A client listing technologies often has not worked out the task himself — and the person who will work it out, for free, in the correspondence, is you.

A caveat: this is not a verdict. Adobe is mentioned in 28.5% of listings and Figma in 18.5%, and often that is simply a habit of phrasing. The sign works in combination with others, not on its own.

What to do: ask one question — "what should the output be and by when". Either the answer appears or it does not, and that is information either way.

Sign 3. A substantial test task demanded before any conversation

How it sounds: "send us three versions on our subject", "do a trial piece and then we will discuss it".

Before the task, the deadlines and the money have been discussed.

When a test task is normal

A small trial proportionate to the job: one example, half an hour of work, on a general subject rather than on the client's real material. That is a reasonable way to remove risk, and refusing it is foolish.

When it is a way of getting work for free

Three signs together:

The test is on the client's real task. Not an abstract example but a specific product going on sale tomorrow.

The volume matches paid work. "Three versions of listings" is the job itself.

It is being asked of many people. Sometimes visible in the phrasing "we will select the best of the applications".

What to do: offer to show something similar from your portfolio — that is exactly why it is asked for in 37.5% of listings. If they insist on the real task, say it plainly: I do that after terms are agreed, and I can show you three finished pieces of this type.

Sign 4. From nowhere, going nowhere

How it looks: an account with no history, a channel created a week ago, no site, no mentions of the company, not one past project.

Why this is a warning. Not because every new account is fraudulent — plenty of businesses genuinely write from personal ones. But because you have nothing against which to check the promises.

What to check in two minutes:

Whether the company has a site and whether it is alive.

Whether it has social media and when the last post was.

Whether what the listing says matches what is visible from outside.

Reviews or traces of past projects.

This is not paranoia but part of the work. Two minutes of checking against an hour on an unpaid test.

Sign 5. Urgency without a cause

How it sounds: "needed yesterday", "urgent, ready to start right now" — while the task is not described and no budget is named.

Why this is a warning. Real urgency usually arrives with specifics: a launch date is burning, a contractor walked out, there is a shoot tomorrow.

A client like that names both the deadline and the money, because he has no time to haggle.

Urgency without specifics usually means one of two things: the person wants to push the price down with pace, or he does not understand the scope himself.

A caveat: in my scrape 13.3% of listings carry the word "urgent", and most of them are genuine. What is worrying is not urgency itself but urgency without detail.

What to do: reply with a question about the date and the scope. Real urgency answers it instantly.

How to apply this without becoming paranoid

The main risk with a checklist like this is starting to screen out everything and never applying at all.

The rule: one sign is not grounds for refusing. Two or more is grounds for asking questions before spending time on a full application.

And the other side, worth remembering. In the scrape 23% of clients explicitly look for someone long-term and 12.9% talk about ongoing volume. A third of the market wants a permanent person, and those people write normal listings.

Dead ends are noticeable precisely against that background.

Where to spend your effort instead of filtering

The economics are simple: an ordinary application takes a minute, an application with the task analysed takes twenty.

A working ratio: eight short applications and two detailed ones a day. The short ones give you statistics and screen out the dead ends naturally — those simply do not reply. The detailed ones go where long-term work is visible.

One or two answers out of ten is a normal conversion, not a failure.

What to do about your job applications today

One action, ten minutes.

Open the five most recent listings you were going to apply to and run each through the five signs.

Usually one of the five falls away immediately, two raise questions and two are fine. That is the real state of the feed — and it explains why some applications go silent through no fault of yours.

What is ready for this on the platform

The job feed — a daily round of the chats where the work is posted. What is collected gets filtered by a model and sorted into seven categories: AI creators, video editing, design, UGC, AI agents and automation, social and marketing, operators. The card holds the substance of the task, the budget if stated, and the client's contact details.

Filtering is half the job. Then you have to sell. The listing passed the check, and now comes the part where most people fall away: write, handle the objection, get to payment. Five assistants cover it, Basic plan:

What to read so this is not guesswork. The "Sales" course, Basic plan. The lesson "4 types of people. How to sell to anyone" gives the four client types with the language for each: Drivers are sold through emotion and scale, Controllers through hard numbers and a face-to-face meeting, Supporters through care for the team, Analysts through tables and time to think. The lesson "Refusals" covers rejection: "no" is an assessment of your approach, not of you.

Ready outreach scripts and a practice partner — Freelance 2.0, prompts 48 and 49. Basic plan or $9 separately.

Three skills for the same job — advanced, files from Creator (or $9 one-off, or 1,000 XP):

And separately on the third sign — the test task. The skill Checking output for real value (advanced: files from Creator, or $9 one-off, or 1,000 XP) helps distinguish a meaningful task from work done for the appearance of work: it looks for whether the result has a specific consumer or is being requested "for assessment". A minute of checking against an hour of unpaid testing pays for itself the first time.

The AI sales assistant — for when the listing has passed the check and you have to write.

This is where the real problem starts for most people. Filtering out the dead ends is a matter of technique. What comes next is entering into correspondence with a stranger, promising him something, naming a price and not losing your nerve when he asks "why so expensive". Plenty of people simply close the tab at that point. Not because they could not do the work — because they cannot sell and do not want to spend years learning how.

The assistant closes exactly that. It does not "write texts": it leads you through the correspondence from the first message to payment, prompting each next move.

What it does at each step.

Reads the person from the listing. You paste the text and get not a summary but a breakdown: who is in front of you, what matters to him and what he is afraid of. Not "needs a designer" but "has been burnt by contractors who disappeared after two weeks" and "afraid of getting a beautiful image that is unreadable on a phone". Those fears are not written in the listing — they are read between the lines, and the person himself is usually not aware of them.

Writes an application aimed at those fears. Not "hello, I was interested in your vacancy", but opening lines that land exactly where it hurts.

Carries you on through the exchange. The client replied — you show a screenshot and get the next move. He asked the price outright — you get a form of words that avoids naming a figure blind and moves the conversation to a call. That is precisely the part that is most frightening to do alone.

Takes apart what has already happened. Upload a transcript of the call and get a portrait of the client plus a separate breakdown of your own mistakes: where you slipped into jargon, where you failed to press.

Lets you practise before the live conversation. A separate mode: the assistant plays a sceptical client and asks awkward questions — "what if the bot is rude to a customer", "why so expensive". Making a mistake in correspondence with it costs nothing.

It accepts not only text but screenshots of exchanges and transcripts — that is, it works on your real deal rather than a training example.

What this changes in substance. Selling stops being a talent you either have or do not. It becomes a procedure where at each step it is clear what to do. The whole cycle from listing to sent application is 3–5 minutes; ten messages in the time it usually takes to write one and then spend half a day failing to send it.

The ten-lesson module of the "Sales" course — where to find clients, four client types, objection handling, working with rejection. Included in the Basic plan.

Registration is free. Three days of the Basic plan switch on with one button, no card needed.

What to take on the platform for this task

What to read next

Where to find AI work — 49 jobs a day and how to see them without reading chats by hand.

A client's type from the text of a listing — what else is readable in the text besides the signs of a dead end.

The market in full: 5 986 jobs over four months — the whole scrape with the category breakdown.

A covering message without a portfolio — what to write in an application once the listing has passed the check.

AI API pricing in 2026 — what a request costs and how to work out your own costs.

A call in your child's voice — the same deception, but over the phone.

Also on AI SKILLS