The advert is the employer's CV

For most of my career I've given candidates the same advice about their CVs. Put the fact that decides things at the top, keep it short, use evidence instead of adjectives, and write for the reader. I've been aware for a long time that many job adverts aren't well written, but this month, with further analysis, I realised just how few job adverts would pass that test.

Since 1 October a piece of software I built has read every job advert I might plausibly apply for. I built it with an AI coding assistant, though I'm not a developer. Each morning it reads the job boards, agency sites and careers pages I've pointed it at, throws out anything that rules me out, such as a requirement for fluent German, and scores the rest from one to five against my experience. I checked those scores against 36 roles I'd scored by hand, and it came within a point on 30. It has also found roles I'd never have come across, including a six-month senior inclusion role I hadn't seen advertised anywhere. So far it has logged 429 roles, and it has taught me more about adverts than about the job market.

What the adverts left out

Titles were the first thing I couldn't pin down. A "Head of" at a 200-person company is often what a 20,000-person employer would call a Senior Manager, and "Director" moves around just as much. Working from the title and any salary it can find, the tool has marked 44 roles as too senior for me and 121 as a stretch, and reading the adverts in full shows some of those calls are wrong in both directions. A recruiter can ask what a vague title on a CV meant. A candidate scrolling a job board can't.

Salary was the second. Looking in both the job board's salary field and the advert text, the tool found a figure for 97 of the 429 roles, leaving 332, roughly three in four, without one. Some boards show pay in a box the tool can't always see, so the true share is probably a little lower. The split by source was more striking. Recruitment agencies' own sites gave a figure for 16 of their 40 roles, while employers' own careers pages gave one for 6 of 65. Because the tool reads each kind of site differently, I had an AI assistant check a sample of 18 of those careers-page adverts against the live pages. Six had closed. None of the other 12 gave a figure, and four mentioned pay only in words, such as "competitive salaries and excellent benefits". It's a small sample, but it points to a gap in the adverts rather than in the tool. On a CV, leaving out the most important number is something I'd question. In adverts it's normal.

The third problem is the requirements the tool can't catch. On top of the 429 roles it kept, it has rejected 50 outright: 27 asked for fluent German, French or Portuguese, 9 were written entirely in another language, 12 were built around labour relations and 2 were for engineering hiring. Those are the hard requirements it can detect. What it misses are the softer ones that settle a shortlist, such as a background in a particular sector, knowledge of a market on another continent, or, in one culture role, business development as the first real requirement. In the adverts I've read in full, those tended to sit well down the page, and I only found them by reading to the end.

About one advert in twenty scored four or five. Some of that reflects how specific my search is, but most of my reading time goes on level, pay and those soft requirements, three things an advert could tell me in its first few lines.

The other reader

Any candidate can now paste an advert into a chatbot and ask whether they're a fit, and I'd assume most people use AI somewhere in the process, CVs included. I do too. When the tool finds a vacancy it goes into an Airtable, sorted into sections by how suitable it looks, and my phone pings me when something scores well.

I read those adverts myself first (the fours and fives, and sometimes a three), and if one still looks worth doing I take it into a separate AI project I use for CVs and ask for a view: apply, skip or not sure. That view often picks up things I hadn't noticed, but it can only judge what the advert says, so where the advert leaves out the level, the pay or what will settle the shortlist, it's guessing.

It then asks whether I want to go ahead, and if I say no, the role is marked as skipped. If I say yes, it drafts a CV and we work through it together, with me pushing back and asking what a recruiter would make of it. I submit every application myself, on LinkedIn or the company's site.

The same problem runs both ways. An advert made of adjectives tells me little about the employer, and a CV accepted as the AI first wrote it tells an employer little about me, so I use AI to question the fit.

Employers are seeing more unsuitable applications. In the CIPD's 2026 resourcing survey of 1,014 UK employers, 58% reported an increase in unsuitable candidates applying over the previous year, while 33% saw an increase in suitable ones. The survey doesn't say why. Candidates using AI to apply faster and more often is a likely part of it, and in a ManpowerGroup study of 80 senior HR and talent leaders in the US and UK, nearly 54% reported that "AI-assisted candidate behavior is making it harder to accurately assess true candidate capability". My suggestion is that vague adverts add to the problem, because an advert that doesn't say what decides the shortlist can't stop anyone, human or machine, from concluding they're a fit.

Writing the advert like a good CV

I've disliked adverts that are clearly just a job description copied and pasted, and have done for most of my career. A job description is a useful internal record of what a role is accountable for, its grade and its reporting line. An advert has to help the right person decide to apply and the wrong person decide not to.

That's the job a CV does, and the same rules work. The deciding facts go at the top: level, salary, location and office days, and the requirements that will settle the shortlist, soft ones included. It should be short. It should use evidence rather than adjectives, because "fast-paced", "dynamic" and "competitive" tell a candidate as little as "results-driven team player" tells a recruiter. And it should be written for the reader. An advert with twenty bullet points on what the firm wants and one line on what it offers is like a CV that is all objectives and no achievements.

As Recruitment Manager at Baker McKenzie, I rewrote our adverts along these lines and introduced a standard template, which kept them consistent while leaving room to add or remove points for each role. I cut most of the bullet points and added context about the team, the work and why the role existed, so the advert did some selling. I can't point to a number that changed as a result. I did it because I knew it was better practice. That context came from the intake meeting with the hiring manager, not the job description, and the intake is where a good advert is really written.

The wording of an advert also affects who applies, not only how many. In a well-known 2011 study, Gaucher, Friesen and Kay found that adverts for male-dominated jobs used more masculine-coded wording, and that this wording left women less interested in the job and less sure they would belong, without changing how capable they felt. I'd expect a long list of "essential" requirements, half of which aren't essential, to have a similar effect on anyone who takes the list at its word. An advert should be inclusive and welcoming as well as truthful and factual, and none of that costs more than a careful rewrite.

A baseline before the law arrives

The EU's Pay Transparency Directive requires employers to give applicants the starting pay or pay range, and member states were due to bring it into national law by 7 June 2026. Most missed that date. Spain's draft Royal Decree went public at the end of June. Ius Laboris describes it as a partial transposition, and both Ius Laboris and PwC note that it leaves out the recruitment-stage rules, so applicants have no right yet to see the pay range. UK employers aren't covered for their UK roles, though those hiring in EU countries will be.

So I'll treat my numbers as a baseline: today, roughly three roles in four show no salary my tool can find. When the recruitment rules reach the countries I'm watching, I'll run the count again.

The advice I'd give a candidate about their CV is the advice I'd now give an employer about the advert. If you hire, when did you last read one of your own adverts as a candidate would, on a phone, without knowing anyone in the team?