There were 707,000 vacancies in the UK in the three months to May 2026, the lowest level since early 2021. Fewer roles on the market means more applicants chasing each one, and most hiring teams read that as good news. It isn't. A cooler market doesn't hand you better candidates, it hands you more of everyone, and the job description is the only filter standing between your inbox and a fortnight of unpaid sifting.
So it's worth asking what a job description is actually for. Most are written to clear internal approval. HR needs the grading right, legal wants the risk covered, the hiring manager wants the responsibilities to line up with the org chart. Every one of those readers matters. None of them is the person you are trying to hire.
The Job Title Nobody Types Into a Search Bar
Look at what candidates actually search for. UK keyword data from Ahrefs puts "warehouse operative job description" at roughly 1,300 searches a month, "support worker job description" at around 1,100, and "care assistant job description" at about 800. Plain, unglamorous, literal titles.
Now look at what employers publish. Customer Success Ninja. Talent Partner (Growth). Retail Experience Associate, Level II. Those titles exist because they solve an internal problem: they signal seniority, they fit the banding, they make the place sound ambitious. They also match nothing a real person types.
This is the gap that quietly costs you applications. Your vacancy isn't losing to a competitor's better offer. It's losing earlier than that, to a search result you never appeared in, because the words on the advert and the words in the search bar were never the same words.
Research the Language Before You Write the Role
Finding those words is ordinary content research, and it's the step almost nobody takes. Three things are worth knowing before you draft: which version of the title carries the search demand, which alternative phrases describe the same work, and what candidates ask alongside it.
That second one catches people out. "Healthcare assistant", "HCA" and "care assistant" are three doors into one job, and a candidate only walks through the door they know. Publish under one label and you've quietly shut the other two.
It's worth looking at who currently ranks for the title you're about to publish. Mostly it will be job boards, and that tells you two useful things. The phrasing they lead with is the phrasing that's winning, so there's no shame in borrowing it. And you are unlikely to outrank them from your own careers page, because Reed and Indeed have spent two decades accumulating the domain authority that settles those positions. That gap is an off-site problem rather than a copywriting one, and closing it is slow work of the kind Hetneo exists for. It will not change what you publish this week.
None of this means stuffing keywords into the advert. It means finding out what the role is called by the people who do it, then calling it that.
Plain Language Is Also the Safer Language
There's a happy accident here. The wording that performs in search is largely the wording that keeps you out of trouble.
Acas guidance on how to advertise a job is blunt about the phrasing to avoid. "Recent graduate" and "highly experienced" both risk age discrimination. "Barmaid" and "handyman" imply you have already decided the gender of your hire. A "German sales rep" should be a "German-speaking sales rep", because one describes a nationality and the other describes the skill you actually need.
The principle Acas lands on is a good editorial rule as well as a legal one: use language that defines the job, not the applicant. Adverts written that way tend to be specific, concrete and free of coded shorthand, which is exactly what makes them findable. Acas also suggests advertising in at least two places, partly because a single online channel can silently narrow who ever sees the role.
Let Search Engines Read the Advert Properly
Once the words are right, make the page machine-readable. Google's job posting structured data documentation sets out four required properties: datePosted, description, hiringOrganization and title, plus a location (or a location type, if the role is remote). Get those in place and the vacancy becomes eligible for Google's dedicated job search experience, where candidates filter by title and location.
One condition catches teams out. The structured data has to match content a candidate can actually read on the page. Marking up a salary band or a location that doesn't appear in the visible advert isn't a shortcut, it's a reason to be dropped.
If your vacancies sit on an applicant tracking system, check what it emits before assuming this is handled. Plenty of platforms publish partial markup, or none.
Where AI Drafting Helps, and Where It Doesn't
AI is genuinely useful for the first pass: turning a messy internal role profile into readable prose, tightening a bloated responsibilities list, or producing three variations of an opening paragraph to react to. It's fast, and reacting to a draft is easier than facing a blank page.
What it cannot do is own the content. It doesn't know your grading structure, your salary bands, your reasonable adjustments, or which of last year's requirements were quietly abandoned. It will also produce confident, generic, faintly inhuman copy if you let it, and candidates recognise that tone immediately.
Keep a named human accountable for accuracy, inclusivity, salary detail and legal compliance. That reviewer is not a formality, they're the reason the advert is publishable.
Measure Applications You Would Actually Interview
Most recruitment reporting stops at volume: views, clicks, applications received. Volume flatters a bad advert. An advert with the wrong title and a vague description can generate plenty of applications, all from people who misread what the job was.
Better questions: what proportion of applicants met the essential criteria? How many reached first interview? Where did the shortlisted ones come from, as opposed to where the bulk arrived from? Those two sources are often different channels, and the reporting that only counts arrivals will point you at the wrong one.
If a change to the title lifts applications by 40 per cent but shortlist numbers stay flat, the change didn't work. It just made more work.
Before You Publish
A short pass, worth doing every time:
- Is the title what a candidate would type, rather than what the org chart calls it?
- Are the alternative names for this role reflected somewhere in the text?
- Would the wording survive the Acas test of defining the job, not the applicant?
- Are the essential criteria genuinely essential, or are some inherited from an old version?
- Is the salary in the advert?
- Does the structured data match what a reader can see?
- Has a named person signed off on accuracy and inclusivity?
Getting found isn't the whole of recruitment. But an advert nobody finds can't be a good advert, however carefully it was approved internally, and the fix usually costs nothing more than deciding to write it for the candidate first.