Skip to content

Most Recruiters Never Check Whether Their Job Ads Actually Work

4 min read
Toy-style AI robot adjusting a clock and arranging job advert performance cards on a desk.

Recruitment agencies often know which job boards feel busy. Fewer know which ads actually led to useful applications, interviews and placements.

A role goes live across several boards and social channels. Applications arrive from different sources. The recruiter is busy speaking to candidates and clients, so performance tracking becomes informal. If the advert feels quiet, someone might refresh the wording. If it has been live too long, someone might re-post it. But the timing and decision are often based on instinct rather than evidence.

For a small or mid-sized agency, that creates a simple operational problem: underperforming ads can sit live for weeks without anyone noticing, while the recruiter loses time checking dashboards manually.

An example workflow: a job-ad performance and re-post timing assistant

This is an example automation workflow, not a real client case study.

Imagine a recruitment agency that posts the same role across its applicant tracking system, two job boards and a couple of social channels. Every week, a lightweight assistant pulls the application counts, source data and dates live for each open role. It compares current performance against the agency’s own historical benchmarks for similar roles.

For example, the benchmark might show that junior finance roles usually receive a certain number of relevant applications in the first seven days, while senior technical roles take longer. The assistant does not compare every role against one generic standard. It uses role type, location, seniority and previous campaign data where available.

The weekly output is a one-page recommendation sheet per open role. It might include:

  • applications received by source;
  • days live on each channel;
  • performance versus the usual benchmark for that role type;
  • a suggested action: refresh wording, reprice, close and re-list, promote, or leave as is;
  • a suggested re-posting time based on when similar roles have historically converted.

The recruiter still decides what to do. No advert is edited, closed or re-posted automatically. The assistant simply makes the review easier and more consistent.

Why this is worth doing

Recruitment is full of small timing decisions. Should this advert be refreshed today or left alone? Is the low response rate normal for this type of role? Is one job board consistently producing volume but not quality? Are we re-posting on a Friday afternoon when our own history suggests Tuesday morning works better?

Those questions can be answered, at least partly, by data the agency already has. The difficulty is that it is scattered across dashboards and exports. A recruiter may need to check several systems just to understand whether a role is performing normally.

A weekly assistant can save one to two hours of dashboard-checking across multiple boards. More importantly, it can stop slow adverts from drifting. If a role is underperforming against the agency’s own benchmark, it appears in the recommendation sheet before it becomes an urgent client problem.

What the assistant should and should not do

This is a low-risk use case when kept as a recommendation tool. It should not automatically change spend, re-post adverts, alter job descriptions or close campaigns. Those decisions still belong with the recruiter, who understands the client relationship and market conditions.

The benchmark data also needs care. Old performance data can become misleading if the market changes, if the agency moves into a new niche, or if job board behaviour shifts. A benchmark should be refreshed periodically and treated as a guide, not a law.

The wording recommendations should be practical rather than magical. The assistant might point out that an advert is missing salary information, has a vague location, or is much longer than similar adverts that performed better. It should not invent claims about the employer or role.

What you would need

The basic inputs are application counts, source data, dates live and past performance for similar roles. Some agencies can pull this from an ATS or job board API. Others may start with a weekly export into a spreadsheet.

The assistant then needs a simple benchmark table: role type, location, seniority, expected response range and typical conversion pattern. An AI model can help draft the short recommendation text, but the underlying comparison should be clear enough for a recruiter to challenge.

For example, a recommendation might say: “This advert is 40% below the usual seven-day application range for similar roles. Consider refreshing the opening paragraph and re-posting on Tuesday morning, when this role type has historically generated stronger initial response.” That is useful because it is specific and reviewable.

A practical first step

Pick one role category you place regularly and review the last ten similar adverts. Note the posting day, source, application volume and whether the campaign produced a placement. Even a basic benchmark is better than pure guesswork.

If your agency is spending time and money on job adverts without a clear view of what works, this is a good candidate for a small AI-assisted workflow. I help professional firms and SMEs turn scattered operational data into practical decision support. If you would like to explore your own version, book a short call.

Want help putting AI to work?

If this article sparked an idea, let's have a short call about what AI could do in your business — no jargon, no obligation.