Is a machine judging you?
No. And you are entitled to a straight answer to that question from any job board or agency you deal with, including this one.
This page is short for an unusual reason: there is no model here at all. Not a small one, not a cheap one, not one rented from somebody else. Nothing you send us is passed to a third party model, because there is no third party model.
The six things we refuse to build
We do not score, rank or grade people. At all.
There is no number attached to you anywhere in this system, because the system has no way to make one. Not a match percentage, not a suitability score, not a shortlist position.
No machine decides who you get to see, either.
This is the one worth watching for, because it hides behind a true sentence. A platform can advertise NO AUTOMATED HIRING DECISIONS and mean it, while its own audit log reads: agent SCORED 412 CANDIDATES, then a named human PROMOTED 12 TO REVIEW. Every rejection has a human name against it, and the human only ever saw twelve people out of four hundred and twelve. The other four hundred were decided by the machine, at the only point that mattered. We do not do the filtering step at all, so there is no gap between what we claim and what a log would show.
No machine ever rejects an application.
Your application goes to the employer with your CV attached, whole and unread by us. We do not filter it, sort it or hold it back. Under Article 22 of the UK GDPR you have the right not to be subject to a decision based solely on automated processing where it significantly affects you, and a rejection from a job is exactly that. The simplest way to honour that right is not to build the machine.
No AI interviews, and we will not add them.
Not a video interview scored by software, not a voice analysis, not a personality inference. For a driver, a picker or a fitter this is insulting as well as unreliable, and anything that reads emotion in a work setting is heading straight at the prohibited list.
No database of people, so nothing to source you from.
Your record does not exist here until you send it, and there is no billion profile pool for anybody to search. If a service already holds a profile of you that you never created, ask it where that came from: Article 14 says you should have been told.
Your CV never trains anything.
Nothing you send us is used to build, tune or improve any model, ours or anybody else's, and nothing you send us leaves this server for a third party model. There is no third party model.
What the software here actually does
All of it is ordinary and none of it is clever, which is the point. The last column is the one that matters.
| What | How it works | What it can decide |
|---|---|---|
| The question box on this site | It matches the words you type against a written list of answers, the way a search box does. There is no model. If it does not recognise your question it says so and hands you to a person, rather than inventing something that sounds right. | Nothing |
| The pay check when an employer writes an advert | It compares the job title against a table of pay ranges written by hand and tells the employer if the money looks low for that role. A lookup table, not a prediction. The employer publishes at whatever figure they like either way. | Nothing |
| Advert screening | Written rules looking for the things we refuse to publish: a fee charged to a jobseeker, no pay figure, discriminatory wording. A person reviews anything it holds. | It can hold an advert for a human to look at. It cannot reject one on its own |
| Search and the free CV check | Ordinary text search over the adverts, and a word by word comparison between your CV and an advert you paste in. Nothing is stored and nothing is scored. | Nothing |
If you are an employer buying this from somebody else
If you use an AI screening tool from anybody, understand where the liability lands. Under the Equality Act 2010 it is the EMPLOYER who answers for indirect discrimination in a selection process, not the vendor whose software produced the ranking, and "the algorithm did it" has never been a defence to anything. Recruitment and worker management are also classed as HIGH RISK under Annex III of the EU AI Act, which brings record keeping, transparency and human oversight duties for anybody operating such a system in scope. Ask any vendor three questions: where did your training data and your profiles come from, can you produce the reasoning behind a single individual score, and who is named as the human who made the decision. A good vendor will answer all three. We are not asking you to trust us instead: we are telling you we do not sell you the thing that creates the problem.
Three claims that should stop you signing
All three are taken from real vendor pages. We are not naming anybody, and the point is not to sneer: it is that the same product's marketing page and compliance page often contradict each other, and you sign the marketing one while inheriting the duties described in the other.
1. A percentage accuracy figure for matching people to jobs
"Ranks candidates against your job requirements with 98 per cent accuracy."
Accurate against WHAT? To measure the accuracy of a hiring prediction you would have to know how the rejected candidates would have performed in the job, and nobody knows that, because they were rejected. The number cannot be right or wrong, which means it is not a measurement. Ask for the definition of the metric, the size of the sample and who audited it. A specific quantified performance claim a seller cannot substantiate is a misleading commercial practice, and consumer protection enforcement now carries direct fines rather than a court case.
2. Scoring anybody on "cultural fit"
"Analyses skillsets, industry experience and cultural fit markers." Elsewhere on the same product, individual scores for "self starter" and "customer empathy".
Cultural fit is the best documented proxy there is for age, class, race and disability, and a personality inference scored out of a hundred is exactly the kind of criterion that produces indirect discrimination at scale and in writing. If it goes wrong, the Equality Act asks the EMPLOYER to justify the criterion as a proportionate means of achieving a legitimate aim. Try justifying a machine's opinion of somebody's empathy.
3. A compliance page and a sales page that disagree
One page: "the agent recommends, it never decides," with named criteria and human oversight. The FAQ of the same product: the AI "looks beyond keywords to understand the nuances of a candidate's career trajectory" with a percentage attached.
Both were written honestly, by different people, for different readers. That is normal and it is the problem: you are sold the confident one and you inherit the duties described in the careful one. Print both, put them side by side, and ask the vendor which is the contract. Ask the same of us.
Will this always be true?
The five refusals above are settled and we are not going to quietly drop one. If we ever use software in a way that affects who gets seen or who gets a reply, this page changes first and says what changed and when, rather than being updated afterwards to match what we already did.
Ask us anything about it. Contact us · What happens to your CV · Twelve questions to ask any job board