Why Jobseekers Don't Trust AI Hiring Tools
A decision you cannot see is a decision you cannot argue with.
The rejection arrives four days later and says almost nothing. Thank you for your interest. We have decided to move forward with other candidates. We will keep your details on file.
You do not know whether a person read it. You do not know whether you were filtered out on something you could have fixed, like a certification you actually hold but recorded in the wrong field, or on something that was never going to change. You cannot ask, because there is nobody to ask.
That experience, repeated enough times, is what the numbers below are measuring.
The trust gap is not a perception problem
AI was supposed to make hiring fairer by removing human bias from the first pass. In one sense it did. A system does not get tired at four in the afternoon or warm to people who went to its school.
But it replaced a bias you could at least argue with by a process nobody will explain. According to Greenhouse, 46% of jobseekers say AI in hiring has actively eroded their trust in the process. Among entry-level candidates in the US and UK, 62% say the same. More than half, 55%, believe they are being evaluated by AI without ever being told.
| Reported figure | |
|---|---|
| Jobseekers who call AI hiring fair | 8% |
| Hiring managers who trust AI for faster, better decisions | 70% |
| Jobseekers who suspect AI evaluates them without disclosure | 55% |
| Jobseekers who blame AI for reduced hiring transparency | 42% |
The two numbers at the top of that table are the important ones, and they should be read together.
70% of hiring managers trust AI to make faster and better hiring decisions. Only 8% of job seekers call it fair.
Seventy against eight. The people operating the system and the people subject to it are not having the same experience of it, and the gap is not close enough to be a misunderstanding. It is worth noticing which group gets to decide whether the tool stays in use.
The law is catching up, unevenly
Regulators have started to treat automated hiring decisions as something that owes people an explanation.
California moved first and hardest. AB 1008 clarified that personal information protections apply to data held in AI systems, and SB 942 introduced disclosure requirements for AI-generated content. The direction is consistent even where the specifics differ by jurisdiction: if a system makes or materially shapes a decision about a person, that person is increasingly entitled to know it was involved.
Rules vary considerably depending on where you live and where the employer operates, and they are moving quickly. The safe assumption for an employer is that "the algorithm decided and we cannot explain it" has a limited shelf life as an answer.
What transparency actually looks like
Here is where most vendors reach for the word "explainable" and leave it there, so let me be specific about what we do and, more usefully, what we deliberately do not.
At Swipejobs, when a job is surfaced to a worker, the reasons come with it. The certification that matched the requirement. The distance from the site. The shift pattern fitting the availability you set. The pay meeting the floor you gave. Those are things you can look at, disagree with, and change. If a job you wanted did not appear, the fix is usually visible in your own profile.
What we do not show is a score. No percentage, no rating, no number telling you how good a candidate you are.
That is a deliberate choice rather than a missing feature. A single number is exactly the thing you cannot argue with. It compresses a dozen specific, checkable facts into one figure whose meaning is opaque to the person it describes, and it invites everyone to treat it as objective, which it is not. "You did not appear for this shift because your forklift certification expired in March" is useful. "You scored 61" is not information. It is a verdict wearing the costume of information.
The reasons are the transparency. The number would undo it.
What you are entitled to ask
If you are dealing with an employer or a platform and cannot tell what happened to your application, these are reasonable questions, and the reaction to being asked is itself informative:
- Was automated decision-making used in assessing my application? Increasingly you have a right to be told, depending on where you are.
- What specific criteria was I assessed against? Not the weighting, not the model. The criteria.
- Which criterion did I not meet? Frequently it turns out to be one fixable, factual thing.
- Can I correct the underlying data? Wrongly recorded or out-of-date certifications are extremely common, and are the easiest thing in this list to put right.
- Is there a human review step, and how do I request one?
An employer who can answer these is running a process they understand. One who cannot may not know what their own tooling is doing, which should worry them more than it worries you.
The test
The question worth asking about any hiring system is not whether it uses AI. Nearly all of them now do, including the ones that never reply at all.
The question is whether it can tell you why. A system that gives reasons can be corrected, argued with, and improved by the person it is deciding about. A system that gives a verdict and no reasoning is asking for trust it has not earned and cannot demonstrate. If an employer cannot explain why the answer was no, they should not be surprised when good people stop asking the question.