00WHAT THE REGULATOR FOUND

A score is
not a verdict.

A person seen from behind at a desk by a window, looking at a laptop with printed sheets of paper and a cup of tea beside it — AI-generated image, not a photograph of a real person or place.

Your application may be rated by software first.

AI tools can score CVs, rank candidates and analyse interview answers. The UK data protection regulator has audited these tools and set out what employers should tell you, and when a person should look again.

ASKfor a human reviewUpdated 16 October 2026

Check a stage

01Principles

Know if AI is used
Ask how it decided
Request a second look
Rights that still apply when software scores you

02THE SHORT ANSWER

Software can rate you,
people should check it.

AI is used at every stage of hiring.
The rules focus on telling you, and letting you challenge.

SCREENING

Scores from your CV

Screening tools often score CVs using keywords set by the employer. UK government guidance warns they can inherit bias from past hiring data.

Check a stage

INTERVIEWS

Answers read by machine

Some tools transcribe and score recorded interviews. Guidance says transcription can be biased against regional and non-native speakers.

See the numbers

YOUR RIGHTS

Ask for a person

The ICO says that when decisions are made without a real person involved, you can contest the decision and ask for a human review.

What to ask

General information only, not legal advice. For advice on your own situation, speak to a qualified lawyer or a free advice service.

03A QUICK STAGE CHECK

Which tool is
rating you?

Pick a stage where an AI tool might assess you. The bar shows how many fairness and access risks UK government guidance flags for that kind of tool.

Based on the UK government's Responsible AI in Recruitment guide (2024). General information, not legal advice.

Stage check

Choose a toolAn application chatbotCV scoring and siftingGames and personality testsRecorded video interviewInterview transcription

An application chatbotCheck the answers

Guidance warns general-purpose chatbots may give wrong answers about the job. A text-only chatbot can also be hard to use for visually impaired applicants.

CV scoring and siftingBias risk

Tools trained on past hiring data can inherit its bias. Gaps in employment may count against parents, carers, disabled and neurodivergent people.

Games and personality testsValidity questions

The guide says many such tests lack scientific validity, AI-driven or not, and may be inaccessible for neurodivergent people or assistive technology users.

Recorded video interviewBias risk

Many systems use eye contact as a sign of engagement, which can disadvantage neurodivergent applicants, carers and people from cultures where eye contact is avoided.

Interview transcriptionAccent risk

Transcription tools can be biased against regional and non-native English speakers and people with a speech impediment, which affects any scoring built on the text.

Source: DSIT, Responsible AI in Recruitment (2024), annex A on key risks by tool type.

04TWO NUMBERS WORTH KNOWING

Ratings can be skewed,
and fixes are possible.

One study of bias in AI CV ranking,
one figure from the UK regulator's audits.

85%

White-associated names preferred

Researchers swapped names on over 550 real CVs and asked three language models to rank them. The models favoured white-associated names 85% of the time and Black-associated names 9%.

NAME BIAS · UNIVERSITY OF WASHINGTON

≈ 300 fixes

Recommendations to AI recruitment providers

The ICO audited AI sourcing, screening and selection tools and made almost 300 recommendations, all accepted, covering fairness, transparency and collecting less data.

AUDITS · ICO 2024

The study tested models in a research set-up, not any specific employer's system. The ICO also found some tools inferred gender or ethnicity from a name, which it said was not accurate enough to monitor bias.

05FOUR THINGS YOU CAN ASK

Questions worth
putting in writing.

The ICO says employers should tell you when automated decisions are made and explain how the tools work. These are the questions its guidance points to.

Is AI or automation used?
Employers need to tell you if automated decisions are being made about your application, and how that may affect it.
How was the decision reached?
If you think an automated decision is wrong, you can ask how it was reached and explain your point of view.
Can a person take a second look?
The ICO says you can contest the decision and request a real person to review it.
Do you need an adjustment?
Guidance says employers should plan reasonable adjustments, such as a manual route, for disabled applicants disadvantaged by a tool.

A laptop on a desk set up for a recorded video interview, with a small round light on a stand, a glass of water and an empty chair — AI-generated image, not a photograph of a real person or place.

06BEHIND THE SCORE

Who is responsible
for the rating?

Close-up of two hands holding a sheet of paper over a wooden desk, with a pen, a highlighter and a cup of tea beside it — AI-generated image, not a photograph of a real person or place.

The ICO found some AI providers wrongly treated themselves as mere processors and tried to pass all responsibility to recruiters, under vague contracts. In one case, each side's privacy policy pointed to the other, so candidates were told almost nothing.

It also found tools that collected far more personal information than needed, including profiles scraped from job networking sites and social media. The rules on automated decisions changed under the Data (Use and Access) Act, and the ICO says its guidance is being updated.

07Questions

Questions people ask

Can an employer reject me using AI alone?

The ICO says employers can use automated decision-making in hiring if they have a valid reason, but safeguards must be in place. You should be told, and you can contest the decision and ask for a real person to review it.

How do these tools rate applications?

UK government guidance says screening tools often use natural language processing to score CVs and personal statements, usually by searching for keywords based on criteria the employer sets.

Are AI hiring tools biased?

They can be. A University of Washington study found three language models favoured white-associated names 85% of the time when ranking CVs. Government guidance says tools trained on historical data can inherit past bias.

What if a tool puts me at a disadvantage because of a disability?

The government guide says employers have a duty under section 20 of the Equality Act 2010 to make reasonable adjustments, which may mean a manual review instead of the AI tool.

How do I ask for a human review?

The ICO says a request can be made verbally or in writing, and recommends following up in writing. If you are unhappy with the response, complain to the organisation first, then to the ICO.

08Sources

A score is
not a verdict.