How can AI help recruiting?
AI helps most with the writing that surrounds hiring. A recruiter or hiring manager supplies the role, the requirements and the process, and the tool produces drafts that would otherwise take an afternoon. Each draft is still checked for accuracy, fairness and tone before it is used.
Job adverts
Drafting an advert from the hiring manager's intake notes, in plain language, with the must-have requirements separated from the nice-to-haves.
Intake preparation
Listing the questions to ask a hiring manager before a search begins, so the brief is complete.
Structured interview guides
Turning agreed competencies into consistent questions and a scoring guide that every interviewer uses.
Outreach and candidate updates
Drafting first messages, scheduling notes and status updates that are clear and polite.
Summarising your own notes
Tidying an interviewer's handwritten notes into a structured record, without adding judgements the interviewer did not make.
Is using ChatGPT for a job advert the same as using AI to screen candidates?
No, and the difference matters. Drafting an advert with a general assistant is writing help: a person reads, edits and publishes it. A tool that sources, scores, filters or ranks candidates is part of the selection decision itself, which brings data protection, discrimination and, in some places, AI-specific rules.
The UK Information Commissioner's Office made the same distinction in its November 2024 audit outcomes report on AI tools in recruitment. Its audits covered sourcing, screening and selection tools and explicitly excluded generative AI used for chatbots and drafting job adverts. The EU AI Act lists AI systems intended for recruitment or selection, including placing targeted job adverts, analysing and filtering applications and evaluating candidates, among its high-risk uses. New York City's Local Law 144 requires a bias audit within one year before an automated employment decision tool is used, a public summary of the results and notice to candidates.
This page is general information, not legal advice. If you are considering a screening or ranking tool, involve whoever handles data protection and employment law for your organisation before any candidate data is processed.
What does a reviewed AI job advert look like?
Brightwater Clinics, a fictional group of three physiotherapy practices, needs a practice coordinator. The hiring manager's notes say: manage appointment diaries for three clinicians, handle patient calls and emails, process payments, 30 hours a week across four days, experience with any booking system helpful, no degree needed.
A one-line prompt produced an advert asking for 'a rockstar coordinator to join our young, energetic team', listing a degree as essential and adding 'fast-paced environment, must thrive under pressure'. Three problems in three lines: the degree requirement contradicts the brief and narrows the pool for no reason, 'young' is an age-related phrase that can discourage or exclude older applicants, and the pressure language describes a job nobody had defined that way.
The reviewed prompt pasted the notes and said: 'Write a job advert under 250 words. Use only these requirements. Separate essential from helpful. Use plain, inclusive language. Do not describe the team's age, personality or culture beyond what is supplied.' The recruiter then checked the hours, the pay line and the application steps against the approved vacancy record before publishing. A useful final check is to ask the tool to list every requirement in the advert, then compare that list with the hiring manager's notes.
Can AI help build a fair interview process?
Yes, if the competencies come from you. Give the tool the three or four things the role genuinely needs, and ask for two questions per competency, a follow-up probe for each and a simple scoring guide describing what a weak, adequate and strong answer contains. Consistent questions asked of every candidate make comparison fairer than unstructured conversations.
Then review the questions for anything that invites personal information you should not be weighing, such as family plans, health or nationality, and for questions that only someone with a particular background could answer well. The scoring guide is a support for the interviewers' judgement, not a replacement for it.
Want to see how this connects to a role?
Explore the planned team trial ↗What should you check before using an AI screening tool?
Ask the provider to show, in writing, how the tool meets the expectations regulators have already set out. The ICO's audit report includes recommendations for recruiters, and the items below adapt several of them into questions a small team can ask. They are a starting point, not a complete compliance review.
Impact assessment
Has a data protection impact assessment been completed before processing begins, and will it be updated as the tool changes?
Roles and contract
Is the provider a controller, joint controller or processor for each use of candidate data, and is that recorded in the contract?
Instructions and minimisation
Can you set explicit instructions for which data fields are processed and for what purpose, and confirm the provider does not reuse that data for its own purposes?
Inferred characteristics
Does the tool infer characteristics such as gender or ethnicity from names or other data? The ICO found some providers doing this and says inferred information of that kind is still special category data.
Candidate information
Who tells candidates how AI will process their information, and what will they be told?
Human review
How will a person review the tool's outputs and correct errors? The ICO says recruiters should not use AI to make automated recruitment decisions where the tool is not designed for that purpose.
Can recruiters paste CVs into ChatGPT?
Not as a casual shortcut. A CV is personal data about someone who has not agreed to it being shared with an AI service, and asking a general assistant to 'rank these five CVs' turns a chat into an unreviewed selection tool. The output can reflect bias in wording or background, and nothing records why one candidate scored above another.
If your organisation wants AI help with applications, treat it as a tool decision: an approved service, a defined purpose, an impact assessment, information for candidates and human review of every outcome. Until then, keep AI to drafting from your own notes.
Do recruiters care if candidates use AI?
Many do, but the more useful question is what the assessment is meant to show. Candidates increasingly use AI to polish CVs and cover letters, so a well-written application tells you less than it once did about writing ability. Interviews and work samples that ask candidates to explain their reasoning, or to talk through how they would check a piece of work, are harder to outsource.
Be careful about trying to detect AI-written text with software. A detector's verdict is an estimate that can be wrong, and rejecting a candidate on the strength of it is itself a decision you may not be able to explain. Say clearly in your process whether and how AI may be used, and design tasks around that.
Where is AI the wrong tool in recruitment?
AI should not reject, shortlist or rank candidates unless the tool has been designed, assessed and approved for that purpose, with a person reviewing the outcome. It should not infer protected characteristics, assess 'culture fit' or summarise an interview into a verdict the interviewer did not reach.
It is also the wrong tool for the moments candidates remember: a rejection after a final interview, an offer conversation, a difficult reference. Draft if it helps, then make sure the message sounds like a person who took the time.
How can a recruiting team practise using AI well?
Run the job advert exercise above with your own fictional brief: everyone drafts with AI, then lists what they removed and why. The recruiters who catch the invented degree requirement and the age-coded phrase are the ones applying judgement, and the exercise makes that skill visible.
Tosheo.ai is being built to teach this inside a planned HR and recruiting path: short lessons, hands-on drafting tasks with real AI tools, feedback against a visible rubric and a manager view of skill progress without chat logs. It trains your existing recruiters to use AI carefully; it is not a screening tool and does not assess candidates. The product is pre-build, so this is planned scope.
Sources and scope
European Commission AI Act Service Desk, Regulation (EU) 2024/1689 Annex III, high-risk AI systems ↗