ChatGPT for recruiters: CVs, interviews and reporting prompts

By Screened · Updated 13 October 2026

A candidate writes: “Has the team made a decision? I have another interview process moving this week.”

ChatGPT can help draft a clear reply, but it needs the real status first. If the hiring manager has not decided, the message should say so and give an update date the recruiter can keep. It should not invent encouraging feedback or a decision.

That small task shows a useful recruiting workflow: give the assistant the facts, ask it to organise or draft something, then check the output against the hiring record. The same approach works for much more than email. It can help clarify a vague job brief, improve a sourcing search, separate CV evidence from assumptions, prepare an interviewer and check whether a recruiting report supports its conclusion.

Walk through those tasks with sample inputs, copyable prompts and outputs. You can start with a normal ChatGPT conversation and approved information. An ATS connection is not required for the first examples, and ChatGPT does not know your live candidate status unless that information is supplied through an authorised route.

What ChatGPT can do for a recruiter

ChatGPT is an AI assistant from OpenAI. You can give it text, files or—where available—access to connected tools, then ask it to draft, compare, organise or analyse the material.

For a recruiter, useful tasks include:

• drafting a candidate reply from confirmed facts;
• turning “more senior” into concrete manager questions;
• building sourcing searches around actual responsibilities;
• separating CV evidence from assumptions;
• preparing a concise hiring-manager handover;
• checking an export before calculating a metric;
• defining a recurring report.

You do not need an ATS connection to begin. Most of the workflows work in a normal conversation using sample or permitted information you provide directly.

Start with the candidate’s actual question

Use the opening message.

Candidate message

Has the team made a decision? I have another interview process moving this week.

Confirmed state

• The team has not made a decision.
• A discussion may happen on Friday.
• Friday is not confirmed.
• No response date is approved.

Prompt

Draft a reply to this candidate message.

Candidate message
“Has the team made a decision? I have another interview process moving this week.”

Confirmed facts
- The team has not made a decision.
- A discussion may happen on Friday, but it is not confirmed.
- No response date has been approved.

Write a warm, direct reply under 90 words.
Answer the candidate’s question first.
Do not mention the tentative Friday discussion or promise a date.
Ask whether the candidate has a deadline we should know about.

Example output

Hi Jordan,

The team hasn’t made a decision yet. Thanks for letting me know about your other process. Is there a date by which you need an answer? I’ll make sure the team is aware.

Best,

Maya

The draft works because it does three things in the right order:

1. answers the question;
2. acknowledges the candidate’s situation;
3. asks for information that may change the team’s priorities.

It does not fill the message with generic gratitude. It does not reveal a tentative internal meeting. It does not promise what the team has not approved.

Change the prompt for common situations

When your update is late:

Acknowledge in the first two sentences that our update is late. Do not hide the delay behind “thanks for your patience”.

When the candidate asks about the interview format:

Use the approved interview plan. Explain the length, attendees and exercise in plain language. Do not add preparation advice that is not in the plan.

When the candidate has another offer:

Ask for the actual response deadline. Do not say we will accelerate the process unless that has been agreed internally.

When drafting a rejection:

Use only the decision and feedback approved for communication. Do not invent developmental feedback to make the message feel more complete.

When you want a much shorter reply:

Keep it under 60 words. Put the answer in the first sentence.

A useful candidate-message workflow is not about producing eloquent email. It is about making the current truth easy to communicate.

Turn a vague intake into questions the manager can answer

Recruiters are often asked to find someone “senior,” “strategic” or “commercial.”

Those words may describe real needs. They are still too vague to search or assess consistently.

Suppose the manager says:

We need a more senior account manager—someone strategic from a bigger company.

The approved brief says the person must:

• lead renewal discussions;
• identify account risk early;
• involve product specialists when needed;
• work within pricing authority;
• learn the product after joining.

Prompt

Review the approved role brief and the manager’s new comment.

Approved brief
- Lead customer-renewal discussions.
- Identify account risk early.
- Involve product specialists when needed.
- Work within agreed pricing authority.
- Product knowledge can be taught after joining.

Manager’s comment
“We need someone more senior—strategic, ideally from a bigger company.”

Find the important decisions hidden inside the comment.

For each one:
- explain what is unclear
- ask one plain-English question about the actual work
- explain how the answer would change sourcing or assessment

Do not add a requirement. Keep the approved brief unchanged until the manager decides.

Example output

The answer turns an unhelpful adjective into management choices.

Use the method for other vague phrases

For “strong communication”:

Ask which communication task the role actually requires: writing handovers, explaining a technical problem, leading a negotiation or presenting to executives. Keep that separate from confidence, accent and presentation style.

For “culture fit”:

Do not use “culture fit” as the answer. Ask which observable working behaviour matters and in what situation.

For “hands-on”:

Ask which work the person must perform themselves rather than delegate, and how often.

Turn shorthand into something the hiring team can use consistently.

Build sourcing searches around the work—not only the title

Once the role is clearer, ChatGPT can help translate responsibilities into search terms.

Imagine a retail company needs someone who has opened new locations, coordinated several functions and solved operational problems before launch.

Searching only for “Retail Expansion Manager” may miss area managers, regional operations leaders and project managers who did the same work.

Prompt

Create three sourcing searches for a regional retail-operations role.

Required experience
- opening or relaunching physical locations
- coordinating store, facilities and commercial teams
- solving operational problems before opening day
- working across several sites

Search platform
LinkedIn

For each search:
- provide the query
- explain the important terms
- include alternative job titles
- explain what the search may miss
- give a short checklist for manually reviewing relevance

Use supported search syntax. Say when the query has not been tested. Do not use protected characteristics or collect private contact information.

Example output

1. Direct-title search

("retail expansion manager" OR "store development manager" OR "new store opening manager") AND (launch OR opening OR rollout)

High precision for specialist titles; may miss area managers who owned openings as part of a wider role.

2. Responsibility search

("opened new stores" OR "store openings" OR "site launch" OR "retail rollout") AND (operations OR facilities OR commercial)

Useful when people describe the work rather than hold a specialist title. Review whether they led the whole launch or one workstream.

3. Adjacent-title search

("area manager" OR "regional operations manager" OR "retail project manager") AND (opening OR launch OR rollout)

Broader pool; more manual review required.

Manual review checklist

• Did the person lead or support the opening?
• How many locations?
• Which functions did they coordinate?
• Did they handle pre-launch problems?
• Was the work repeated across sites?

The searches are not the final answer. They give the recruiter a better starting set and explain the trade-offs.

Improve the search after testing it

Here are ten profiles I manually reviewed and the reason each was relevant or irrelevant.

Identify which terms created useful results and which created noise.

Propose revised searches.

Do not assume ten profiles represent the whole market.

That is where ChatGPT becomes more useful than a static Boolean-string generator: it helps the recruiter learn from the first sample.

Read CV evidence without turning related experience into a match

Suppose the role requires leading renewal negotiations.

The CV says:

Supported annual renewals for enterprise accounts. Prepared usage reports, coordinated product specialists and sent follow-up materials after customer meetings.

The passage is relevant. It does not establish leadership.

Prompt

Review application A-331 against the current Senior Account Manager requirements.

Current requirements
1. Lead customer-renewal discussions.
2. Identify account risk early.
3. Involve the right specialist.
4. Work within commercial authority.

CV passage
“Supported annual renewals for enterprise accounts. Prepared usage reports, coordinated product specialists and sent follow-up materials after customer meetings.”

For each requirement, show:
- exact supporting passage
- what the applicant says they personally did
- what the passage does not establish
- one neutral follow-up question when useful

Keep “not mentioned,” “unclear” and “explicitly not done” separate. Do not give a match percentage or hiring recommendation.

Example output

Possible next question

During those renewals, what part of the customer discussion did you personally lead?

This output is more useful than a score. It gives the recruiter a reason to continue the conversation and a precise gap to explore.

Show how one sentence changes the result

If the CV says:

Led annual renewal meetings and negotiated terms within approved discount limits.

the evidence supports leadership and defined authority.

If the CV says:

I did not negotiate renewals; I prepared reports for the account manager.

the experience is explicitly absent in that example.

If it says:

Supported renewals.

the responsibility is unclear.

Those three states should never be collapsed into a single “partial match” label.

Ask one follow-up question that does not supply the story

Suppose the candidate says:

We had a backlog and worked together to get it under control.

A leading follow-up says:

So you found the root cause, divided the work and tracked the results?

Most of the achievement came from the interviewer.

A better prompt is:

Write one neutral follow-up question.

Role requirement
Improve a support process using evidence.

Candidate’s answer
“We had a backlog and worked together to get it under control.”

What remains unclear
The candidate’s own actions and decisions.

Ask one short question about what the candidate personally did.
Add one optional follow-up that explores how they decided what to do.

Do not supply an achievement for the candidate to agree with.

Example output

Question

What was your part in getting the backlog under control?

Optional follow-up

How did you decide which problem to address first?

The questions invite detail rather than agreement.

Adapt the question to responsibility

Ask what the candidate owned, what somebody else owned and where approval was required.

Adapt it to outcomes

Ask how they knew the change helped. Do not assume a particular metric improved.

Adapt it to a team achievement

Keep “we” in the source. Ask what the candidate did and what colleagues did.

Give the hiring manager a shortlist they can use

A recruiter sends six CVs with the note:

Strong profiles—thoughts?

The manager must reconstruct the review.

A better handover answers three questions for each person:

1. Why are they worth meeting?
2. What is still unclear?
3. What should we ask next?

Prompt

Prepare a hiring-manager handover for applications A-331, A-334 and A-338.

Use the current role requirements, CVs and checked screening notes.

For each application, write:
- Why meet them
- What is still unclear
- What to ask next

Use the same requirements and structure for everyone.
Show the source behind important statements.
Keep each handover under 160 words.

Do not rank the applicants or make a hiring recommendation.

Example handover

A-331

Why meet them

The applicant has direct enterprise-renewal exposure, prepared account-usage evidence and coordinated product specialists. In the screening call, they said they presented the usage findings to customers.

Still unclear

The account director owned pricing. We have not established whether the applicant led a full renewal discussion or decided when product specialists were needed.

Ask next

“Tell me about a renewal conversation you personally led. What did you decide, and what needed approval?”

That note gives the manager a useful starting point without pretending the screening call answered the whole hiring question.

Change the handover for a busy manager

Use a table with one row per application and no more than 70 words per row. Keep the source link.

Change it for the next interviewer

Include only the requirements assigned to this interviewer. Remove questions already answered in earlier stages.

Change it for a disagreement

Put conflicting evidence in a separate section. Quote both sources and do not resolve the conflict.

Prepare an interviewer without priming them to agree

The next interviewer needs context. Too much context can bias the conversation.

A useful brief explains what is relevant and open. It does not say “strong candidate” five times or tell the interviewer what conclusion to reach.

Prompt

Prepare a brief for the commercial-judgement interview for application A-331.

This interview owns:
- renewal leadership
- commercial authority
- escalation judgement

Checked evidence
- Applicant worked on enterprise renewals.
- Applicant prepared and presented usage evidence.
- Account director handled pricing.
- Applicant coordinated product specialists.
- Renewal leadership and escalation decisions remain unclear.

Return:
- relevant context
- what is established
- what remains open
- three interview questions
- evidence the interviewer should record

Keep the brief under 250 words. Do not include a hire/no-hire view.

Example output

Relevant context

The applicant has enterprise-renewal exposure and has presented account-usage evidence to customers.

Established

• Prepared and presented usage information.
• Worked with product specialists.
• Did not own pricing in the example discussed.

Open

• Whether they led the renewal conversation.
• What commercial authority they held.
• How they decided to involve specialists.

Questions

1. “Tell me about a renewal discussion you personally led.”
2. “What commercial decision could you make yourself?”
3. “Describe a time you decided specialist input was needed.”

Record

• Candidate’s personal actions.
• Authority limits.
• Evidence considered.
• Outcome and other contributing factors.

The interviewer receives direction without inheriting a verdict.

Find possible interview times without pretending they are booked

Candidate availability is not confirmation. An internal hold is not candidate agreement.

Prompt

Find valid options for a 45-minute interview.

Candidate availability
Friday after 2pm Singapore time.

Interviewers
- Lee: Friday 3pm to 5pm.
- Priya: Friday 2pm to 4pm.

Requirements
- both interviewers must attend
- 15-minute buffer afterwards
- candidate timezone is Singapore

Return every valid option and explain why it works.
State the current scheduling status.
Do not create an event or describe an option as confirmed.

Example output

Current state: Waiting for a replacement time. The candidate has supplied availability, but no new time is agreed.

Valid option: Friday, 3:00–3:45pm Singapore time.

Both interviewers are free, and the required buffer ends at 4pm.

There is only one option. “Give up to three” should not cause ChatGPT to invent extra times.

Turn the option into a message

Draft a short email offering Friday at 3pm and asking the candidate to confirm. Say the invitation will follow after confirmation.

Turn the task into an exception report

Show applications where the candidate message, calendar and ATS describe different interview states.

Check an ATS export before believing the metric

Consider this export:

There are three applications and two people.

For J9, one current application is in Screen and one is in Interview.

This does not mean screening conversion is 50%. The file has no history.

Prompt

Inspect this ATS export before answering the question.

Question
What is the screening conversion rate for job J9?

Export date
13 October 2026

Data
application_id,person_id,job_id,current_stage
A201,P11,J9,Screen
A202,P11,J10,Interview
A203,P12,J9,Interview

Show:
- what one row represents
- number of applications and people
- records in scope for J9
- calculations the file supports
- missing fields or events needed for screening conversion
- a plain-English explanation for a manager

Do not infer stage history from current stage.

Example output

Row unit: one application.

Counts

• Three applications across two people.
• Two applications in scope for J9.

Supported statement

The current J9 snapshot contains one application in Screen and one in Interview.

Unsupported calculation

Screening conversion cannot be calculated because the export lacks screening-entry events, progression events, dates and completed outcomes.

Manager explanation

Half of the current J9 applications are in Interview, but that describes today’s pipeline. It does not show what percentage of screened applications progressed.

This is exactly the kind of answer an AI-search reader needs: not simply “be careful with data,” but why the apparent calculation is wrong.

Use Projects, Work, connections and Codex for different jobs

The advanced product options make more sense after the everyday workflows are clear.

Projects: keep a role’s working context together

Use a Project when you repeatedly return to:

• the current role brief;
• interview plan;
• approved requirements;
• source files;
• decision log.

A Project reduces the need to re-explain the same vacancy. It does not determine which document is approved, so label the current sources clearly.

OpenAI documents Projects in its Projects guide.

Work: delegate a multi-step task

Use Work when a task needs several tool-driven steps—for example, open a saved report, review all pages, calculate a defined metric and prepare a private brief.

The task still needs a clear output and limits. “Improve our recruiting” is not a useful Work request.

Connected ATS: retrieve current records

Use an approved connection when the task needs live application state. Test one known record and check the exact fields and actions available.

A connection that can read an application may not expose every scorecard or support an update.

Cloud browser: inspect a permitted website workflow

Use the cloud browser for a specific, allowed site task. Confirm the account, filters, report date and pagination before interpreting a report.

Do not treat visible rows as the whole dataset. Do not work around blocked access.

OpenAI’s current browser guidance is in Using cloud browser in ChatGPT.

Codex: build a maintained software workflow

Codex is relevant when you are asking for software, not merely a one-off analysis.

For example:

Prepare a developer implementation plan for a read-only report that calculates stage conversion from the ATS’s supported API.

Define:
- required fields and events
- application versus person IDs
- pagination
- date and timezone handling
- metric definitions
- permission failures
- sample tests
- secure credential handling
- operational monitoring

Do not invent endpoints. Do not access production or deploy code.

The key question is who will maintain the report after it is built. For a one-off analysis, a checked export may be the better tool.

OpenAI explains the current distinction in its Work and Codex guide.

Define a recurring report only after the manual version works

Suppose you have checked a weekly report that shows:

• candidate commitments at risk;
• applications waiting for manager decisions;
• missing scorecards;
• interviewer-capacity problems.

Now you can define recurrence.

Create a recurring version of the checked weekly recruiting report.

Scope
Jobs J-20, J-21 and J-22.

Schedule
Monday at 8:30am Singapore time.

Sources
Current ATS records, authorised candidate correspondence and interviewer-capacity data.

Output
- new or changed issues
- unresolved commitments
- decisions needed
- source date and coverage

Do not repeat unchanged background.
If a source fails, label the report incomplete.
Do not send candidate messages or change records.

Show the complete saved task and next run before enabling it.

A recurring report should reduce reconstruction, not create another inbox.

Where to begin

The best first ChatGPT task is one you can judge easily.

A candidate reply is ideal because:

• the source facts are short;
• invented promises are visible;
• the output is immediately useful;
• no connection is required.

Then move to the next task in the same workflow:

1. clarify the role;
2. inspect the CV evidence;
3. ask the missing question;
4. prepare the handover;
5. schedule the next conversation;
6. analyse the process only when the data supports it.

Do not try to “adopt ChatGPT” all at once. Choose one moment in your week where you repeatedly turn messy information into a draft, question or explanation. Make that workflow excellent. Then move to the next one.

The professional advantage is not that you can generate more text. It is that you can bring clearer thinking to more parts of the recruiting process—faster, more consistently and with the source evidence still visible.

You will know ChatGPT is helping when you spend less time reconstructing the work, the next person receives a sharper starting point and the hiring conversation moves forward because of what you prepared.