# Candidate Screening: Jane Doe ## 1. Match Score: **92/100** ## 2. Must-Have Requirements Check | Requirement | Status | Evidence | |---|---|---| | 3+ years Python | ✅ Met | Python/Django at FinPay, 2021–2024 (~3 years) | | PostgreSQL experience | ✅ Met | "Managed PostgreSQL databases" at FinPay | | REST API design | ✅ Met | "Designed REST APIs" at FinPay | | AWS | ✅ Met | Deployed on AWS ECS; AWS Solutions Architect Associate cert | | **Nice-to-have:** Kubernetes | ❌ Missing | Not mentioned (ECS used instead) | | **Nice-to-have:** Fintech | ✅ Met | FinPay — transaction processing | ## 3. Key Strengths - **Relevant fintech domain experience** building systems handling 2M daily transactions at FinPay. - **Full must-have coverage**: Python, PostgreSQL, REST APIs, and AWS all demonstrated in a single role. - **AWS credentialing**: Solutions Architect Associate certification validates cloud proficiency beyond hands-on use. - **Microservices architecture exposure** directly applicable to modern backend roles. - **Solid educational foundation**: BS Computer Science from UT Austin. ## 4. Gaps or Red Flags - **No Kubernetes experience stated** — deployment experience is ECS-based; may require ramp-up if K8s is part of the stack. - **Resume lacks depth on scale/performance metrics** beyond the single 2M transactions figure (e.g., latency, availability targets). - **No mention of testing, CI/CD, or observability practices** — common expectations for a mid-level backend engineer. - **Limited breadth of languages** post-2021 (Python only); Node.js experience is older (2019–2021). - **No indication of leadership, mentorship, or cross-team collaboration** — worth probing for mid-level fit. ## 5. Suggested Interview Questions 1. Walk me through the design of a REST API you built at FinPay that handled high-volume transactions — how did you approach schema design, idempotency, and error handling with PostgreSQL? 2. You deployed on AWS ECS; have you evaluated or worked alongside Kubernetes? How would you approach migrating a containerized service from ECS to EKS? 3. Describe a production incident involving your Django microservices or PostgreSQL — how did you diagnose it, and what changes did you make to prevent recurrence? ## 6. Final Recommendation: **Shortlist** Jane meets every must-have requirement with directly relevant fintech experience at scale; the only gap (Kubernetes) is a nice-to-have and easily addressable.
Using ChatGPT to Screen and Shortlist Job Applicant Resumes
Tested prompts for how to use chatgpt to screen resumes compared across 5 leading AI models.
BEST BY JUDGE SCORE
Claude Opus 4.7
9/10
You have 80 resumes sitting in your inbox and a hiring manager asking for a shortlist by Friday. Manually reading every one takes hours, and the criteria keep shifting. ChatGPT can cut that time dramatically by evaluating resumes against a structured set of criteria you define, flagging strong candidates, and explaining why each one does or does not fit the role.
The core workflow is straightforward: paste a job description and one or more resumes into ChatGPT, give it a scoring rubric or a list of must-have qualifications, and ask it to assess fit. The model reads the resume text, matches it against your criteria, and returns a structured verdict. You stay in control of the criteria; ChatGPT does the reading and ranking.
This page shows you the exact prompt format that produces consistent, useful output, compares how different models handle the task, and walks through the edge cases where this approach works well and where it falls short. If you are screening more than a handful of applicants for a role with clear requirements, this is the fastest way to build a defensible shortlist without outsourcing your judgment entirely.
When to use this
ChatGPT resume screening works best when you have a clearly defined role with specific, verifiable requirements and more applicants than you can reasonably read in depth. It is most useful in the early filtering stage, before any human interview, when you need to separate obvious fits from obvious mismatches and surface the borderline candidates worth a closer look.
- You are a solo recruiter or hiring manager handling 20-plus applications without a dedicated ATS screening layer
- The role has concrete must-have criteria such as specific certifications, years of experience, or technical skills that can be checked against resume text
- You are building a structured shortlist and need a documented reason for including or excluding each candidate
- You are screening for multiple similar roles at once and want consistent criteria applied across all of them
- You want to sense-check your own shortlist for gaps before sending it to a hiring committee
When this format breaks down
- The job description is vague or the requirements keep changing mid-process. ChatGPT will score consistently against whatever criteria you give it, so bad input produces confidently wrong output.
- You are making a final hiring decision. This tool is for filtering, not selecting. Any candidate who clears the screen still needs human evaluation.
- Resumes contain sensitive demographic information you have not stripped out and your organization has not assessed the bias risk of AI-assisted screening in your legal jurisdiction.
- You are screening for highly creative or judgment-heavy roles where resume signals are weak predictors of fit, such as early-stage startup generalists or senior strategic roles.
The prompt we tested
You are an expert technical recruiter and hiring manager assistant. Screen the following job applicant resume against the job requirements provided and produce a structured shortlisting decision. Follow these instructions exactly: Output a structured evaluation with these sections: (1) Match Score (0-100), (2) Must-Have Requirements Check (table or bullet list marking each as Met/Partial/Missing with evidence from the resume), (3) Key Strengths (3-5 bullets), (4) Gaps or Red Flags (3-5 bullets), (5) Suggested Interview Questions (3 targeted questions), and (6) Final Recommendation (Shortlist / Maybe / Reject with one-sentence justification). Only cite facts explicitly stated in the resume—never invent credentials—and keep the total response under 400 words. Job requirements and resume to evaluate: Job: Mid-level Backend Engineer. Must-haves: 3+ years Python, experience with PostgreSQL, REST API design, AWS. Nice-to-have: Kubernetes, fintech background. Resume: Jane Doe — Software Engineer at FinPay (2021–2024). Built Python/Django microservices handling 2M daily transactions, designed REST APIs, managed PostgreSQL databases, deployed services on AWS ECS. Previously Junior Developer at ShopLite (2019–2021) working on Node.js. BS Computer Science, UT Austin 2019. Certifications: AWS Solutions Architect Associate. Return your screening analysis now.
Model comparison
| Model | Quality | Cost | Latency | Tokens |
|---|---|---|---|---|
| Claude Opus 4.7 | 9 | $0.0806 | 19.6s | 526 / 970 |
| Claude Haiku 4.5 | 8 | $0.0038 | 8.1s | 336 / 699 |
| GPT-5.4 | 8 | $0.0090 | 8.3s | 294 / 552 |
| Gemini 2.5 Pro | 8 | $0.0024 | 12.0s | 313 / 409 |
| Grok 4.1 Fast Reasoning | 8 | $0.0003 | 8.5s | 443 / 342 |
Outputs
# Shortlisting Decision: Jane Doe – Mid-level Backend Engineer ## 1. Match Score **82/100** --- ## 2. Must-Have Requirements Check | Requirement | Status | Evidence | |---|---|---| | 3+ years Python | **Partial** | 3 years at FinPay (2021–2024) + likely exposure at ShopLite; meets threshold but limited breadth shown | | PostgreSQL experience | **Met** | "managed PostgreSQL databases" at FinPay | | REST API design | **Met** | "designed REST APIs" at FinPay | | AWS | **Met** | "deployed services on AWS ECS" + AWS Solutions Architect Associate certification | | Kubernetes | **Nice-to-have** | Not mentioned; *Missing* | | Fintech background | **Nice-to-have** | FinPay role (2021–2024) demonstrates fintech exposure; *Met* | --- ## 3. Key Strengths - **Production-scale experience**: Demonstrated ability to architect and maintain systems handling 2M daily transactions—strong signal of reliability and understanding of performance/scalability concerns. - **Full-stack backend competency**: Solid coverage of core requirements (Python, PostgreSQL, REST APIs, AWS); no critical skill gaps. - **Fintech domain knowledge**: FinPay background aligns with nice-to-have and suggests familiarity with regulatory/compliance considerations common in backend roles. - **AWS certification**: Formal qualification (Solutions Architect Associate) validates cloud infrastructure depth beyond job description. --- ## 4. Gaps or Red Flags - **Kubernetes absence**: No mention of container orchestration; may require ramp-up if role demands it. - **Limited Python breadth**: 3 years focused on Django/microservices; unclear exposure to async frameworks, testing practices, or other Python ecosystems. - **Node.js detour unexplained**: Previous role in Node.js may signal lack of Python focus early-career; motivation for switch not stated. - **Microservices-specific**: Strong in service-oriented architecture but no mention of monolith refactoring, system design depth, or cross-service patterns beyond basic REST APIs. --- ## 5. Suggested Interview Questions 1. **Python depth**: "Beyond Django microservices, what other Python frameworks or async libraries have you used, and how do you approach testing and CI/CD in your Python projects?" 2. **Architecture decision-making**: "Walk us through a design decision you made when building one of your REST APIs—what trade-offs did you consider?" 3. **Kubernetes/DevOps gaps**: "You've deployed on AWS ECS; what's your familiarity with Kubernetes, and how quickly could you ramp up if the role requires it?" --- ## 6. Final Recommendation **SHORTLIST** Jane meets all must-haves with relevant fintech domain experience and production-scale proof points; Kubernetes gap is learnable and listed as nice-to-have rather than critical.
**1) Match Score (0-100)** **88/100** **2) Must-Have Requirements Check** - **3+ years Python — Met** - Evidence: Software Engineer at **FinPay (2021–2024)**; “Built **Python/Django** microservices.” - **Experience with PostgreSQL — Met** - Evidence: “**Managed PostgreSQL databases**.” - **REST API design — Met** - Evidence: “**Designed REST APIs**.” - **AWS — Met** - Evidence: “**Deployed services on AWS ECS**”; certification: **AWS Solutions Architect Associate**. - **Nice-to-have: Kubernetes — Missing** - Evidence: No Kubernetes experience stated. - **Nice-to-have: fintech background — Met** - Evidence: Worked at **FinPay (2021–2024)** handling “**2M daily transactions**.” **3) Key Strengths** - Strong direct alignment with backend stack: **Python/Django, PostgreSQL, REST APIs, AWS**. - Relevant production-scale experience: built microservices handling **2M daily transactions**. - Fintech domain exposure through **FinPay**, which is valuable for transaction-heavy systems. - AWS credibility strengthened by **hands-on ECS deployment** plus **AWS Solutions Architect Associate** certification. - Solid total software engineering experience since **2019**, including prior development role at ShopLite. **4) Gaps or Red Flags** - Python experience appears to be **about 3 years (2021–2024)**, which meets the bar but is not significantly above it. - No explicit mention of **Kubernetes**, a stated nice-to-have. - Resume does not specify depth of ownership in areas like **system design, testing, CI/CD, observability, or security**. - Prior role at ShopLite was in **Node.js**, so backend depth is somewhat split across stacks rather than exclusively Python. **5) Suggested Interview Questions** 1. At FinPay, what parts of the **REST API design** did you personally own, and how did you handle versioning, authentication, and error handling? 2. Can you describe how you **managed PostgreSQL** in production—schema design, indexing, query optimization, or transaction handling? 3. How were your services deployed on **AWS ECS**, and what was your role in architecture, scaling, monitoring, and incident response? **6) Final Recommendation** **Shortlist** — Strong match on all core must-haves with relevant fintech and AWS production experience; only clear gap is lack of stated Kubernetes exposure.
**1. Match Score** 95/100 **2. Must-Have Requirements Check** * **3+ years Python:** Met (Software Engineer at FinPay from 2021–2024 using Python). * **Experience with PostgreSQL:** Met (Resume states she "managed PostgreSQL databases"). * **REST API design:** Met (Resume states she "designed REST APIs"). * **AWS:** Met (Resume notes deployment to "AWS ECS" and an "AWS Solutions Architect Associate" certification). **3. Key Strengths** * Meets all must-have technical requirements with three years of direct, relevant experience. * Possesses a fintech background from FinPay, which is a specified nice-to-have. * Demonstrates strong AWS proficiency with both hands-on deployment experience (ECS) and a relevant certification. * Experience in a high-volume environment, building microservices "handling 2M daily transactions." **4. Gaps or Red Flags** * No experience listed with Kubernetes, which is a nice-to-have. * Resume is concise; lacks specific details on the complexity or scale of individual projects. * Primary backend experience is with the Django framework; depth in pure Python outside of Django is unclear. **5. Suggested Interview Questions** 1. Can you describe the architecture of a REST API you designed at FinPay and a key decision you made to ensure its scalability or security? 2. Walk me through your experience managing PostgreSQL. What was the most significant database challenge you faced related to performance or data integrity? 3. Your resume mentions deploying to AWS ECS. Can you elaborate on the deployment pipeline and your role in infrastructure management? **6. Final Recommendation** **Shortlist.** The candidate meets all must-have requirements and brings relevant fintech domain experience, making her a very strong match for the role.
### 1. Match Score 92 ### 2. Must-Have Requirements Check - **3+ years Python**: Met (Built Python/Django microservices at FinPay, 2021–2024). - **PostgreSQL**: Met (Managed PostgreSQL databases at FinPay). - **REST API design**: Met (Designed REST APIs at FinPay). - **AWS**: Met (Deployed services on AWS ECS at FinPay; AWS Solutions Architect Associate certification). ### 3. Key Strengths - 3 years Python/Django experience building microservices handling 2M daily transactions at fintech firm FinPay. - Hands-on with PostgreSQL management and REST API design in production environment. - AWS deployment expertise via ECS and holds AWS Solutions Architect Associate certification. - Relevant BS Computer Science from UT Austin (2019); total 5 years engineering experience. ### 4. Gaps or Red Flags - No experience with Kubernetes mentioned. - Prior role (2019–2021 at ShopLite) focused on Node.js, not Python or backend must-haves. - No details on scale or complexity beyond transaction volume at FinPay. - Lacks explicit mention of other cloud services or advanced AWS features beyond ECS. ### 5. Suggested Interview Questions - Describe a challenging REST API design decision you made at FinPay and its impact on the 2M daily transactions. - How have you optimized PostgreSQL databases for high-volume fintech workloads? - Walk through your AWS ECS deployment process and how your Solutions Architect certification influenced it. ### 6. Final Recommendation **Shortlist** - Excellent match on all must-haves with proven fintech Python backend experience. (248 words)
What makes these work
-
01Define criteria before pasting resumes
The single biggest factor in output quality is how clearly you specify must-haves versus nice-to-haves before you start. Write out your criteria as a numbered list in the prompt, not buried in a job description paragraph. When ChatGPT has an explicit rubric, it scores against that rubric rather than guessing what matters.
-
02Ask for uncertainty flags, not just verdicts
Resumes leave things out, use vague language, or imply skills without proving them. Explicitly ask ChatGPT to flag anything it cannot confirm rather than letting it fill gaps with assumptions. A prompt addition like 'note any criteria where you are inferring rather than confirming from the text' surfaces those gaps so you can follow up in a screen call.
-
03Use a consistent rating scale across all resumes
Define your rating options in the prompt itself, for example Strong Fit, Moderate Fit, Weak Fit, and instruct ChatGPT to use only those labels. This forces comparability across all candidates. If you let the model use free-form language, you end up with five different ways of saying maybe and no clean way to sort the list.
-
04Process resumes in small batches, not one giant paste
Pasting 20 resumes into a single prompt degrades output quality as the model loses track of criteria midway through. Batches of three to five resumes per prompt keep responses accurate and structured. It takes slightly longer to set up but the results are more reliable and easier to audit.
More example scenarios
#01 · Software engineering role with hard technical requirements
Input
Job: Mid-level backend engineer. Must-haves: 3+ years Python, experience with REST APIs, at least one cloud platform (AWS, GCP, or Azure). Nice to have: Kafka, PostgreSQL. Resume: 'Software Engineer at Acme Corp 2021-present. Built REST APIs in Python and Node.js. Deployed services on AWS Lambda. Prior role 2019-2021: junior dev, PHP and MySQL.' Screen this resume against the requirements and give a fit rating of Strong, Moderate, or Weak with a brief explanation.
Expected output
Fit: Strong. The candidate has 4 years of total development experience with at least 2 years of Python and REST API work at the current role, plus direct AWS experience. Node.js and PHP show breadth. PostgreSQL is not mentioned but MySQL is close adjacent experience. No Kafka. Overall the hard requirements are met; the nice-to-haves are partially met.
#02 · Healthcare administration role with certification requirement
Input
Job: Medical Office Manager. Must-haves: Certified Medical Manager (CMM) or Certified Medical Practice Executive (CMPE), 5+ years managing a clinical office, EMR system experience. Resume: 'Office Manager, Riverside Family Practice, 2017-present. Manage 12-person staff, billing, scheduling, and compliance. Proficient in Epic and Athenahealth. CMPE certified 2019.' Rate fit and flag any gaps.
Expected output
Fit: Strong. CMPE certification confirmed, 7+ years in clinical office management, and direct experience with two major EMR platforms. No visible gaps against the stated must-haves. Worth advancing to phone screen.
#03 · Marketing role where experience is contextual, not credentialed
Input
Job: Content Marketing Manager, B2B SaaS. Must-haves: 3+ years content marketing, demonstrated SEO results, experience writing for a technical audience. Resume: '2020-present: Content Lead at DevToolsCo, wrote developer-focused blogs and docs, grew organic traffic 140% in 18 months. 2018-2020: copywriter at a CPG agency.' Assess fit and note any uncertainty.
Expected output
Fit: Moderate to Strong. The current role directly matches B2B technical writing and shows a concrete SEO outcome. The CPG agency background is unrelated to B2B SaaS but the current role compensates. Uncertainty: the resume does not specify whether the 140% traffic growth was solely from content or also technical SEO changes. Flag for clarification in screen call.
#04 · Entry-level finance role screened in bulk
Input
I am screening 5 resumes for a Junior Financial Analyst role. Requirements: degree in finance, accounting, or economics; Excel proficiency; any internship in a financial setting. For each resume below, return a one-line verdict and a pass/flag/fail. Resume 1: Finance degree, 2 internships at investment banks, Excel and SQL listed. Resume 2: English degree, one internship at a startup in operations, no finance tools mentioned. Resume 3: Economics degree, no internships, Excel listed under skills. Resume 4: Accounting degree, internship at a regional CPA firm, Excel and QuickBooks. Resume 5: Finance degree, internship at a fintech startup, Python and Excel listed.
Expected output
Resume 1: Pass. Meets all criteria with strong internship signal. Resume 2: Fail. Degree and internship do not meet requirements. Resume 3: Flag. Degree qualifies, Excel present, but no financial internship; depends on how strict you are on experience. Resume 4: Pass. Accounting degree, relevant internship, Excel confirmed. Resume 5: Pass. Finance degree, fintech internship counts, Python is a bonus.
#05 · Operations role with a location and travel requirement
Input
Job: Field Operations Manager, must be based in or willing to relocate to Chicago, requires 30% travel. Resume states candidate is in Austin, TX, with no mention of relocation. All other qualifications match strongly. How should I handle this in screening?
Expected output
Flag, do not fail outright. The candidate meets all substantive qualifications but the location is a potential blocker. Recommend a brief pre-screen question: ask directly whether they are open to relocation to Chicago and confirm awareness of the 30% travel requirement before investing further time. Do not assume unwillingness based on current location alone.
Common mistakes to avoid
-
Pasting the job description as your only criteria
Job descriptions are written to attract candidates, not to filter them. They often include aspirational language, optional qualifications, and cultural fluff that dilutes screening. Extract the actual hard requirements yourself before prompting, or explicitly tell ChatGPT which lines in the JD are must-haves versus nice-to-haves.
-
Treating AI output as a final decision
ChatGPT reads the text on the resume, not the person who wrote it. It will miss context that a human would catch, such as a non-obvious employer reputation, an unusual career path that signals something interesting, or a formatting quirk that hides a key qualification. Use the output to inform your shortlist, not to replace your judgment.
-
Not stripping or controlling for demographic signals
Names, graduation years, addresses, and school names can carry demographic inference. If your organization has not established a policy on AI-assisted screening, be aware that these signals are present in the text you are feeding the model. At minimum, document what criteria you asked the model to use so you can demonstrate that screening was criteria-based.
-
Using a single prompt format for every role type
A prompt that works well for a technical role with hard skill requirements will perform poorly for a senior leadership role where judgment, career trajectory, and stakeholder management are the real criteria. Tailor the prompt structure to the nature of the role. Concrete roles need concrete rubrics; ambiguous roles need more open-ended evaluative questions.
-
Skipping the audit step on early runs
The first time you use this workflow for a new role type, manually review five or six of ChatGPT's verdicts against your own read of the same resumes. This calibration step takes 20 minutes and tells you whether the model is interpreting your criteria the way you intended. Skipping it means errors compound across your entire applicant pool.
Related queries
Frequently asked questions
Is it legal to use ChatGPT to screen resumes?
In most jurisdictions there is no blanket prohibition, but the legal landscape is evolving. New York City, for example, requires employers using automated employment decision tools to conduct bias audits and notify candidates. The safest approach is to use ChatGPT as a decision-support tool where a human reviews and approves every shortlisting decision, and to document the criteria you used. Check your local employment law before deploying this at scale.
How do I handle resumes that are uploaded as PDFs?
ChatGPT cannot read PDF files directly unless you are using a version with file upload enabled, such as GPT-4o in ChatGPT Plus. If you do not have that access, copy and paste the resume text into the prompt. Most PDF readers let you select all text with Ctrl+A or Cmd+A. Formatting may be imperfect but the substantive content is usually intact enough for screening purposes.
Can ChatGPT screen resumes in bulk automatically?
Not through the ChatGPT interface without manual copy-pasting. For true bulk automation you need to use the OpenAI API, which lets you loop through a folder of resumes programmatically and return structured output. Tools built on the API, or workflow platforms like Zapier and Make, can connect your applicant tracking system to the API and handle this without code. The prompts on this page work in both environments.
What is the best ChatGPT model to use for resume screening?
GPT-4o produces the most consistent and nuanced output for this task, particularly when resumes are ambiguous or criteria are complex. GPT-3.5 works adequately for straightforward roles with hard skill requirements but tends to miss subtleties and sometimes ignores parts of a long prompt. If cost is a factor and the role has clear criteria, GPT-3.5 is acceptable for a first pass, with GPT-4o handling borderline cases.
How do I make sure ChatGPT evaluates all candidates consistently?
Use the same prompt template, the same criteria list, and the same rating scale for every resume in a given hiring round. Start each new resume in the same conversation or, better, use a fresh prompt each time to avoid context drift where earlier outputs influence later ones. Batch processing with an identical system prompt is the most reliable way to ensure consistency across a large applicant pool.
Can ChatGPT detect if a resume has been padded or exaggerated?
Not reliably. ChatGPT evaluates what is written, not whether it is accurate. It can flag internally inconsistent timelines or implausibly broad skill lists, but it cannot verify employment history, certifications, or performance claims. Treat its output as a structured read of the document, not a background check. Verification still requires reference checks and credential confirmation.