The ATS-Friendly Fresher Resume: How Screening Really Works

You have probably heard that "75 per cent of resumes are rejected by robots before a human sees them", and that you must stuff your resume with keywords to get through. Both claims are repeated everywhere, and both are much shakier than they sound. The figure is often traced to a resume-writing company's marketing rather than to published research, and recruiters describe their applicant tracking systems mostly as databases they search, not judges that reject on their own.

That does not mean the system is irrelevant. Real things go wrong: a resume the software cannot read, a "no" to a required question, a skill described in words the recruiter does not search for. This article separates those real risks from the myths and gives you a resume that handles both a parser and a person.

• What an applicant tracking system actually does, and what the real risks are.
• Layout rules that keep your resume readable by software.
• The section order and content that suit a fresher.
• How to run a five-minute parse test, and how to use AI as a second reader.

An applicant tracking system (ATS) is software companies use to manage hiring: job posts, applications, interview stages and communication. Common products include Greenhouse, Lever, Workday, SmartRecruiters and many others. When you apply through a careers page, your resume usually goes into one of these.

Broadly, it does four things with your application:

1. Parses your resume: extracts text and tries to fill fields such as name, email, education and employers.
2. Asks screening questions on the application form: work authorisation, location, graduation year, notice period. Some are "knockout" questions: a particular answer can disqualify an application automatically.
3. Stores and routes applications to recruiters and hiring managers, who review them, often grading candidates against criteria.
4. Lets recruiters search and filter all candidates, including people who applied for other roles earlier.

Some systems also offer matching or ranking features, sometimes using AI, which an employer can switch on. You cannot see whether a particular company uses them or how. So the safe approach is not to "beat the robot" but to make your resume easy for any software to read and easy for a person to judge.

• "A secret score rejects most resumes." There is no reliable evidence for a universal hidden cut-off, though practices vary and you cannot see inside each company.
• "Hide keywords in white text." It is easy to spot when a recruiter copies the text, and looks dishonest.
• "Copy the whole job description into your resume." It reads as spam to the person who opens it.
• "Only Word documents work." Text-based PDFs parse well in most modern systems. Follow the employer's instructions if they ask for a format.

Graphic templates from design tools often look attractive and parse badly. Plain, well-tested options include the widely used "Jake's Resume" LaTeX template (editable online in Overleaf), open-source builders such as Reactive Resume, and a clean single-column document in Google Docs or Word.

One page. At entry level, one page is not a restriction; it shows you can prioritise. Everything an employer needs from a fresher fits.

Firstname LastnameCity, India · phone · email · linkedin.com/in/yourname · github.com/yourname · portfolio (optional)

Links should be clickable in the PDF and should work. Use the same name everywhere: resume, LinkedIn, GitHub.

Useful if it adds something the rest of the page does not, such as your target role or a career change: "Final-year ECE student building backend services in Java; deployed a payment-retry service with idempotency (link)." Skip generic summaries ("hard-working, passionate learner").

Languages: Java, Python, SQLBackend: Spring Boot, REST APIs, JPA/HibernateData and tools: PostgreSQL, Redis, Docker, Git, GitHub Actions

List only what you could answer questions about. Order by relevance to the roles you apply for. Ten to fifteen items is plenty.

Call the section "Projects", not "Academic Projects" or "Mini Projects"; the label lowers expectations before anyone reads a word. Two or three projects, each like this:

Payment Retry Service | Java, Spring Boot, PostgreSQL | live demo · code- Prevented duplicate charges in all 12 simulated failure scenarios by enforcing  idempotency keys with a database unique constraint.- Designed a payment state machine (created, processing, succeeded, failed) and  safe retries with exponential back-off for a flaky provider.- Cut median status-check latency from 404 ms to 15 ms (200 requests) by adding  a composite index.

Every project needs at least one link a stranger can open. A project without evidence is the weakest line on a resume.

Degree, college, university (use the official names), years, and CGPA or percentage exactly as on your marksheet. Background checks compare these (article 41). Add relevant coursework only if your projects do not already show it.

Internships, freelance work, teaching assistant roles, open-source contributions (article 9), hackathon results and competitive programming ratings, each with a link where possible. If you have a gap after graduation, a line describing what you did in it (article 2) belongs here.

1. Copy test. Open your PDF, select all, copy, and paste into a plain text editor. Does the text come out in the right order, with nothing missing or merged? If columns interleave or your name and email disappear, fix the layout.
2. Search test. In the pasted text, search for three technologies a typical job description in your target role would mention. Are they there, spelled the same way?
3. Link test. Click every link in the PDF.
4. AI second reader. Paste the plain text (contact details removed) into an assistant:

Below is the plain text extracted from my resume PDF.1. Does anything look garbled, out of order or merged, as if extraction failed?   Quote the exact lines.2. List the sections and the fields a parser would likely extract   (name, education, each project, skills).3. Read it as a recruiter skimming for 30 seconds: what is the first thing you   notice, and what is the first thing that makes you doubt?Do not give me a percentage "ATS score"; you are not an ATS.[paste extracted text]

Asking the assistant not to give a score matters. A number like "ATS score: 78%" looks precise and means nothing, because the assistant is not the employer's system and does not know its settings.

Some assistants accept a PDF. That is a useful extra check: if the assistant's reading of your PDF jumbles sections, an employer's parser may too. But the copy test above is the more direct check.

Here are my resume bullets. For each one:1. Does it say what I achieved, how it was measured and what I did?2. Is it specific to me, or could any student have written it?3. Does it start with a strong past-tense verb, and is it under 25 words?4. Suggest a tighter version using ONLY the facts already in the bullet.   If a number or detail is missing, ask me for it instead of adding one.[paste bullets]

Then apply the Follow-Up Test to every rewritten bullet. If the assistant made a bullet sound bigger than what you did, change it back.

Before your application sprint, send the PDF to someone who does not know your work and ask: "In 30 seconds, what does this person build, and what is the most impressive thing here?" If they cannot answer, your proof is buried. Move your strongest project to the top of the Projects section and make its link impossible to miss.

• Two-column or graphic templates that look good and parse badly.
• Contact details only in the header, which some systems drop.
• A skills list longer than your projects section.
• CGPA or college name that does not match your documents.
• Chasing an "ATS score" from a website or an assistant instead of fixing real problems.
• Two pages for a fresher, usually because of a long skills list, coursework and certificates.

ProjectsLibrary Booking System | Java, Spring Boot, MySQL | live demo · code- Built REST APIs for search, issue and return of books with role-based access  for students and librarians.- Prevented double issue of the same copy with a transactional check;  covered by [n] tests for concurrent requests.Real-time Chat | Node.js, Socket.IO, Redis | live demo · code- Delivered messages between [n] concurrent test clients with median latency  of [x] ms on a local test.- Stored message history in Redis with a 7-day expiry to keep memory bounded.

• Next: Tailor Without Lying: Matching Your Resume to Job Descriptions with AI adapts this resume to each role without inventing anything.
• For more resume examples and section advice, see Writing a Resume That Survives the First 10 Seconds.
• For using AI on your resume more broadly, see Improve Your Resume and Profile with AI, Honestly.