The phrase 'ATS keywords' gets thrown around like there is one magic list. There is not. Modern Applicant Tracking Systems — Greenhouse, Lever, Workday, iCIMS, Taleo — each score resumes differently, and the weight they put on different keyword types has changed meaningfully over the last two years.

This post is a 2026 update on what actually moves the needle inside an ATS, based on how these systems parse documents and what recruiters filter on.

How ATS scoring really works

Most ATS platforms do two things with your resume: parse it into structured fields (name, email, work history, education, skills) and index the full text for recruiter search. Your 'score' in the ATS is not actually a single number — it is whether a recruiter's filter matches your parsed fields and search terms.

The uncomfortable truth
There is rarely an auto-reject score. What kills applications is recruiter filters — a recruiter typing 'Python AND Kubernetes AND senior' into the search box will never see you if those exact terms are missing.

The four keyword tiers that matter

Tier 1 — Hard skills (highest weight)

Concrete technologies, programming languages, frameworks, cloud platforms. These are the terms recruiters filter on most aggressively. Mirror the job description exactly: if it says 'TypeScript', write 'TypeScript', not 'TS' or 'typed JavaScript'.

Tier 2 — Job title variants

ATS search frequently looks for title matches. If the role is 'Senior Software Engineer' and your last title was 'Senior Engineer II', consider adding the normalized title as a parenthetical: 'Senior Engineer II (Senior Software Engineer)'.

Tier 3 — Methodologies and practices

Agile, Scrum, CI/CD, TDD, code review, on-call. These matter less than hard skills but are still filtered on, especially by non-technical recruiters.

Tier 4 — Soft skills (lowest weight)

Communication, leadership, collaboration. Almost never filtered on directly. Useful in bullet context but never a replacement for hard skills.

The exact-match problem

ATS parsers typically do not understand synonyms. 'ML' and 'machine learning' are two different search terms. 'Postgres' and 'PostgreSQL' are two different search terms. When in doubt, include both forms.

Led a machine learning (ML) platform migration from AWS SageMaker to a custom Kubernetes-based (K8s) inference stack.

What recruiters actually search

Based on publicly shared recruiter dashboards and our own analysis of thousands of job descriptions, the searches that convert candidates into interview pipelines tend to combine:

  • One primary technology (Python, Go, React)
  • One infrastructure or scale signal (AWS, Kubernetes, distributed systems)
  • One seniority anchor (Senior, Staff, Principal, or years of experience)
  • Occasionally one domain term (fintech, ML, infra, security)

The skills section is still worth having

Some advice columns argue the skills section is dead. It is not. A dedicated 'Skills' or 'Technologies' block remains one of the most reliable ways to get every relevant keyword indexed, especially for skills you have used but did not have room to mention in a bullet.

Keyword stuffing still fails

Do not paste white-colored keywords at the bottom of your resume. Modern ATS parsers detect this, many strip formatting before parsing, and recruiters spot it immediately when they open the file.

Automate the keyword match

The cleanest way to get keyword coverage right is to run a diff between the job description and your resume. Our free tool does this: paste both, get back a tailored resume plus a gap list of every keyword that was missing.