ATS Keyword Optimization for Resumes: The 2026 Guide

Over 95% of Fortune 500 companies and an estimated 75% of EU employers with 250+ staff use an Applicant Tracking System (ATS) to filter resumes before a human ever opens them (source: Jobscan 2023). Yet the median resume matches only 26% of the keywords in its target job description — which is why perfectly qualified candidates keep getting the automated rejection email twelve minutes after applying.

ATS keyword optimisation is the deliberate, evidence-backed practice of mirroring the exact skills, tools, certifications, and job-title language that appears in a posting, then weaving those terms into the parts of your resume that the parser reads first. Done well, it lifts your match rate from the 20s into the 70s — the band where recruiters start clicking "Review" instead of "Reject".

This guide explains what ATS keywords are, how European recruiters use them, how to extract them from a posting, and how AI tools like Alchema automate the whole loop without crossing into keyword-stuffing territory.

What are ATS keywords?

ATS keywords are the words and phrases that an applicant tracking system indexes from a job description and then searches for — or semantically matches against — your resume. They fall into five buckets:

• Hard skills and tools: Python, SQL, Kubernetes, Figma, SAP FICO, HubSpot
• Certifications and credentials: AWS Solutions Architect Associate, PMP, PRINCE2, ISO 27001 Lead Auditor, Scrum Master
• Regulations and frameworks: GDPR, HIPAA, SOX, MDR, EU AI Act, NIS2
• Soft skills: stakeholder management, cross-functional collaboration, conflict resolution
• Exact job titles: Senior Product Manager, Data Engineer, Ingenieur d'études, Werkstudent Marketing

Modern parsers (Workday, SAP SuccessFactors, Greenhouse, Lever, Teamtailor, Recruitee) no longer care about hidden white-text tricks. They build a structured profile of your resume, extract entities using NLP, and score the match against the posting. What they still reward is exact-phrase alignment in the right section: a skill listed in your Skills block counts, but a skill mentioned only in a three-year-old job title buried on page two counts less.

How do you optimise a resume for ATS?

Think of it as a three-pass workflow.

Pass 1 — Extract. Paste the job description into a keyword extractor (Jobscan, Alchema, Resume Worded) or use a large language model with a structured prompt. Ask for the top 15 keywords ranked by frequency and importance. Note which ones appear in the "required" versus "nice to have" sections.

Pass 2 — Map. For each keyword, ask: do I genuinely have this skill, and can I prove it with a project, metric, or certification? If yes, place it. If no, cut it. Inserting a skill you can't defend in the first interview is a waste of a keyword slot and a reputation risk.

Pass 3 — Place. Keywords belong in four locations, in decreasing order of ATS weight:

1. The professional summary — first 30 words, where parsers and humans both start reading.
2. The skills section — a scannable bulleted or tagged list of 12-20 core skills.
3. Experience bullets — each bullet should embed 1-2 keywords in context with a result.
4. Certifications and education — exact credential names, no abbreviations unless you also include the expanded form.

Avoid putting critical keywords only in the header, footer, a text box, a table cell, or an image. Those are the zones where parsers still fail most often (source: Greenhouse parsing benchmarks 2024).

Good vs bad keyword placement

Bad — keyword dumping, no evidence

Skills: Python, SQL, AWS, Kubernetes, Docker, Terraform, Python, SQL, cloud, DevOps, agile, scrum, Python, data

Good — keyword with quantified evidence

Senior Data Engineer, Delivery Hero (Berlin) — Built a Python and Airflow ingestion pipeline on AWS that loaded 2.4 TB/day from 17 sources into Snowflake, cutting reporting lag from 24 h to 40 min and saving EUR 180k/year in manual BI work. Owned Terraform modules and on-call rotation for the data platform.

The good version proves Python, Airflow, AWS, Snowflake, Terraform, data engineer, and data platform — seven keywords — while also giving the recruiter a concrete metric.

How many keywords should you target?

LinkedIn Talent Solutions' 2023 research found that resumes matching 70% or more of the required skills in a job description were shortlisted 4.8x more often than those matching under 40%. But matching more than 85% of keywords often signals keyword stuffing and triggers recruiter scepticism.

The sweet spot for most EU professional roles is 10-15 core keywords, each appearing 1-3 times across the resume in natural context. Tools like Alchema's keyword coverage score will flag both under-matching (below 55%) and suspicious over-matching (above 90%) so you stay in the high-trust band.

How does an ATS actually read my resume?

When you click "Submit" on a job board, the parser runs roughly this pipeline:

1. File validation — is it a supported format (PDF, DOCX, RTF)? Image-only PDFs and scans fail here.
2. Text extraction — OCR fallback for scanned files, native text extraction otherwise.
3. Section segmentation — the parser looks for canonical headers (Experience, Education, Skills, Certifications). Creative headers like Where I've Left My Mark confuse it.
4. Entity extraction — NER models pull out job titles, companies, dates, skills, locations, degrees.
5. Normalisation — MSc, Master of Science, M.Sc. collapse to one entity. Same for tools with aliases.
6. Scoring — a weighted match against the posting. Required skills weigh more than nice-to-haves; recency weighs more than ten-year-old roles.

That pipeline is why formatting matters. A gorgeous two-column resume with a sidebar looks great to a human but can scramble the section segmentation step, causing the parser to file your senior experience under "Education".

EU-specific ATS nuances

Inside the EU, three quirks matter.

• GDPR and lawful basis. Applicant data must be processed with a lawful basis — usually "taking steps at the request of the data subject prior to entering into a contract" (Art. 6(1)(b) GDPR). Good ATS vendors offer data-minimisation and deletion after a set period; you can ask about it.
• ESCO taxonomy. Many EU public employment services (including EURES) map job postings to the ESCO skills and occupations taxonomy. Using ESCO-standardised skill names on your resume boosts matching for EURES and public-sector roles.
• Language matching. A French posting expects French keywords even if the company is international. A German Projektleiter is not the same parser entity as Project Manager in a bilingual environment. Alchema auto-generates locale-appropriate variants for bilingual resumes.

Should you use AI to generate keywords?

Yes, but use it as a research assistant, not a ghostwriter. AI tools are excellent at extracting and ranking keywords from a posting, flagging missing coverage, and suggesting where to integrate them. They are unreliable at inventing experience for you or defending fabricated achievements in an interview.

A healthy workflow: AI extracts the top 15 keywords, you tick the ones you can prove, you write bullets with real metrics, then AI does a final pass to check coverage and readability. This is the core loop inside Alchema, built to stay on the right side of the keyword-stuffing line.

Frequently asked questions

What are ATS keywords on a resume? ATS keywords are the skills, tools, titles, and qualifications extracted by the applicant tracking system from the posting. Hard skills, certifications, regulations, soft skills, and exact job titles are the five categories that drive match scores.

How do I find the right keywords for my posting? Paste the job description into a keyword tool or LLM, extract the most frequent nouns and skill phrases, and weight any term appearing 2+ times or marked "required".

Will stuffing keywords get me rejected? Yes. Modern parsers use semantic scoring, and recruiters manually review shortlists. Aim for natural integration with evidence, not density.

Does an ATS read PDFs or Word better? Both, as long as the file is text-based (not a scan) and uses standard section headers. Avoid tables and headers/footers for critical content.

How many keywords should I include? 10-15 core keywords, each appearing 1-3 times in context with quantified evidence.

Are EU ATS systems different from US ones? The engines are the same. Language, photo/DOB conventions, and GDPR lawful basis differ by country.

Industry-specific keyword patterns

Different industries have different keyword density norms, and using the wrong vocabulary in the right role flags your resume as generic.

Engineering and software. Postings emphasise languages (Python, Go, TypeScript, Rust), frameworks (React, Next.js, FastAPI, Spring Boot), infrastructure (Kubernetes, Terraform, Docker, AWS/GCP/Azure), practices (CI/CD, TDD, code review), and data (SQL, dbt, Snowflake). Density is high — 20-30 concrete technical keywords across a senior resume is normal.

Sales and GTM. Keywords skew toward methodology (MEDDIC, SPICED, Challenger Sale, value selling), tools (Salesforce, HubSpot, Outreach, Gong, Clay), metrics (ACV, quota, pipeline, win rate, CAC), and segmentation (SMB, mid-market, enterprise, ICP). Density is medium — 12-20 keywords.

Product management. Frameworks (RICE, ICE, Kano, JTBD, North Star, OKRs), tools (Jira, Linear, Amplitude, Mixpanel, Figma), disciplines (discovery, experimentation, roadmapping, pricing), and seniority signals (cross-functional, stakeholder, executive). Density is medium.

Marketing. Channels (paid search, SEO, content, email, ABM), tools (HubSpot, Marketo, GA4, Looker, Braze), metrics (CAC, LTV, MQL, SQL, attribution, ROAS), and frameworks (AARRR, funnel, lifecycle). Density varies.

Finance, legal, compliance. Regulations (GDPR, MiFID II, SOX, Basel III, CSRD, CBAM, EU AI Act), tools (SAP, Oracle, Workday, Bloomberg), certifications (CFA, ACCA, CISM, PRINCE2). Density is high but precise — 15-25 specific regulation and certification names.

Match your density to your function; dropping 25 technical keywords onto a finance resume reads as copied from a different posting.

The skills section format that wins

Three formats parse well and read cleanly:

Grouped list

Languages:    Python, Go, TypeScript, SQL
Cloud & infra: AWS (EC2, S3, Lambda), Kubernetes, Terraform
Data:         dbt, Snowflake, Airflow, Postgres

Tagged inline

Skills: Python · Go · TypeScript · SQL · AWS · Kubernetes · Terraform · dbt · Snowflake · Airflow · Postgres

Two-column (only in ATS-safe tables, not visual sidebars)

• Works only if you test the parsed output. Risky unless you've verified with a target ATS.

Pick grouped list or tagged inline. Both score well on coverage and readability.