Most ATS advice circulating in Israeli tech WhatsApp groups is wrong, or at best calibrated for the 2015 American market. The reality in 2026: companies like Wix, Monday, and Check Point use Greenhouse, Comeet, or Workday; AllJobs and Drushim run their own legacy parsers; and the gap between "passes the bot" and "gets a recruiter call" is wider than candidates think. Here is what actually happens to your CV between upload and a human eye, and the concrete moves that matter.

How an ATS actually reads your CV

Forget the folklore about "75% match scores" and "AI screening." Most Israeli-market ATSes do three things in sequence: parse the document into structured fields, index the text for keyword search, and surface a candidate list ranked by recruiter-defined filters. Greenhouse and Comeet (used by Monday, Wiz, Riskified, AppsFlyer) lean heavily on parsing. Workday (Check Point, Mobileye) is more rigid about field mapping. Drushim and AllJobs run older keyword-index logic that gets confused by anything fancy.

The parser walks your file looking for predictable section headers — Experience, Education, Skills — then assigns every block of text to a field in their candidate database. When it can't figure out where a chunk belongs, it dumps it into a generic "other" bucket that recruiters never search. A two-column layout, a header inside a text box, or a creative section name like "Where I've Built Things" is enough to make 30% of your CV invisible.

The recruiter then runs Boolean searches against the indexed text: "React" AND "Node.js" AND ("AWS" OR "GCP"). If your CV doesn't contain those literal strings, you don't appear in the result set. There is no semantic understanding. Reactjs and React.js are different tokens. K8s and Kubernetes are different tokens. This is the single biggest mistake mid-career engineers make.

The Israeli-tech keyword pitfalls

The Israeli market has its own vocabulary problems that generic ATS guides ignore.

Spelling variants that actually matter

Full Stack vs Fullstack vs Full-Stack — Wix and Monday job postings use "Full Stack." AllJobs recruiters search "Fullstack." Use both somewhere on the page.
Node.js vs NodeJS vs Node — Greenhouse parsers treat these as distinct. Pick two and use them.
Backend vs Back-end vs Back End — Same problem. The job description tells you which the recruiter is searching for. Mirror it.
QA vs Quality Assurance vs SDET — Israeli job posts increasingly use SDET; your CV from 2019 probably doesn't.

Military service: the 8200 question

If you served in 8200, Mamram, Talpiot, Ofek, or any technical unit relevant to the role — name it explicitly. Do not write "Intelligence Corps" and assume the recruiter will infer. Israeli tech recruiters search for the unit name as a keyword. Wiz, Check Point, and most cyber companies have hiring managers who explicitly filter for these.

If you served in a non-technical role (combat, logistics, education), include service dates and unit name without overselling. "IDF, 2014–2017" with a one-line description is enough. Padding a non-technical service record with technology buzzwords trips both the ATS (false positives in the experience-years calculation) and the human reviewer.

US-style vs Israeli-style headers

Drop the US objective statement. Drop the photo (yes, even though Israeli LinkedIn culture loves them — most ATS parsers either strip images or get confused by them). Keep the header tight: name, role, city (Tel Aviv, Herzliya, Haifa — not "Greater Tel Aviv Area"), one phone, one email, LinkedIn URL. That is the entire header.

File format: PDF or DOCX?

PDF, exported from a single-column document, generated by a real word processor (Google Docs, Word, Pages). Not "Print to PDF" from a design tool. Not Canva. Not Figma export. Those produce PDFs where the text is positioned glyph-by-glyph and parsers reconstruct it as gibberish.

The test: open your PDF, select all the text with Cmd+A, copy it, paste into a plain text editor. If the result reads top-to-bottom in a sensible order, the parser will handle it. If it's scrambled, your two-column layout or text-box header is the problem. Fix the source document, do not fix it in the PDF.

DOCX is acceptable for Comeet and Greenhouse. Avoid it for Workday, which has historically choked on Word-generated tables. When a portal lets you upload either, choose PDF.

Bullet structure that survives both filters

Every bullet should answer three questions: what did you do, what was the measurable outcome, and what stack did you use. Verb-result-metric, with the technology embedded.

A bullet without a number is a job description. A bullet with a number is a claim. ATSes don't care about the difference, but the human reading the shortlist absolutely does.

Before

Worked on the checkout team to improve performance and stability of the payment service, collaborating with backend and DevOps engineers.

After

Reduced p95 checkout latency from 1.4s to 320ms by rewriting the payment-routing service in Go (from Node.js), cutting Stripe webhook timeouts by 87% across 4M monthly transactions.

The "after" version contains: a verb (Reduced), a metric (1.4s → 320ms), a second metric (87%), a scale anchor (4M monthly), the stack (Go, Node.js, Stripe), and an implicit team context (rewrote a service, which means owned a project). One bullet, six pieces of recruiter-relevant signal, every keyword a Boolean search would find.

Three to five bullets per role is the right density for a 3–8 year CV. Ten bullets per role signals junior-level over-explanation.

The skills section trap

A standalone "Skills" or "Technologies" section helps the parser fill the structured Skills field, which recruiters do filter on. But it hurts you when it lists tools you used once for a hackathon. The recruiter sees "Kubernetes, Kafka, Spark, Terraform, Airflow" on a CV from someone who spent four years writing React, calls them in for a backend role, and the candidate is exposed in the first 10 minutes of the screen.

The rule: list a technology in your skills section only if you'd be comfortable being asked a 20-minute deep-dive question on it. Everything else goes inline in the relevant bullet, which gives it appropriate context ("built a one-week PoC with Airflow" is honest; "Airflow" in a flat skills list is not).

Group skills by category (Languages, Frontend, Backend, Infra, Data) rather than dumping a 40-item comma-separated list. Categorized lists parse cleanly into Greenhouse's skill taxonomy. Flat lists often get truncated.

Hebrew, English, or both?

One CV, English-only, for any company whose job description is posted in English. That covers ~90% of Israeli tech roles at the companies in scope here. Do not submit a bilingual side-by-side document — parsers split it into garbage and recruiters find it visually noisy.

Keep a Hebrew version for AllJobs postings written in Hebrew and for non-tech-fluent staffing agencies. Mirror the structure exactly; do not "improve" the Hebrew version with extra detail.

The checklist

Single-column layout, no text boxes, no headers/footers with content the parser needs. Test it: copy-paste the PDF text into a plain text file. If it reads in order, you're fine.
Standard section headers: Experience, Education, Skills, Projects (optional), Military Service. Not "My Journey" or "What I Bring."
Mirror the exact spelling from the job description for your top 5–8 keywords. If the post says "Node.js," your CV says "Node.js" — not "Node" or "NodeJS."
Three to five bullets per recent role, each with verb + metric + stack. Older roles (5+ years back) can be one or two bullets.
Name your IDF unit if it's technical and relevant. 8200, Mamram, Talpiot, Ofek, C4I, Matzpen — spell them out.
Skills section is short and honest. Categorized, not a flat dump, and only includes things you'd survive a deep-dive interview on.
PDF from a real word processor, single column, no Canva, no Figma, no design-tool exports. File name: FirstnameLastname_Role_2026.pdf, not resume_final_v3.pdf.
Run the keyword check against the actual job description before each application. Paste both into the match analyzer and fix the gaps it surfaces — the missing keywords are usually the difference between landing in the recruiter's inbox or their archive.

What to do tomorrow morning

Open your current CV. Do the copy-paste-into-plain-text test. Count the bullets per role and check how many contain a number. Open the last three job descriptions you're targeting and circle every keyword that doesn't appear verbatim on your CV. Fix those three things first; everything else is optimization. If you're rebuilding the document from scratch, the resume builder enforces the structural rules above by default — single column, parser-safe sections, ATS-friendly headers — so you can spend your time on the bullets, which is where the actual work lives.