If you applied to ten Israeli tech jobs last week, your CV passed through three pieces of software before any human looked at it — and most likely none of those pieces of software were the same one. Israel runs on a different ATS stack than the US, and the recruiter UI on top of it shapes what gets read and what gets dropped.

Three systems do roughly 80% of the work. Comeet for local startups and most homegrown unicorns. Greenhouse for the international R&D centers and the bigger growth-stage companies. Workable for SMBs, agencies, and series-A startups that need something off-the-shelf and cheap. Lever exists in Israel but it's rare — I've seen maybe one or two listings using it in the last year. Bullhorn shows up at staffing agencies, not direct employers.

What each system does to your CV before a recruiter sees it is different enough that "ATS-friendly" as a generic concept is almost useless. Let me break down what's actually happening.

What an ATS parser actually does

Every ATS does three things when your PDF lands:

Tokenization — splits the text into words, normalises case, strips punctuation. This is where weird formatting breaks things.
Section assignment — assigns chunks of text to fields: Experience, Education, Skills, Summary. This is where the parser guesses what your "Where I've Built Things" section is, and it usually guesses wrong.
Ranking signals — pulls keywords matching the job's required and preferred skills, computes a match-style score, sorts you in the recruiter's candidate list.

The recruiter doesn't see your PDF first. They see a structured candidate card the parser built — name, current role, years of experience, parsed skills as tags, and a list of bullets pulled from the most recent job. If those fields are wrong or empty, you're at the bottom of the list before anyone read a word you wrote.

Comeet — the Israeli default

Comeet is Tel-Aviv-built, used by Wix, Monday, Lightricks, ironSource, AppsFlyer, and a long tail of Israeli startups. If you applied to a homegrown company directly through their careers page and the URL contained comeet.co or the embedded form looked clean and minimal, you were in Comeet.

What it does well: Comeet is genuinely Hebrew-aware. RTL parsing works, Hebrew section headers ("ניסיון תעסוקתי", "השכלה") get recognised, and the keyword matcher does basic morphological matching on Hebrew verbs — which most international ATS products do not. If you submit a Hebrew CV, Comeet handles it; if you submit it to Greenhouse, you're rolling dice.

Where it chokes: multi-column PDFs. Comeet's text extractor reads column-by-column in single-column mode by default, so a two-column CV gets serialised as "left column top to bottom, then right column top to bottom" — which scrambles your sections beyond recognition. The "Skills" sidebar gets read after the entire experience section ends, so the parser thinks your skills are somehow part of your education, and the keyword scoring breaks. I've seen strong candidates get scored below noise on Comeet purely because of a designer-built two-column template.

Recruiter UI quirk: Comeet shows the parsed candidate card alongside an "AI summary" — three bullets the system pulls from your most recent role. If your most recent role's bullets are weak, that's literally the first thing the recruiter sees. Front-load your strongest bullet at the top of your current job, not buried in position three.

Greenhouse — the international R&D default

Greenhouse runs Mobileye, Wiz, Riskified, JFrog, Check Point's product groups, and basically every Israeli R&D center owned by a US parent. If the careers page felt American — Greenhouse-branded job board, "Apply" button, structured questionnaire about visa status and US salary expectations — you were in Greenhouse.

What it does well: structured field mapping. Greenhouse parses your PDF into very specific fields and is rigid about them. If your CV has clear "Experience", "Education", and "Skills" sections with English headers, Greenhouse extracts them cleanly and the recruiter UI populates correctly.

Where it chokes: anything non-standard. A summary section called "About Me", a skills section called "Tech I Use", a Hebrew header in an otherwise English CV — Greenhouse treats it as unstructured text and dumps it in a generic notes field. The parsed candidate card shows blank fields, the recruiter sees an empty profile, and your application gets sorted to the bottom.

Recruiter UI quirk: Greenhouse lets recruiters Boolean-search across all candidates with operators — "kafka" AND ("python" OR "go") AND "tel aviv". If your skills section uses different casing or spacing than the search query, you don't appear. "Node.js" with a dot, "NodeJS" without, "Node" alone — these match different searches. Pick one variant per skill and use it consistently. I prefer the dotted form because that's what most JDs use.

Workable — the SMB and series-A default

Workable runs a long tail of Israeli companies — series-A startups, scale-ups not big enough for Greenhouse, and a chunk of the consulting and agency market. The parser is the most permissive of the three.

What it does well: messy CVs survive better here than anywhere else. Multi-column? Workable handles it. Weird section header? Workable still recognises it. Tables? Mostly fine.

Where it chokes: not on the parser side — on the recruiter side. Workable's recruiter UI is the weakest of the three for filtering and ranking, so recruiters using it tend to skim less and revert to the original PDF. That sounds like a positive — and it is, if your CV is strong on visual scan. It's a negative if your strength was the keyword density that the parser would have surfaced. Workable recruiters skip nuance more than Greenhouse or Comeet recruiters do, because the tool doesn't help them surface it.

What this means for how you write the CV

Before I get to the keyword tactic, run the match analyzer against the actual JD you're applying to. It scores how your CV's keyword density and section structure match the job, and tells you what's getting dropped before a human sees it. Then come back for the rest.

A few rules survive across all three systems:

Single-column PDF, always. Two-column templates look great on screen and break on Comeet. There's no upside.

English section headers — "Experience", "Education", "Skills", "Summary". Not "Where I've Built Things", not "My Stack", not "Career Highlights". Greenhouse won't recognise the creative header. Comeet will, but you're hurting yourself with the other 60% of applications.

Keyword mirroring without stuffing. If the JD says "Kafka" and you have it, write "Kafka" — not "message queues" or "event streaming". The parser is doing exact-match tokenization, not semantic search. Mirror the JD's exact stack vocabulary.

Hebrew and English variants for split applications. Maintain two versions: a clean English CV for Greenhouse-driven roles, and a Hebrew or bilingual CV for Comeet-driven Israeli companies that want to see Hebrew. Don't translate your English CV to Hebrew with Google Translate — the result reads like a robot wrote it and Israeli recruiters will spot it in two seconds.

A concrete before/after

Here's a bullet that the parser would token-split badly:

Owned end-to-end the backend/infra of payments — including Java/Kotlin services, Kafka pipelines, and Postgres → Snowflake ELT.

The slash-separated tokens (backend/infra, Java/Kotlin, Postgres → Snowflake) get split or merged inconsistently. Comeet might tokenize "Java/Kotlin" as a single token "java/kotlin" that matches neither "Java" nor "Kotlin" in the JD. Greenhouse usually handles slashes but the unicode arrow breaks it.

Rewrite:

Owned backend and infrastructure for payments. Built Java and Kotlin services, Kafka event pipelines, and a Postgres-to-Snowflake ELT processing 12M events per day.

Same content. Every keyword (Java, Kotlin, Kafka, Postgres, Snowflake, ELT) is now a clean token. The parser pulls all six. The recruiter scanning the candidate card sees all six.

What to do tomorrow morning

Open your CV. Three checks:

Single column? If not, switch to a single-column template. This morning. Before you apply to anything else.
Section headers in English and standard? "Experience", "Education", "Skills", "Summary". Rename anything creative.
Slash-separated tech tokens? Replace every Java/Kotlin with Java and Kotlin. Every Postgres/MySQL with Postgres and MySQL. Every frontend/backend with frontend and backend. Five-minute job.

That's the floor. After that, the 12 CV mistakes article covers what the human reads after the ATS hands you off, and the military service section piece covers the one Israel-specific section that none of these parsers know what to do with — but every Israeli recruiter scans for first.

The ATS isn't smart. It's a series of dumb pattern-matchers stacked on top of a PDF parser that was written in 2014. You don't beat it with cleverness. You beat it by writing for what it actually reads.