Greenhouse's Ryan Bot: The 2 Must-Have Skills That Auto-Reject You

Greenhouse's support docs, updated August 2026, quietly formalized a filter most applicants never see: an AI recruiter named Ryan reads your resume against a "Must Have" skills list, and if any of them are missing, you never reach a human. The rule sounds mild until you notice recruiters are handling 412% more applications per seat than in 2022. Here is how the "2 Must Haves" floor actually fires, why Greenhouse's public "we don't auto-reject" line does not match its own partner documentation, and what to change on your resume before you send the next one.

What Ryan actually does inside Greenhouse

Ryan is an AI recruiter, sold through Greenhouse's partner directory as "AI Screened," that reads every resume against a recruiter-defined Must Have skills list and disqualifies any resume missing a Must Have before a human opens it. Greenhouse's own support article (29122364864539, updated August 2026) describes the flow: Ryan "reviews resumes against your job requirements, runs structured phone screens when your team sends an invite, and returns screening results to Greenhouse."

The loop is fully wired into your candidate stage:

• Ryan ingests the job description.
• The recruiter tags Must Haves and Nice to Haves inside the AI Screened app.
• Resumes get parsed and checked against Must Haves.
• Candidates who pass get scheduling, confirmations, reminders, and the phone screen itself, all from Ryan.
• Scores and transcripts push back into Greenhouse, and stage moves can be automatic: "Candidates advance or archive automatically."

Greenhouse's example in its own doc is a Node.js developer role. If "the candidate's resume doesn't mention any JavaScript or NodeJS experience," Ryan disqualifies. Note the verb: mention, not demonstrate.

The "at least 2" floor is a floor, not a ceiling

The support article is explicit: "At a minimum you need 2 Must Haves." Greenhouse recommends recruiters "start with only listing things your candidates absolutely must have." That guidance is aspirational. With application volume up triple digits, recruiters treat Must Haves as a volume-control lever, and stacking three, four, or five is the norm.

The 2.74x pool problem hiding inside a Node.js job

Ryan's Must Have filter systematically favors candidates who redundantly list adjacent skills over candidates who are actually more skilled. That is not opinion. It falls out of the underlying talent pool math when you match it to Greenhouse's own example.

In Refolk's index of professional profiles, here is what the pool for Greenhouse's own Node.js example looks like across the US:

Only 36% of "Node-relevant" engineers list Node.js alone without also listing JavaScript. The other 64% list both. Ryan cannot tell whether "both listed" means "actually knows JavaScript" or "padded the skills bar." It just checks strings.

Why the mechanism is keyword-density, not skill

Greenhouse's doc says disqualification triggers when a resume "doesn't mention" a Must Have. Ryan's check runs on parsed resume text. That has two consequences applicants keep missing:

1. A senior engineer who wrote "built the JS runtime layer for our checkout service" without ever typing "JavaScript" fails a JavaScript Must Have.
2. A bootcamp grad who put "JavaScript, Node.js, React, TypeScript" in a skills bar passes.

The rewrite is not about lying. It is about making sure the literal Must Have term appears near real evidence of the work. Which is the exact chore Refolk handles: paste the posting, and Refolk rewrites your own resume so the terms Ryan is grepping for sit next to the accomplishments that prove them.

How to reverse-engineer the 2 Must Haves in any JD

Assume 3 to 5 Must Haves per posting, not 2, and read the JD like a keyword contract. Recruiters under volume pressure stack Must Haves as a triage tool, so the "start with 2" guidance in Greenhouse's docs almost never survives contact with a real requisition.

Here is the read-the-JD checklist that actually maps to Ryan's behavior:

1. Find the hard-noun skills in the first three bullets of "Requirements." Named languages, frameworks, tools, and certifications. These are the Must Have candidates.
2. Look for repetition across the summary and requirements block. If a term shows up in both, it is almost certainly a Must Have.
3. Watch for "required" vs "preferred" language. "Preferred," "bonus," "nice to have," and "familiarity with" mark Nice to Haves. "Must," "required," "strong experience in," and "expert-level" mark Must Haves.
4. Count named tools in one product family. If a JD names Snowflake, dbt, and Airflow, all three are probably Must Haves, not two.
5. Cross-check the company's other open roles. Recruiters reuse Must Have templates across a family of postings.

The goal is not to guess. It is to make sure every plausible Must Have appears verbatim in your resume, next to a bullet that proves it. Refolk does this per posting, which matters because the Must Have list for the same title varies across companies more than candidates expect.

Must Have vs Nice to Have: candidates optimize the wrong list

Must Haves are binary knockouts. Nice to Haves are where scoring happens. Most applicants pad the Nice to Have side of the resume because it feels safer, and lose on the Must Have side without realizing it.

The practical rule: get to 100% Must Have coverage before you write a single "nice to have" bullet. If a JD lists Kubernetes as a Must Have and you have run it in production, the word "Kubernetes" needs to be on the page. "Container orchestration" does not count to a string-matching parser.

The Greenhouse contradiction candidates should know about

Greenhouse's public candidate blog says the platform "never uses AI to rate candidates or auto-reject applications." Its own support docs describe AI Screened doing exactly that, with automatic stage advancement or archival wired into Greenhouse. Both statements are technically compatible and practically misleading.

The seam:

• Greenhouse the core ATS does not, by itself, auto-reject.
• AI Screened, sold in Greenhouse's partner directory and documented in Greenhouse's support hub, does.
• Greenhouse AI, launched March 2026, adds native ML scoring of resumes against scorecards on top of AI Screened.

So when a candidate reads "Greenhouse doesn't auto-reject" and treats it as reassurance, they are reading marketing about one layer while the layer above it (Ryan) has already archived them. Assume any Greenhouse-powered posting could be running AI Screened, Greenhouse AI, or a third-party screener like Brainner, Skima AI, or Applicant AI, all of which write pass/fail signals back into Greenhouse.

Parser quality now beats resume quality

In 2026, Greenhouse reads your resume twice: once as a file for the human recruiter, and once as parsed text for the AI layer. A beautifully designed resume that scrambles skill keywords in the parse fails Ryan even if a human would have loved it.

The parser hazards that actually kill applications:

• Skills inside images or icon graphics. The parser sees nothing.
• Two-column layouts with skills in a sidebar. Parsers often read left-to-right across columns and split your skill list into other sections.
• Non-selectable text from a resume exported as an image PDF.
• Custom fonts that render fine visually but break glyph extraction.
• Skill terms only in the header or a graphical badge, never in a bullet.
• Nonstandard section headings like "What I'm Great At" instead of "Skills" or "Experience."

The fix is boring: single-column, selectable-text PDF, standard section names, and every Must Have skill term written into both a Skills line and at least one Experience bullet where the work happened. This is what Refolk generates by default when it drafts your resume from your history.

A concrete pre-submit checklist for Greenhouse postings

Before you hit Submit on any Greenhouse posting in 2026, run this list. It is the shortest path to passing an AI resume screener without gaming it.

1. Identify the top 5 likely Must Haves from the JD using the read pattern above.
2. Confirm each appears verbatim, spelled exactly as the JD spells it (Node.js vs NodeJS vs Node), in your Skills section.
3. Confirm each appears again in at least one Experience bullet with evidence: scope, scale, or outcome.
4. Remove skills that sound similar but are not the JD's exact term, or add both.
5. Convert the file to a single-column, selectable-text PDF.
6. Read the plain-text version (copy-paste into a note app). If a Must Have does not survive the copy-paste, the parser will not see it either.
7. Draft the cover letter to reinforce the same Must Have terms, not a different vocabulary.

Steps 2, 3, and 7 are the tedious ones and the ones people skip when they are on their 30th application of the week. That is where Refolk earns its keep: it rewrites your resume and cover letter against each posting's Must Have list, and scores how well your history actually matches before you send.

Ryan is not alone in the Greenhouse marketplace

AI Screened is the most visible auto-disqualify layer inside Greenhouse, but it is not the only one, and the norm across the marketplace is now keyword-gated screening.

• Brainner recommends mandatory vs preferred criteria against a claimed 3.5B data points and writes results into Greenhouse.
• Skima AI writes a 0 to 100 SKIMA Match Score into custom Greenhouse fields.
• Applicant AI sits in front of Greenhouse, Lever, and Workable as a pre-filter before the ATS sees you.
• Greenhouse AI (March 2026) adds ML scoring against job-specific scorecards natively.

Across Greenhouse's 7,000+ customers and 40M+ applications processed, the practical baseline in 2026 is: assume at least one AI layer has already run before your resume gets a human read, assume it is checking strings against a Must Have list, and write for both audiences at once.

FAQ

How do I know if a job I'm applying to uses AI Screened?

You often cannot tell from the posting itself. Signals include a mention of an "AI phone screen" or "Ryan" in the application flow, a scheduling link that arrives within minutes of submission, or a company known to be a Greenhouse customer that has publicly discussed AI screening. The safer assumption in 2026 is that any Greenhouse-powered posting could be running AI Screened, Greenhouse AI, or a third-party pre-filter, and to write your resume so it passes a string-matching Must Have check regardless.

Is it "keyword stuffing" to list a skill I only touched once?

There is a real line, and it is about evidence, not counts. Listing "JavaScript" once in your Skills section and once in a bullet that describes what you actually built with it is honest resume writing. Listing "JavaScript, TypeScript, Node, React, Vue, Angular, Svelte, Ember" when you have shipped one React app is the pattern that gets flagged by phone screens and eventually references. Cover the Must Haves you can defend, and let the rest go.

Does a great cover letter save me if a Must Have skill is missing?

No, not against Ryan. Greenhouse's own doc says the Must Have check runs on the resume, and a missing Must Have triggers disqualification before the human read where a cover letter would matter. Use the cover letter to reinforce the Must Have vocabulary already in your resume, not to compensate for gaps in it. If a Must Have is genuinely missing from your background, apply anyway, but expect a lower hit rate on that specific posting.

What is the fastest way to tailor a resume per posting without spending an hour each time?

The manual version is the pre-submit checklist above: extract 5 likely Must Haves, confirm verbatim coverage in Skills and Experience, verify the PDF parses cleanly. That runs 15 to 30 minutes if you already know your history. Refolk shortens that to a paste-the-posting step: it drafts the tailored resume and cover letter from your own history, uses the JD's exact Must Have vocabulary, and scores your fit so you can decide whether the posting is worth sending in the first place.