Greenhouse AI Resume Screening, What It Does

Short answer: Yes, through Greenhouse Talent Matching, and with real limits worth understanding before you budget around it. Talent Matching reads resumes and application responses against criteria you set and assigns each applicant a match score in one of five bands: strong, good, partial, limited, or needs manual review. It is scoped to the first Application Review stage, a recruiter switches it on for a specific job rather than it running across your inbound flow automatically, and Greenhouse states it is "assistive AI, not automated-decision-making" that "does not automatically advance or reject candidates". Greenhouse publishes no price for it.

This article previously reported that Greenhouse declined to score candidates for role fit. That was true of its earlier position and is no longer accurate: Talent Matching now assigns match scores. The section below has been rewritten against Greenhouse product documentation checked on August 23, 2026.

Does Greenhouse use AI to screen resumes?

Greenhouse uses AI on resumes, but not to screen them in the sense most recruiters mean. Checked on August 11, 2026, its AI touches resumes in three places: a resume review feature that parses applications quickly while redacting identifying details, Talent Filtering that searches and filters the pile, and Talent Rediscovery that pulls relevant past candidates back out of your own database. None of those produce a fit score.

The word "screening" is doing a lot of hidden work in this question, which is why the answer confuses people. Parsing is reading a resume into structured fields. Filtering is narrowing a list to the people who match a search. Screening, as most talent teams use the word, means evaluating every applicant against the role's requirements and telling you who is worth a recruiter's hour. Greenhouse does the first two well. It does not attempt the third.

One practical note before anything else: greenhouse.io now redirects to greenhouse.com, so older documentation and bookmarks still work but the product pages have moved.

Can we use AI to screen resumes or rank candidates in Greenhouse?

Yes to both, with one important caveat. Greenhouse Talent Matching screens resumes against criteria you define and sorts applicants into five match bands, which is a ranking. What it will not do is run automatically across your whole inbound flow: a recruiter activates it per job, and it only evaluates candidates sitting in the first Application Review stage.

That scope is the thing to plan around. If your team opens Talent Matching on every requisition as a deliberate step, you get scored, banded candidates and the review gets much faster. If nobody opens it, nothing happens, and a pile of 600 applications sits exactly where it was. It is a tool a recruiter picks up rather than a process that runs on its own, and Greenhouse is explicit that it never advances or rejects anyone by itself.

Greenhouse also stops short of a single composite fit score on purpose. You get a band, not a number, and no ordered list from best to worst inside a band. For most structured-hiring teams that is a feature rather than a gap, because it keeps a human making the call. If what you need is the whole inbound pile read as it arrives and returned in rank order with a written reason per candidate, that is a different shape of tool, and the Greenhouse pricing page covers what the platform costs either way.

What Greenhouse AI actually does, feature by feature

Greenhouse names nine AI capabilities on its own AI recruiting page. Here is what each one does and, more usefully, what it does not touch.

Capability	What it does	What it is not
Resume review	Parses resumes quickly while redacting identifying details to support blind review	Not scoring. It structures and anonymizes, it does not rate fit
Talent (re)discovery	Surfaces qualified talent from your existing database and reduces duplicate applications and spam	Not outbound sourcing across the open web
Talent Filtering	Keyword search across resumes and internal notes, with required and preferred toggles	Not a ranking. It returns matches, in no particular order of merit
Automated hiring plans	Generates structured interview stages, questions and scorecard attributes	Not candidate evaluation. It builds the process, not the shortlist
Key takeaways	Automated interview transcription, analysis and summaries	Not a hiring recommendation
Sourcing team productivity	Surfaces interested candidates faster with automatic sentiment tracking and routing	Not applicant screening. This works on sourced leads
AI insights	Creates report filters and builds reports from text prompts	Not predictive analytics on individual candidates
Offer forecasting	Predicts the likelihood a candidate accepts an offer	Not a fit score. It models acceptance, not suitability
Continuous improvement	Analyzes candidate feedback to improve the hiring process	Not applicant-level evaluation

Read that column of "what it is not" and the pattern is obvious. Greenhouse has invested heavily in AI that speeds up the work around the decision: writing the plan, finding the record, summarizing the interview, forecasting the close. It has pointedly not built AI that makes the shortlist for you.

Does Greenhouse score or rank candidates?

Yes, in bands rather than in rank order. Greenhouse Talent Matching evaluates candidate qualifications against criteria you define, analyzes resumes and application responses, and sorts applicants into five categories: strong match, good match, partial match, limited match, and needs manual review. It highlights the relevant skills and industry experience behind each result rather than returning a bare number.

Notice what Greenhouse still declines to do. There is no single composite fit score and no straight one-to-forty ranked list. That distinction is deliberate and it is consistent with the position Greenhouse has argued publicly for years: its AI recruiting material says AI can inform, summarize and surface insights but the decision is "always yours, not handed to an algorithm," and its buying guidance warns against systems using black-box composite scoring. Five explainable bands is the compromise between that stance and the practical need to triage a large pile.

The operational limits matter as much as the capability, and Greenhouse documents them plainly. Talent Matching is only available on candidates in the first Application Review stage of your pipeline. It does not process inbound applications on its own; a recruiter opens Talent Matching on a given job to activate scoring, and both Talent Matching and resume parsing have to be enabled first. Greenhouse also recommends calibrating on four to six key skills, because the match score is distributed across whichever skills you pick, and a longer list of matched terms does not necessarily produce a higher score.

Greenhouse is explicit that this is "assistive AI, not automated-decision-making," that it "does not automatically advance or reject candidates," and that recruiters and hiring managers remain responsible for all hiring decisions. For a US employer that is a genuinely useful sentence to have from a vendor, because the employer keeps legal responsibility for a selection procedure regardless of who built the model.

So the real question for a high-volume team is not whether Greenhouse scores, it is whether scoring that a recruiter has to switch on per job, on one pipeline stage, matches the shape of the problem. If a requisition draws 40 applications, this is plenty. If forty requisitions each draw 600, the constraint moves from "can it score" to "does it score everything, everywhere, without anyone remembering to turn it on."

What is Greenhouse Talent Filtering?

Talent Filtering is keyword search over your applicants. You search resumes and internal notes for job titles, skills, locations and similar terms, and you can toggle each keyword between "Preferred," which behaves as an OR, and "Required," which behaves as an AND. Greenhouse also generates suggested keywords from the job post so you are not starting from a blank box, and AI-enabled filters can surface candidates by preferred years of experience and related skills.

It is a good version of what it is, and it is faster than reading. But a filter answers a different question from a screen. A filter tells you who used the words you searched for. It cannot tell you that one of those people has the certification and the shift availability the role actually requires while another matched on a keyword buried in a list of technologies they touched once in 2019. It also silently penalizes the candidate who described the same skill in different words, which is the exact failure mode keyword filtering has always had.

What is Greenhouse Real Talent?

Real Talent is Greenhouse's answer to a problem that got much worse quickly: spam, fraud and cheating in application pipelines. It sorts and filters candidates into a tiered inbox based on application quality and the likelihood of fraudulent or spam activity.

This is the closest Greenhouse comes to automated triage, and it is worth understanding precisely what it is triaging. Real Talent is sorting on authenticity, not on suitability. It is trying to answer whether this is a real person making a genuine application, which is a genuinely useful thing to know now that a candidate can generate and submit fifty tailored applications in an afternoon. It is not trying to answer whether that real person can do the job.

Both questions need answering, and confusing them causes buying mistakes. A pipeline cleaned of spam is still a pipeline of 400 unranked genuine applicants.

What this means if your pile is 600 deep

Greenhouse's approach works well when a recruiter can realistically read every application. It strains when they cannot, and the strain shows up in ways that are easy to miss because nothing appears broken.

The shortlist quietly becomes a function of arrival time. Recruiters work from the top of the queue, attention degrades through a long pile, and the applicant who arrived on day nine gets a different quality of review from the one who arrived on day one. Nobody decided that. It is just what happens when the review capacity is smaller than the pile.

Keyword filters then compound it, because the recruiter has to guess the vocabulary in advance. Search "RN" and you miss "registered nurse." Search both and you still miss the person who wrote their license number and their unit. The candidates you never see are invisible by definition, so the process never generates evidence that it is failing.

This is the gap a dedicated AI resume screening layer fills, and it is why teams running Greenhouse buy one rather than replacing the ATS. Greenhouse stays the system of record for requisitions, candidate records, scheduling, scorecards and offers, which is what it is genuinely excellent at. The screening layer reads every inbound application as it lands, applies the knockout and eligibility criteria you wrote for that specific requisition, asks your screening questions, scores role-fit against those criteria and hands back the pile ranked with a plain-language reason under each candidate.

The reason matters more than the ranking. Greenhouse's objection to composite scoring is correct as far as it goes, and the answer to it is not to stop ranking. It is to make the ranking explainable: criteria you set explicitly, visible evidence per candidate, and no automated rejection at any point. If a screening tool cannot show a hiring manager exactly why candidate 12 outranked candidate 40, Greenhouse's warning applies to it and you should listen.

Greenhouse AI compared with a screening layer

	Greenhouse AI	A screening layer
Reads every inbound application	Scored on the first Application Review stage, once switched on per job	Yes, every application evaluated against your criteria as it arrives
Produces a ranked shortlist	Five match bands, no composite score or rank order	Yes, ranked with a written reason per candidate
Knockout and eligibility criteria	Expressed as search filters you run manually	Set once per requisition and applied to everyone automatically
Spam and fraud tiering	Yes, through Real Talent	No, that stays with your ATS
System of record	Yes, this is the core product	No, it sits alongside your ATS
Interview process and scorecards	Yes, including AI-generated hiring plans	No

Questions worth asking before you rely on any of this

If you are evaluating whether Greenhouse's AI is enough on its own, the useful test is not a feature list. It is whether a recruiter can currently explain, for a specific requisition you closed last quarter, why the five people who got phone screens were chosen over the other 300. If the honest answer is that those five were near the top of the queue and looked fine, you have a throughput problem that filtering will not solve.

Ask any vendor in this space, including whoever you already pay, the same five questions. Can you see the criteria behind an individual candidate's position, in plain language, for a specific requisition? Has the tool been bias-audited, when, and can you get the results? Does the system ever reject an applicant without a human, or does it only rank and flag? Are candidates told that AI or automated review is used? Is there a record you could hand a regulator showing who decided what, and when?

Greenhouse's answer to several of those is that Talent Matching is assistive, never advances or rejects a candidate on its own, and shows the skills behind each match band rather than a bare number, which is a defensible position and a useful one to have in writing. A screening layer has to answer them directly instead, which is why explainability and human decision rights matter so much more once ranking enters the picture.

One last thing worth separating out, because it gets conflated with screening constantly. Offer forecasting predicts whether a candidate will accept, and when that number comes back low the cause is usually not the quality of your evaluation. It is that the offer is not competitive for the level and the market, which is a question about what the role actually pays elsewhere rather than anything an ATS can fix. Screening decides who you talk to. Compensation decides who says yes.

The short version

Greenhouse uses AI for resume parsing and redaction, keyword filtering, talent rediscovery, interview summaries, hiring plan generation, offer forecasting and spam and fraud tiering through Real Talent. It scores applicants into five match bands on one pipeline stage when a recruiter turns it on, stops short of a composite fit score, and advises buyers against tools that score opaquely. That makes Greenhouse a strong applicant tracking system whose screening is real but narrow: one pipeline stage, activated per job, five bands rather than a ranked pile. Teams hiring at volume generally add a screening layer that reads everything automatically, ranks transparently and never auto-rejects, rather than replacing an ATS that is doing its actual job well. What none of it comes with is a price, which we break down in full in our guide to Greenhouse recruiting pricing. If you are weighing that decision, the Greenhouse alternative comparison lays out where each product fits, and the same question applied to other vendors is covered in does Workday use AI to screen resumes and iCIMS AI resume screening, where the answers turn out to be very different. iCIMS in particular does rank applicants, which makes the contrast with Greenhouse's position unusually sharp. HireVue AI resume screening lands somewhere else again: it evaluates candidates thoroughly, but only the ones who complete an assessment or interview. And Workable sits at the opposite end from Greenhouse entirely, including resume screening in its Recruiting plans at no extra charge and metering a more autonomous agent at one credit per candidate.

The whole field, sorted by whether each vendor will actually rank the pile rather than just store it, is laid out in our AI candidate screening software roundup.