Greenhouse

Greenhouse sells a hiring system, also called an applicant tracking system. Two of its AI products sort and screen applicants. Real Talent sorts the people who applied into match groups and flags fake applications. Voice AI holds a screening interview by voice. Both are paid add-ons. Real Talent’s price comes through a request form. Voice AI’s price depends on the size of your organisation.

By Ashish Manchanda Published 14 Oct 2026 Last updated 14 Oct 2026

Greenhouse sorts applicants and screens them by voice, and your people decide

In one line

Real Talent groups applicants by how well their CVs match the role and flags likely fraud. Voice AI interviews applicants by voice and scores the answers. Neither AI product moves anyone on or turns anyone away by itself.

Buy it if

You want AI screening inside the system that already holds your applicants. Or you use another hiring system and want a voice interview that connects to it. You accept that your company carries the legal duties.

Do not buy it if

You need an uptime promise and a security addendum for voice interviews. You need voice interviews outside the United States now.

Price

Real Talent features sit under Greenhouse’s Core, Plus and Pro plans. Voice AI is a yearly subscription that scales with the size of your organisation.

What Real Talent and Voice AI do with an applicant

Real Talent is a group of AI tools that sit on the applications you receive. Talent Matching is the part that sorts them into match groups. It works on the candidates in the first Application Review stage of a job. Your team sets a list of weighted skills and experience for the role. Greenhouse calls this a calibration. It suggests focusing on four to six key skills. Talent Matching then reads each CV against that list. It uses only the CV. It does not use names or contact details.

Each candidate lands in one of five groups. They are Strong, Good, Partial, Limited, and Needs manual review. A side panel shows why a person got their group, broken down by experience, industries and skills. Inside each group, candidates sit in order of application date. Greenhouse does not give candidates one number to rank them. The system blocks you from adding a protected trait such as gender to the list. It warns you if a skill could stand in for one.

Your team can switch any AI feature off for the whole company. For Talent Matching, you can also switch it off by office, department or job. A candidate can ask for a person to review their application. If you turn on the candidate opt-out, a candidate who opts out lands in Needs manual review. A Past candidates tab shows earlier applicants with filters instead of match groups. A free assistant, still a test version, turns a plain description of a role into a working setup.

Real Talent also checks for fraud. It looks at an applicant’s phone number, email address, location and IP address. An IP address is the internet address the application came from. IPQualityScore, an outside company, returns 26 signals. This check does not use AI. A separate identity check uses CLEAR, which looks at biometric signals, documents and trusted data sources. CLEAR keeps the selfie and the government ID. Greenhouse receives the pass or fail result, name, email, phone number and IP address.

Voice AI holds a live, two-way voice interview. It asks questions that fit the role. Your team sets a marking guide, and Voice AI scores every answer against it. It runs in 11 languages. The result is an overall score and a score for each question. A written reason cites what the candidate said. Voice AI does not score accent, tone, fluency, pace, stutters, pauses, personality or confidence. It does not read emotion or build voiceprints. Cheat detection listens to the voice only. It does not watch the screen or video. Cheat and fraud signals sit beside the score and are not part of it.

A hiring system can invite a candidate to a Voice AI interview when they reach a chosen stage. The scores, transcript and reasons then go back to the candidate’s record. In the Zapier example, the conversation took 25 minutes. Your people make every decision to advance, reject or rank an applicant.

Greenhouse also sells Sourcing Automation, an add-on for finding people who have not applied. AI can draft each message step, and a person must approve it before it is sent. Emails stop when the person replies.

Voice AI has its own page in our interview guide. For how it compares with other voice interviewers, read Greenhouse Voice AI.

What Greenhouse and Zapier have seen with Voice AI

Zapier, a fully remote AI automation company

Zapier used Voice AI for its own hiring. 273 applicants had the conversation over 75 days. Greenhouse estimates it saved 228 recruiter hours.

Zapier candidates

97% of candidates rated the experience excellent. 6% chose to opt out. Those who opted out got the same consideration.

Greenhouse, across Voice AI

Candidates gave Voice AI 4.5 out of 5 on average. 90% gave it 4 or 5 stars. 10% opted out.

One Greenhouse customer, not named

For one role, 500 people applied. Voice AI screened 200 whom the recruiters had passed over. It sent 19 back, and 17 cleared the recruiter’s review. 5 reached the final round and 1 was hired.

Who Greenhouse suits

You already run Greenhouse for hiring

You can add Real Talent and Voice AI to a Greenhouse subscription on the Core, Plus or Pro plan. Scores, transcripts and written reasons go back to the candidate’s record.

You use a different hiring system

Voice AI also works without the Greenhouse system. It connects to more than 60 hiring systems. They include Workday, SAP SuccessFactors, Oracle Recruiting Cloud and UKG. Ezra, the product behind Voice AI, is still sold on its own, to Greenhouse customers and others.

Too many applications arrive that are spam or fake

Real Talent checks fraud signals, and it can check identity too. Fraud and spam detection comes with the Plus plan.

You hire outside the United States

Voice AI is available in the United States, with more regions planned. Ask Greenhouse which countries it can interview candidates in now.

Where Greenhouse stops, from legal duties to usage caps

Your company carries the legal duties

Greenhouse’s AI Addendum makes your company solely responsible for its AI policies, candidate notices, consents and human oversight. Greenhouse is not responsible for errors in AI output. Your company also covers Greenhouse if someone sues it because your company broke a law. Under the master subscription agreement, that means employment, data protection or AI laws. The Voice AI Terms add fines and penalties to that.

Some places require a candidate’s consent before Voice AI records

Voice AI records audio and video and writes a transcript. Depending on where your applicants are, you may need a separate, clear yes from each one first. You must also give every applicant a clear notice before the session. An applicant who opts out goes to an alternative process that you define.

Voice AI has fewer contract protections than the main product

Under the Voice AI Terms, Greenhouse’s service level agreement does not apply to Voice AI. That agreement promises uptime and gives you money off your bill when it fails. The Security Addendum does not apply either. The Voice AI Terms point to a separate security controls page for Voice AI instead.

A fraud flag can be wrong

Under the Real Talent Terms, fraud detection may flag a genuine application as fraudulent. You may not use Real Talent results as a factor in deciding if someone can get a job. That includes identity checks. The Terms also bar any use covered by the US Fair Credit Reporting Act. CLEAR’s identity check is not a background check. A candidate can decline it, and declining does not count as a failure.

The match tool is not meant for the EU’s high-risk uses

Annex III of the EU AI Act lists AI that filters job applications and evaluates candidates. Under the Real Talent Terms, you acknowledge that Talent Matching is not intended for those high-risk purposes. Greenhouse also expects Voice AI to count as a high-risk AI system when it evaluates candidates. If you hire in the EU, ask Greenhouse how you should use Talent Matching.

Several outside companies handle candidate data

Talent Matching sends details read from the CV to OpenAI. Voice AI uses Deepgram, Daily.co, and OpenAI or Google language models. Anthropic and OpenAI host the AI models behind other Greenhouse features. Amazon Web Services stores data in the United States, Germany and Australia.

Greenhouse keeps a licence to anonymised AI data

Greenhouse does not train AI on customer data for Talent Matching or Voice AI. The AI Terms also give Greenhouse a licence with no end date. It covers what goes into the AI and what it produces, in anonymised form. Greenhouse can use it to improve and train its services and models. The AI Addendum lets Greenhouse use customer data and feedback to develop AI features. Greenhouse promises to make a fair effort to keep personal data out of training.

Greenhouse can slow your AI features if you use them heavily

The AI Addendum lets Greenhouse switch off or reduce AI features if you pass what it decides is “fair usage”. That decision is Greenhouse’s alone. The Addendum also lets it charge extra fees for AI features.

Warden AI, an outside firm, audits Talent Matching and the voice interviewer every month. The Talent Matching audit covers 10 traits that the law protects, such as sex and race. Local Law 144 is a New York City law on automated hiring tools. Warden’s report on Talent Matching under that law comes from controlled tests. It is not a formal certification or a compliance ruling. Both Warden reports used Warden’s own data set of real candidate profiles, not yours.

How Greenhouse is priced and what happens when you leave

Real Talent features sit under Greenhouse’s Core, Plus and Pro plans. Talent Matching sits under Core. Fraud and spam detection sits under Plus. You request Real Talent pricing through a form.

Voice AI is a yearly subscription that scales with the size of your organisation, not per user. Bought on its own, it is priced by headcount with a fair-use allowance of interviews. It is also an add-on for Core, Plus and Pro. Customers usually start with a contract of 12 months or longer. Greenhouse offers an optional pilot at the start, during which you can leave the yearly contract.

Under the master subscription agreement, fees follow your employee headcount. Greenhouse can raise fees at renewal. Each subscription renews for one year unless either side gives notice 30 days before the renewal term. Payments cannot be cancelled or refunded, except in the cases the agreement names. Late fees carry interest of 1.5% a month. Each side’s liability is capped, with exceptions, at the fees from the 12 months before the event. Sourcing Automation has its own cap of $5,000.

During the term, a free access key lets you export your data. After the agreement ends, Greenhouse queues your data for deletion 90 days later. Within 30 days of the end, you can ask for an export file. Greenhouse charges its data migration rates for it. Voice AI recordings, transcripts and scores stay for the length of the contract unless you delete them sooner. Greenhouse deletes them within 30 days of your request. They stay in backups for 30 days.

You could also find and screen people with agents built on your own systems. We have mapped what that looks like. The HR Agent Map

Eight questions to put to Greenhouse

Voice AI sits outside your service level agreement and Security Addendum. Will you write an uptime promise and security terms into our order form? Good answer: A written uptime figure with money back when it fails, and security terms that name the controls.
Your AI Terms keep a licence to anonymised data going into the AI and coming out of it. Which of our candidate data would that cover, and can we exclude it? Good answer: A written list of what is covered, and a clause that excludes our candidate data.
Which countries can Voice AI interview candidates in today? Good answer: A named list of countries, with dates for any that are still planned.
Some hiring systems let a recruiter advance or reject a candidate from inside Voice AI. Who sets that, and who clicks? Good answer: A named person on our team sets it. The record shows who clicked and when.
What does Voice AI cost for our headcount, and how many interviews does the price cover? Good answer: A price in writing, an interview number per month, and the price of going over.
Can we see the full Warden AI bias audit reports for both products? We want the tables for each group. Good answer: The full reports for both products, with tables we can show a regulator.
You can slow or switch off AI features if we use them too much. What monthly limits will you commit to for our account? Good answer: A stated limit per month, a warning before we reach it, and no switch-off without notice.
Your Terms say Talent Matching is not meant for high-risk uses under the EU AI Act. How do you expect us to use it for applicants in the EU? Good answer: A written answer that says what use is allowed, and who carries the risk if we apply it.

Alternatives to Greenhouse

The alternatives are Workable, Jobvite and Lever, or buying nothing.

Workable also builds AI into its hiring system. Workable Agent runs the work from finding people to a shortlist. It searches a database of more than 400 million profiles and gives each candidate a score out of 100. You add it to any Workable plan and pay in prepaid usage units that Workable calls AI credits. Evaluating a candidate costs 1 credit. Sourcing a passive candidate costs 2. Bundles start at 5,000 credits for $600.

Jobvite and Lever belong to the same company, Employ. Jobvite’s Talent Fit ranks the most qualified candidates for a role and explains each match. Its Zero-Click Intelligent Sourcing starts a search when a role opens. Jobvite lets customers start with the core system and add extra parts. Lever builds AI matching, with criteria your team can change, into its hiring system.

Buying nothing means no new contract terms on candidate data and no renewal to track. It also means people on your team sort applications and run screening calls by hand. Our guide to AI sourcing and screening covers what the rest of the market offers.

More in this guide

The AI sourcing and screening guide
Compare the 41 with their own page what each of the 41 products with its own page here does, what it costs, what it can decide alone and which have an outside bias audit.
Mid-market companies which AI sourcing and screening tools suit a company of 200 to 2,000 people, the prices you can see, how the bill grows and what to ask before signing.
Enterprise companies which platforms replace an applicant tracking system and which add to one, what they cost, how the contracts end and the hiring rules by place.
Hiring in India what Indeed charges for search in India and how Naukri sells Resdex, which products host data or launched in India, what India’s data protection law asks of employers and ten questions to ask sellers.
High-volume hiring which AI screening products can reject applicants with no person involved, how they text and call, and what they cost at volume.