What counts as AI recruiting software?
AI recruiting software is any software that applies AI to a step of the hiring workflow. That is a wider definition than most vendors use, and the width is the whole point. Sourcing, screening, scheduling, interviewing, assessing, reference checking and job post auditing all have AI-native products in 2026, and they solve different problems for different people.
Most buying guides define the category as whatever the author sells. A sourcing company will tell you AI recruiting software finds and contacts candidates. A video interview company will tell you it screens them. Both are describing one layer of a stack that has twelve.
Full transparency: this post lives on Nova's blog, and Nova Recruiter appears in exactly one of the twelve categories below. It is absent from the other eleven, and the section near the end says plainly which problems it does not solve.
One boundary worth naming. Onboarding AI, the Enboarder and Kallidus end of the market, picks up after the offer is signed. It is adjacent to hiring rather than part of it, so it is not in the twelve.
The twelve categories at a glance
Vendor lists verified on each company's own site and newsroom in August 2026. Where a product now sits inside a larger platform, the parent is in brackets.
The twelve categories, one by one
1. AI sourcing platforms
The category that grew fastest, and the one where the marketing is loudest. Every vendor here advertises a profile count and they cluster between 750 million and over a billion, which means database size tells you almost nothing. What separates them is what happens after the search: how many outreach channels are genuinely included, whether the ranking is relevance only or relevance plus a quality signal, and whether seats are capped. Watch for email-only outreach sold as multichannel. Juicebox lists Gmail, Outlook and a mailbox count under its outreach section, and nothing else. We put seven of them side by side in a separate post, and if you are comparing against LinkedIn's own tools, the tier differences matter.
2. Applicant tracking with native AI
Every serious ATS now ships AI features, which makes "AI-powered ATS" a description rather than a category. The thing to watch here is the gap between announced and shipped. Greenhouse named four agents in 2026, Job Kickoff, Notetaker, Candidate Insights and Analytics Chart, and several carried a Q3 2026 launch window at announcement. That is normal in this category and it is also how a roadmap gets sold as a feature set. Ask which capabilities are generally available today, and get the answer in writing.
3. Talent intelligence and internal mobility
The enterprise end of the map. These systems score an entire workforce and an entire applicant history against roles, which is how internal mobility programs get built. They are also the slowest to implement and the most expensive. Greenhouse, which competes with Workday, writes in its own buying guide that Workday implementation "typically runs 12+ months". Below a few hundred employees this layer solves a problem you do not have yet.
4. Recruiting CRM and nurture
A CRM is what you need when the bottleneck is not finding people, it is losing them. It tracks candidates across months and across roles, and the AI is mostly about timing: when to re-engage, which pipelines have gone cold. The thing to watch is overlap. Several sourcing platforms now bundle a CRM module and several CRMs now bundle sourcing, so buying both often means paying twice for one pipeline view.
5. Interview scheduling
The least glamorous category and the one with the clearest return, because what it removes is pure coordination. The value scales with panel complexity: for one interviewer in one time zone, the calendar link you already have is enough. For a four-person panel across two continents it saves hours per role. The honest test is how many reschedules a week your team absorbs by hand.
6. Interview recording and notes
These tools transcribe the interview and map what was said to the questions on your scorecard, which is the part that changes decisions rather than just saving time. Two things to watch, and both are legal before they are technical. Recording consent rules differ by country and sometimes by state. And a transcript is evidence, which means it is discoverable.
7. Asynchronous video interviewing
Candidates record answers to fixed questions and the system scores them. It is also the category with the longest public track record of getting this wrong. HireVue announced in January 2021 that it was removing facial analysis from its scoring, saying visual analysis no longer added significant value, after an independent algorithmic audit by ORCAA and an FTC complaint from EPIC. The audit itself was not comprehensive and flagged that candidates with different accents, and candidates who gave very short answers, were scored less reliably (Fortune, SHRM). If you buy in this category, ask precisely what is being scored and what evidence supports it.
8. Skills and coding assessment
The only layer that measures the work instead of the description of the work, which is why it moves quality of hire more than anything upstream of it. The caveat is narrow and important: an assessment measures what it measures. A take-home that tests algorithms tells you about algorithms, not about whether someone can ship in your codebase. Match the test to the job, and cut the ones nobody reads.
9. Conversational AI and high volume screening
Built for hourly and frontline hiring, where the volume is thousands of applicants and the win is answering all of them within a day. It is also the most consolidated category on the map. Mya went to StepStone in 2021 and Crunchbase now lists the company as closed. Paradox, whose assistant is Olivia, has been part of Workday since October 2025. Neither is an independent product any more, so check who owns what before you shortlist here.
10. Reference checking
Structured reference collection scored against competencies, instead of a phone call with whoever the candidate picked. A small category, and shrinking: SkillSurvey now sits inside iCIMS and Searchlight moved to Multiverse in 2024 and into apprenticeships. Most useful at senior levels, where the cost of a bad hire justifies an extra step.
11. Resume parsing and matching
Infrastructure rather than something you buy on its own. It turns a document into structured fields so that everything above it can search. You are almost certainly paying for it already, inside your ATS or your sourcing tool. Worth knowing the names anyway, because parsing quality sets a ceiling on everything downstream: if the parser misreads a job title, no amount of AI ranking above it recovers the candidate.
12. Job description and language auditing
Audits the job post before it goes out and flags language that narrows the applicant pool. Cheap, fast, and it acts at the top of the funnel where a fix compounds across every application. The honest limit: it audits language, not outcomes. A perfectly worded post for a role with a ten-stage process and no salary range will still convert badly.
Seven acquisitions that took names off the map
Buying guides for this category, including ones updated this month, still list products that stopped being independent companies years ago. That matters for a shortlist in two concrete ways. The roadmap you are buying belongs to the acquirer now, and the standalone pricing you were quoted may not survive the integration.
Sources: Workday newsroom (Paradox, HiredScore), iCIMS newsroom (SkillSurvey), Textkernel (Sovren), StepStone (Mya), Kallidus (Sapling). Verified August 2026.
Two more worth knowing that are not in the table. Moonhub's team went to Salesforce in 2025, and Bravado has closed.
The pattern is not random. The acquirers are the platforms that already own the system of record, and what they are buying is the AI layer that sits on top of it. Workday now owns both a conversational agent and a talent orchestration engine, which is two of the twelve categories bought rather than built.
What is actually changing in 2026
From features to agents
The shift is from AI that helps a recruiter do something to AI that does it and reports back. In sourcing that means an agent that filters, shortlists, drafts the outreach and handles replies, with the recruiter approving rather than executing. The honest version of that claim always includes the approval step: an agent that creates and sends with no human in the loop is a compliance problem, not a feature. We broke down what each level of automation actually removes in a separate post.
MCP has become table stakes
Model Context Protocol lets an assistant like Claude or ChatGPT drive a tool directly, so the workflow happens in the assistant instead of in the vendor's interface. Twelve months ago that was a differentiator. It is not one now. Five of the vendors on this map publish one, each documented on their own site in August 2026: Gem calls it GeMCP, Greenhouse ships Greenhouse MCP, Juicebox documents a public endpoint, and Pin and SeekOut both list one in their product menus. The list outside this map is longer. If a vendor sells you MCP as unique to them, that tells you something about the vendor rather than about the protocol.
The EU AI Act arrived on 2 August 2026
From that date, AI systems used to evaluate, filter or rank candidates are regulated as high risk in the EU. The obligations that follow are practical rather than abstract: tell candidates AI is involved, keep a human able to review and overturn the outcome, document how the system works, and keep records. The responsibility falls on the employer deploying the system, not only on the vendor selling it, which is why it belongs in a procurement conversation.
Where Nova Recruiter fits
One category out of twelve, the first row of the table above: AI sourcing platforms. Nova Recruiter searches 800M+ public profiles, ranks them with the Talent Score, and runs outreach across Nova, LinkedIn, InMail and email. Pricing is usage based rather than per seat, with a free tier, Starter at EUR 49 and Growth at EUR 199 a month. It integrates with 60+ ATS platforms, and Nova complies with GDPR and the EU AI Act, with an ISO 27001 certification in progress.
What it is not: not an ATS, not a CRM, not an assessment tool, not an interview platform. If your bottleneck is inbound application volume or interview quality, the products that fix it live in categories 2, 7 and 8, and none of them are ours.
What AI recruiting software still does not do
Across all twelve categories there are three things none of these products do, and a vendor claiming otherwise is worth a second look.
- It does not make the hiring decision. Under the EU AI Act that has to stay with a person who can review and overturn the system.
- It does not tell you whether someone will thrive on your team. It can surface the evidence from an interview. The judgment is human.
- It does not build the relationship. The reason a senior candidate takes the call is still a person, and the reasons they ignore everything else have not changed just because the outreach got automated.
What has changed is the size of the pile of work that never needed a human in the first place.
Bottom line
Buy in the layer where your bottleneck actually is, and buy it knowing which twelve layers exist so you do not pay twice for the same capability. For most small and mid-sized teams the two layers with the clearest return are sourcing, where the time saved is largest, and assessment, where the lift on decision quality is largest. Everything else is worth having once those two work. And whatever you shortlist, check the date on the buying guide you found it in, plus who owns the vendor now.
Frequently asked questions
What is AI recruiting software?
Any software that applies AI to a step of the hiring workflow. In practice that spans twelve distinct categories, from finding candidates who never applied through to auditing the job post before it goes out. No single product covers all twelve, and most vendors cover one or two.
What is the difference between AI recruiting software and an ATS?
The difference is not AI. Modern applicant tracking systems are full of it. The difference is direction: an ATS manages the people who applied to you, and sourcing tools find and contact the people who did not. Most teams run both, with the ATS as the system of record.
How much does AI recruiting software cost?
It depends on the layer, and only some vendors publish a price. In sourcing, paid entry pricing verified in August 2026 runs from EUR 49 a month for Nova Recruiter Starter, $99 for Pin Solo and for Juicebox Starter on annual billing, and $149 for SeekOut Recruit Core on annual billing. Gem and hireEZ publish nothing at all. Free tiers exist at Nova Recruiter and at Pin, both capped. Talent intelligence platforms are quoted per contract and land an order of magnitude higher.
Is AI recruiting software legal under the EU AI Act?
Yes, and it is regulated. From 2 August 2026, systems that evaluate, filter or rank candidates are classified as high risk in the EU. That means candidates have to be told AI is involved, a person has to be able to review and overturn the outcome, and the system has to be documented. The obligation sits with the employer deploying it as well as with the vendor, so ask for the documentation during procurement, not after.
Can AI recruiting software integrate with my ATS?
Usually, and the count is a fair proxy for how mature a vendor is. Nova Recruiter integrates with 60+ ATS platforms. Two things worth checking on any vendor: whether the sync is one way or two way, and whether the integration is on your plan or only on the top tier.
Does AI recruiting software replace recruiters?
No, and in the EU it is not permitted to. What it replaces is the part of the job that was never judgment: searching, filtering, first touch outreach and scheduling. The parts that decide whether you land the hire, evaluating and persuading, stay human.
About the author
Andrea Marino
Co-founder and CEO
Andrea Marino co-founded Nova and is its CEO, where he leads the B2C business and Marketing. Italian-born and based in Milan, he studied International Management at Bocconi and CEMS, and previously worked at the digital transformation consultancy Alkemy.
