Dice

Dice's employer help article on Candidate Match (updated 16 November 2021) explains that a deep learning model 'using predictive analytics' reviews each applicant's tech skills and past, present and desired job titles. It places applicants into one of five classifications from Unlikely Match to Excellent Match and is retrained on recruiter feedback.

https://www.dice.com/support/employer/jobs/jobs---prioritize-apps-using-candidate-match.html

The candidate-facing Dice Match FAQ (updated 24 January 2024) says the models are calibrated on millions of tech job postings and use a skills taxonomy and 'Dice Knowledge Graph' to map skills and titles. It claims the models are 'over 96% accurate compared to how human experts would judge' qualification but does not publish the evaluation method.

https://www.dice.com/support/candidate-help/finding-a-job/dice-match.html

DHI Group's US patent 11,321,671 B2, 'Job skill taxonomy' (granted 3 May 2022), describes how the skills taxonomy behind Dice matching is built. It uses synonym files to map skill terms from job-description corpora to canonical skills, and machine-learned skill relationships (Word2Vec, LSA) to build a skill hierarchy.

https://patents.google.com/patent/US11321671B2/en

DHI's FY2025 Form 10-K (filed 12 February 2026) says Candidate Match and Search on Dice are powered by IntelliSearch, 'a proprietary machine-learning technology'. It also says the company has an internal AI usage policy, but that policy is not published.

https://www.sec.gov/Archives/edgar/data/1393883/000139388326000041/dhx-20251231.htm

No AI principles page, responsible-AI policy, model or system card, instructions for use, explainability statement or ISO 42001 certification exists, and the Q2 2026 AI Applicant Score has only marketing descriptions. Searched dice.com (Knowledge Center, hiring product pages, What's New, privacy policy, terms, sitemaps), dhigroupinc.com (corporate responsibility page, governance documents including the July 2024 Code of Conduct, 293 press releases) and the FY2025 10-K.

Have a URL that disputes this?→

The Dice Match FAQ says the models exclude demographic information and that Dice regularly measures them for bias across age, gender, race and other traits. It adds that 'a recent study of our models shows no evidence of perpetuating gender bias', but the study is unpublished, undated, not attributed to an independent auditor, and limited to the candidate-facing Dice Match score.

https://www.dice.com/support/candidate-help/finding-a-job/dice-match.html

The May 2019 Candidate Match launch release claims the tool's intelligent automation 'removes unconscious bias' and 'ensures equal consideration' because every application is run through the screening process, but cites no fairness testing or audit.

https://dhigroupinc.com/investor/investor-news/news-details/2019/New-Candidate-Match-Solution-from-Dice-Leverages-Proprietary-Matching-Algorithms-to-Speed-Recruiting-Tech-Talent/default.aspx

The AI risk factor in DHI's FY2025 10-K says 'AI algorithms may be flawed' and datasets 'may be insufficient or contain biased information', but discloses no bias testing or audit of Dice's matching or scoring.

https://www.sec.gov/Archives/edgar/data/1393883/000139388326000041/dhx-20251231.htm

No NYC Local Law 144 audit summary, named independent auditor (BABL AI, DCI, ORCAA, Warden AI, Holistic AI, Credo AI, ConductorAI or other), audit date or impact-ratio data was found for Candidate Match, the AI Applicant Score or IntelliSearch.

Where we lookeddice.com; the Knowledge Center; the dice.com/hiring sitemaps (~950 URLs); DHI's press-release archive (293 releases, 2016-2026); the FY2025 10-K. Warden AI's trust directory returns 404 for dice, dhi, dhi-group, dice-com and dicecom (control check: real customer slugs return 200).

Have a URL that disputes this?→

No EU AI Act statement, Article 26 deployer guidance, Article 27 fundamental rights impact assessment template, or provider/deployer role split exists.

Where we lookeddice.com hiring product pages, the AI Hiring Trends hub, the Knowledge Center, the privacy policy and terms, the dice.com/hiring sitemaps, and DHI's investor and corporate-responsibility pages.

Have a URL that disputes this?→

Dice's Terms (last updated 3 December 2025) state that the company makes 'no representation' that site materials are appropriate or available for use outside the United States. This fits a US-only footprint and the absence of any EU deployer documentation.

https://www.dice.com/about/terms-and-conditions

The Dice Match FAQ says the models 'do not factor in any demographic information' and that all personally identifiable information is removed before a match score is calculated. It also says employers must enter skills, titles and location. The commitment is stated only for the candidate-facing Dice Match, not the employer-side scores.

https://www.dice.com/support/candidate-help/finding-a-job/dice-match.html

Dice's privacy policy (last updated 3 March 2026) has an 'AI and Automated Decision-Making Technologies' section. It lists only customer support, network monitoring and security, and fraud and spam uses, and never mentions scoring or ranking applicants for employers. The policy also says Dice may request race and ethnicity, gender and veteran status, and that it sells publicly available, indexed or purchased personal information to employer customers.

https://www.dice.com/about/privacy-policy

DHI's corporate responsibility page (2025 reporting period) says TrustArc audits the brands' privacy practices annually and that a Data Protection Officer is engaged. Security is aligned to NIST CSF and reviewed quarterly by a Security Council, and zero data breaches are reported for 2025. No SOC 2, ISO 27001 or ISO 42001 certification is claimed.

https://dhigroupinc.com/corporate-responsibility/overview/default.aspx

The Knowledge Center's candidate-database-quality article (updated 7 December 2020) says every profile goes through AI-driven fraud screening, with exceptions reviewed by Dice's compliance team. It also cites a third-party analysis finding that 97% of profiles belong to genuine technologists.

https://www.dice.com/support/employer/searching/search---candidate-database-quality.html

There is no training-data statement, protected-attribute exclusion list or data-use disclosure for the employer-side AI Applicant Score, which the What's New page says scores education, industry and field of study. The Candidate Match article confirms recruiter feedback retrains the model, but no retention, consent or opt-out terms for that use are published.

Have a URL that disputes this?→

Recruiters see a Candidate Match level next to each applicant and can sort applications by Match. Clicking 'Match' in the application preview shows which job skills were mapped and which resume skills they matched, an in-interface explanation of the score (article updated November 2021).

https://www.dice.com/support/employer/jobs/jobs---prioritize-apps-using-candidate-match.html

The Q2 2026 entry says the AI Applicant Score 'automatically ranks every applicant' and scores each one across technical skills, experience, education, industry and field of study to give recruiters 'transparency' for shortlisting. Application Management Filters 'remove ineligible candidates before review begins' by work authorization, location, application date and fit status. Screeners (Q1 2026) filter out applicants with role-specific questions before they reach the pipeline.

https://www.dice.com/hiring/whats-new

The AI Applicant Score product video page (published 13 May 2026) is headlined 'AI does the ranking for you' and says each candidate is 'automatically ranked against your job requirements'. It describes no override or human-review controls.

https://www.dice.com/hiring/employer-resources/video/ai-applicant-score

The Knowledge Center's sourcing guide documents Unbiased Sourcing Mode, a one-click toggle in TalentSearch that hides factors commonly associated with bias, such as name, background and schools.

https://www.dice.com/support/employer-experience/finding-tech-professionals.html

The candidate FAQ tells job seekers that only they can see their Dice Match score, that it has 'no determining weight' when they apply, and that employers cannot filter out lower scores. None of this addresses the separate employer-side Candidate Match levels or AI Applicant Score.

https://www.dice.com/support/candidate-help/finding-a-job/dice-match.html

The Knowledge Center, hiring product pages and What's New log contain no public documentation of: a way to turn off AI scoring or fit-status filtering; a score-override control; decision audit logs; per-jurisdiction settings; or a guarantee of human review for applicants who are filtered out.

Have a URL that disputes this?→

No public status page exists. status.dice.com has no DNS record, dice.statuspage.io redirects to Atlassian's Statuspage marketing site, and dhigroupinc.statuspage.io exists but returns 'Your page is inactive'. dice.com/.well-known/security.txt returns 404 and HackerOne reports no 'dice' program. No model-update changelog, AI incident channel or published fairness monitoring results were found.

Have a URL that disputes this?→

DHI's corporate responsibility page describes ITIL-based incident, problem and change management and reports 2025 service metrics: 99.655% availability against a 99.8% monthly uptime target, three critical-feature disruption events and one performance issue. These are group-level uptime figures, not AI monitoring.

https://dhigroupinc.com/corporate-responsibility/overview/default.aspx

Dice's Privacy & Trust Center invites job seekers to report job postings that seem 'discriminatory' through a Report control routed to the Compliance team or by email to compliance@dice.com. It says this feedback helps 'improve our algorithms and processes'. This is a misuse channel, not an AI incident channel.

https://www.dice.com/privacy-center/how-to-report-misuse

The What's New page is a quarter-by-quarter product release log for 2019-2026. It records launches such as Match Scores for Recruiters (Q2 2019), Dice Match (Q1 2023) and the AI Applicant Score (Q2 2026), but it tracks features, not model versions, retraining or performance changes.

https://www.dice.com/hiring/whats-new

The Post Jobs page describes the recruiter-facing Match Score ('Our tech taxonomy ranks candidates by skills—not just keywords') and claims that 71% of applicants are high-quality matches, without explaining the method.

https://www.dice.com/hiring/post-jobs

The Find Candidates page documents: IntelliSearch, where recruiters paste a job description and 'our AI instantly identifies key skills and matches candidates'; a 'Likely to Switch' prediction of whether candidates are open to new opportunities; and Unbiased Sourcing.

https://www.dice.com/hiring/find-candidates

Dice's 'Trust Gap in Tech Hiring 2025' report on its AI Hiring Trends hub advises recruiters to: guarantee human review; offer a second human review of AI rejections; 'use AI to surface candidates, not eliminate them'; and regularly audit AI decisions against human judgment. This is general good-practice advice, not documentation of Dice's own tools.

https://www.dice.com/hiring/recruitment/reports/trust-gap-in-tech-hiring

Dice's Terms (last updated 3 December 2025) disclaim 'any and all responsibility for any employment decisions made by any Employer' and make customers indemnify Dice for claims arising from their hiring or employment decisions. They contain no AI-specific terms for deployers.

https://www.dice.com/about/terms-and-conditions

The What's New log lists IntelliSearch for Bullhorn (Q2 2021) and IntelliSearch for Greenhouse (Q4 2021), plus Easy Post and Easy Apply integrations with iCIMS, Greenhouse, Bullhorn and SmartRecruiters. These bring Dice's matching into customers' ATS workflows.

https://www.dice.com/hiring/whats-new

No DPA, subprocessor list, AI Applicant Score help article, AEDT candidate-notice template, or compliance guidance for NYC LL 144, Illinois HB 3773, Colorado or the EU AI Act was found. dice.com/about/dpa, /hiring/dpa, /hiring/security, /hiring/trust and /hiring/responsible-ai all return 404.

Have a URL that disputes this?→