A year ago most employers that participated in our CandE Benchmark Research Program told us that they were not using AI recruiting technologies to screen, match, and/or rank job applications.
In 2025, nearly 40% said they did. With another 21% considering it for next year.
Prior to this year when we’ve asked companies about the AI recruiting technologies they utilize, we didn’t specifically ask about AI application screening. But that’s only because it’s still so new to the market.
There were also increases in utilization of basic chatbots, conversational AI chatbots, sourcing candidates from external sources (matching / virtual assistant), and sentiment analysis of candidate and/or employee open-ended feedback (a key differentiator of Survale Enterprise and the CandE and EmployE Benchmark Research Programs). See the table below.
What underscored the increased utilization of the AI application screening was the untenable increase in applications employers received this year, especially for larger employers. A big part of that increase were more serial applications, those candidates who are applying for jobs they’re not qualified for, and automating the process.
And even though less than 3% of the candidates told us this year they used AI application platforms to apply for multiple jobs at once (like LazyApply.com, Jobhire.ai, and others), job application volumes have increased dramatically. Fake candidates are also on the rise and North Korea’s IT army has been hacking the global job market, which is straight out of a Tom Clancy novel.
Again, the application volumes have been untenable for larger employers. In our 2023 CandE research:
- 49% of employers with 10,000+ employees globally told us they had 51+ applications per job posting.
- 13% told us they had 200+ applications per job posting.
But this year in 2025:
- 64% of employers with 10,000+ employees globally told us they had 51+ applications per job posting.
- 21% told us they had 200+ applications per job posting.
That equates to a 31% increase since 2023 when there were 51+ applications per job posting and a 62% increase when there were 200+ applications per job posting.
Mercy me. This is why, current AI-bias lawsuits and all, more and more employers with application volume will be implementing and utilizing AI application screening technologies, which today is offered by many mainstream ATS’s and talent intelligence platforms.
Combine that with more and more candidates using generative AI to improve their resumes and cover letters. Plus, there are those tech savvy candidates injecting meta data into their uploaded applications to try to game the AI screening technologies.
The AI recruiting-and-hiring escalation between candidates and employers is here to stay.
To keep AI application screening as fair as possible, while improving better selection outcomes, we recommend that employers:
1. Begin with High-Quality, Representative Training Data
More bias comes from the data itself, not algorithms. Employers should ensure that diverse training datasets reflect gender, race, age, background, and experience variety. This includes balanced samples so that historically overrepresented groups don’t dominate the model.
2. Use Transparent and Explainable AI Models
Employers should select AI systems that allow them to understand what features the model uses – skills, keywords, and/or total years of experience – and why candidates are ranked a certain way. This transparency enables auditing and the correction of hidden bias.
3. Conduct Regular Bias Audits
Audits should identify any disparate impact across gender, race/ethnicity, age, disability, veteran status, and other protected groups depending on local law.
4. Always Keep Humans in the Loop
AI shouldn’t make final hiring decisions. Employers must ensure a human review of AI-flagged rejections and rankings. This helps catch errors and contextual nuances algorithms can miss.
5. Standardize Job Requirements and Scoring
To reduce subjective interpretation, write clear, measurable criteria – for example, “2 years Java experience,” not “strong coding skills”. Use structured scoring rubrics for all candidates as this uniformity reduces inconsistent or biased filtering.
6. Remove Sensitive Attributes From Inputs
AI technologies must not process protected demographic data unless specifically allowed for bias mitigation, and instead focus on skills, achievements, and experience.
7. Follow Regulatory & Ethical Guidelines
Current legal frameworks affecting AI hiring include EEOC (U.S.) guidance on algorithmic fairness, NYC Local Law 144 (audits required for automated employment decision tools), California’s AI hiring law (prohibits automated decision systems from discriminating against protected classes), EU AI Act (high-risk category for hiring tools), and ISO/IEC 42001 (AI management systems). Employers should document compliance and maintain risk assessments.
8. Involve DEI, Legal & HR Teams Early
AI procurement and deployment should not be left solely to IT or vendors. Employers should make it cross-functional where HR/TA defines job criteria & fairness needs, DEI reviews for demographic impacts, legal ensures regulatory compliance, and IT/Data teams validate technical safeguards.
9. Train Recruiters on AI Literacy
Employers need to educate staff on what AI can and cannot do, how to spot biased outputs, when to override AI recommendations, and how to escalate algorithmic issues. This helps to bridge technology and human judgment.
10. Communicate With Candidates
Transparency builds trust and reduces legal exposure, which is why employers should disclose when AI is being used and how it influences screening and hiring decisions. Also, let candidates know how they can request accommodations if necessary.
Unfortunately, job seekers believe that artificial intelligence is selecting them out of the running for jobs, especially those dispositioned after they’ve applied because they’ve been deemed unqualified. This was one of the top negative open-ended comment themes in our 2025 CandE Benchmark Research. All of our 2025 research briefs and reports will be available at the end of the year and early 2026.
Take serial applicants out of the equation, those candidates who apply for anything and are usually perpetually unqualified, and you’re left with people who want a job or really need a job and feel they’re somewhat if not fully qualified. Letting them know that recruiting teams are screening their applications and not AI, if that is the case, has been something employers have been posting on their career sites and included in candidate communications. If AI screening technologies are used, then emphasizing that “humans are making the final decisions” is just as important.
Will this messaging make a definitive and positive brand-impact difference in the future? Possibly, or maybe not. Again, the AI recruiting-and-hiring escalation between candidates and employers is here to stay. But transparency has always been a primary differentiating factor in our candidate experience benchmark research for the past 14 years, and since recruiting is in the business of “no” – saying no to a lot more candidates over and over again – communication and feedback loops will always be the greatest recruiting investment.
| AI Recruiting Technologies Utilized (partial list) | 2025 | 2024 |
| Screening, matching, and/or ranking job applications | 39% | NA |
| Basic chatbot recruiting automation (binary Q&A “customer-service” on career site) | 38% | 34% |
| Sourcing candidates from external sources with a virtual assistant | 36% | 22% |
| Generative AI that analyzes job postings for unintentional bias and offer recommendations on how to create job listings with more inclusive language | 27% | 28% |
| Conversational AI chatbots that answer candidate questions | 23% | 19% |
| Assessing candidates to identify team personality and culture fit | 21% | 24% |
| Sourcing existing internal candidate database/s with a virtual assistant | 20% | 22% |
| Sentiment analysis of candidate and/or employee open-ended feedback | 20% | 17% |
