[Survey of 503 Recruiters] 64.0% of New Graduate Recruiters Have Rejected Candidates Due to 'AI Documents'

SHIFT AI Inc. (Headquarters: Shibuya-ku, Tokyo; CEO: Shota Kiuchi; hereinafter 'the Company'), which operates the generative AI learning community 'SHIFT AI' with the No. 1 number of users under the mission of 'Making Japan an AI advanced country', conducted a survey on the actual use of AI in recruitment among 503 recruiters and HR personnel nationwide. Key Findings of the Survey 1. 53.9% of recruiters use AI. Reaches 72% for new graduate recruiters The percentage of those 'using' AI in recruitment activities is 53.9% overall. Compared to 61.3% for mid-career recruiters, new graduate recruiters lead at 72.0%, showing that the adoption of AI in new graduate recruitment is advancing. 2. 76% of new graduate recruiters 'sense AI documents'. 66.9% for mid-career The total percentage of those who 'sense' it is 57.1% overall, 66.9% for mid-career, and 76.0% for new graduates. Three out of four new graduate recruiters feel the presence of AI documents. Even in mid-career, those who 'strongly sense' it is 29.0%, significantly exceeding the overall average (18.1%). 3. The 'too perfect problem' emerges as a unique challenge in mid-career recruitment The top reason for sensing an AI document is 'refined expression but lacking specific details' (56.4% overall). The second reason, 'too perfect to differentiate', is 56.6% among mid-career recruiters, significantly higher than the overall (43.2%), indicating that the homogenization of resumes makes differentiation difficult. 4. About half of recruiters have rejected candidates due to AI documents The total percentage of those who 'have rejected' is 48.5% overall, 52.0% for mid-career, and 64.0% for new graduates. On the other hand, 23.8% of mid-career recruiters answered 'cannot judge', revealing the reality that AI detection itself is becoming difficult. 5. Internal rule development is lagging, leaving responses to individual judgment on the front lines Only 14.9% have 'clear rules that are well-known'. The total for 'no rules/left to the person in charge' and 'don't know' reaches 38.6%, showing that guideline development is not keeping up with the rapid increase in AI documents on the front lines. 6. 'Designing evaluation criteria' is the top task that cannot be replaced by AI overall. 'Cultural fit' stands out in mid-career 'Designing recruitment standards and evaluation criteria' tops the list at 30.0% overall. For mid-career recruiters, 'assessing the candidate's personality and cultural fit' is the highest at 38.3%, showing a particularly strong demand for human involvement in tasks that assess individual aptitude. Survey Details Q1. Do you use AI in your recruitment activities? The percentage of those 'using' AI in recruitment activities is 53.9% overall. It is notably high among new graduate recruiters at 72.0%, about 10 points higher than mid-career recruiters (61.3%). The percentage of those 'not using' AI is 38.2% overall, with fewer non-users among new graduate recruiters (20.7%) than mid-career recruiters (34.7%). The data shows that the new graduate recruitment scene is leading in recruitment DX. Q2. In which processes do you use AI in recruitment activities? (Multiple answers) Among AI users (Overall n=271, Mid-career n=152, New graduate n=198), the most common process is 'primary screening and automatic evaluation of application documents' (Overall 50.2%, Mid-career 65.8%, New graduate 53.0%). It is particularly high among mid-career recruiters at 65.8%, indicating a high need for automation in mid-career recruitment where document volume is large. This is followed by 'automatic generation of job descriptions and requirements' (Overall 48.7%) and 'automatic extraction of scout candidates' (Overall 44.3%), showing that AI use is spreading across the entire upstream recruitment process. Q3. Do you feel that candidates are using generative AI to create application documents? The total for 'feel' is 57.1% overall, 66.9% for mid-career, and 76.0% for new graduates. Among new graduate recruiters, 'feel to some extent' reaches nearly half at 49.8%, showing that the perception of AI documents is widely shared. For mid-career recruiters, 'strongly feel' is 29.0%, significantly higher than the overall average (18.1%), confirming a rapid increase in AI-created resumes. Q4. When you feel an application document is AI-generated, on what basis do you judge it? (Multiple answers) The top reason for feeling it is an AI document is 'expressions are refined, but lack specific episodes or numbers' (Overall 56.4%, Mid-career 57.8%, New graduate 57.4%), which is almost the same regardless of the recruiter type. Notably, the second reason, 'the writing is too perfect, making all candidates look equally accomplished', is 56.6% among mid-career recruiters, significantly higher than the overall 43.2%. While AI documents were generally considered problematic for being 'thin on content', this survey revealed a new perspective that the problem with AI documents is that they are 'too perfect to reject', meaning they are 'too flawless'. Q5. Have you ever rejected a candidate because their application document was highly likely created by generative AI? The total for 'have rejected' is 48.5% overall, 52.0% for mid-career, and 64.0% for new graduates. In particular, about two out of three new graduate recruiters have experienced rejecting candidates, making it clear that AI documents are affecting actual recruitment decisions. On the other hand, 23.8% of mid-career recruiters answered 'cannot judge, so there is no standard to base it on', highlighting a unique challenge in mid-career recruitment where the improved quality of resumes makes AI detection itself difficult. Q6. What difficulties do you face when dealing with application documents suspected to be AI-generated? (Multiple answers) The most common concern is 'there are no internal standards or rules for rejecting based on AI generation' (Overall 35.0%, Mid-career 35.5%, New graduate 44.0%). This is followed by 'don't know how to distinguish if it is AI-generated' (Overall 30.6%, Mid-career 35.9%, New graduate 39.3%). New graduate recruiters have deep concerns across all items, and 'not troubled' is significantly lower for new graduates (16.4%) compared to mid-career (27.4%). It seems that mid-career recruiters have certain countermeasures since they can judge based on track records and skills. Q7. How are you currently dealing with application documents suspected to be AI-generated? (Multiple answers) The top response overall is 'lowering the evaluation if the possibility of AI generation is high' at 31.8%. This is even higher among new graduate recruiters at 42.9%, showing a strict evaluation stance towards AI documents. In contrast, the most common response for mid-career recruiters is 'digging deeper into skills and experience during interviews to directly verify true ability' at 34.7%, an approach that aligns with the characteristics of mid-career recruitment by verifying ability in interviews rather than relying on documents. 'Abolished or reduced document screening and increased casual interviews' also reached 15.7% overall, indicating a movement to review document screening itself. Q8. Does your company have rules or guidelines regarding the use of AI in recruitment selection? Only 14.9% have 'clear rules that are well-known to recruiters'. The total for 'no rules/left to the person in charge' and 'don't know' reaches 38.6%, showing that many companies are using AI without keeping up with guideline development. Among new graduate recruiters, 'rules exist but have not permeated the front lines' is high at 38.9%, highlighting the issue of rules becoming a dead letter. Q9. Do you personally feel resistance to leaving recruitment selection to AI? The AI-tolerant group (want to actively leave it to AI + acceptable if it's a supplementary role) is 43.7% overall, 52.9% for mid-career, and 61.1% for new graduates. New graduate recruiters have the highest AI tolerance, and in particular, 'acceptable if it's a supplementary role' is 44.4% for new graduates, significantly higher than 33.1% for mid-career. On the other hand, 'AI should not be used in recruitment in general' remains at 13.1% overall, showing that resistance to complete reliance on AI is still strong. Q10. As a recruiter, which tasks do you think cannot be replaced by AI? (Multiple answers) The top task considered irreplaceable by AI overall is 'designing recruitment standards and evaluation criteria' at 30.0%, followed by 'negotiating and closing offer conditions' at 28.8%, and 'digging deeper into skills and experience and evaluating true ability in interviews' at 27.8%. For mid-career recruiters, 'assessing the candidate's personality and cultural fit' is the highest at 38.3%, showing a particularly strong demand for human involvement in tasks that directly assess individual aptitude. 'I think AI will be able to replace all tasks' remains at 17.3% overall, 14.1% for mid-career, and 11.6% for new graduates, indicating that trust in the human role in recruitment remains strong. Discussion This survey revealed that the 'AI vs AI' structure in the recruitment scene is rapidly becoming a reality. In the chain where applicants create documents with AI and companies screen and judge with AI, recruiters are facing new challenges. In particular, this survey highlighted the 'too perfect problem', which has been overlooked in previous surveys. 56.6% of mid-career recruiters and 41.6% of new graduate recruiters pointed out that the problem with AI documents is not that they are 'thin on content' but that they are 'too perfect', indicating that the recruitment scene is facing a common difficulty regardless of the recruiter type. Just as an exam where everyone gets a high score fails to differentiate, the homogenization of document quality is causing recruiters to lose the opportunity to evaluate candidates based on their true abilities. In response to this situation, the recruitment scene is currently exploring countermeasures mainly in two directions. One is 'strengthening deep dives in interviews' (mid-career 34.7%), and the other is 'lowering evaluations' (new graduates 42.9%), which are symptomatic treatments. However, both are responses left to individual judgment, and the fact that over 30% of the total voiced that 'there are no internal rules' or 'don't know how to distinguish' indicates an urgent need for governance design as an organization. Furthermore, the reality that only 14.9% have 'clear rules that are well-known' suggests that companies face similar challenges not only in dealing with AI documents but also in their own use of AI for recruitment. The current state of the recruitment market is that both those who use AI and those who are subjected to it are operating based on front-line judgments without established rules. What will be important in future recruitment is not only to establish countermeasures for AI-generated documents but also to step into 'designing a selection experience that cannot be replicated by AI'. The fact that over 80% of recruiters answered that they 'do not think AI can replace all tasks' shows that confidence in the human role in recruitment is still strong. Assessing cultural fit, negotiating offers, and building trusting relationships with candidates are areas where AI is weakest, and where recruiters should focus their time and energy. In response to these changes in the recruitment scene, our company is working to strengthen recruitment capabilities in the AI era through AI utilization support directly linked to practical work, and reskilling and implementation support for companies and recruiters. We will continue to contribute to creating an environment where both recruiters and companies can make appropriate decisions by continuously surveying and disseminating information on the social implementation of generative AI in the recruitment market. Survey Overview Survey Period: May 20 to May 24, 2026 Survey Target: Male and female recruiters and HR personnel aged 25 to 69 nationwide Number of Subjects: 503 *Recruiters may hold concurrent roles in multiple areas. The breakdown is 248 mid-career recruiters, 275 new graduate recruiters, 165 part-time/arubaito recruiters, and 113 outsourcing/freelance recruiters. Survey Method: Internet Research / Conducting Agency: SHIFT AI Inc. <> When quoting or reprinting this release, we kindly ask that you clearly state our credit (e.g., Survey by SHIFT AI). [Company Overview] Under the mission of 'Making Japan an AI advanced country', SHIFT AI Inc. operates the generative AI learning community 'SHIFT AI', which has the No. 1 number of users* and allows people to learn about the business application of AI technologies, including generative AI. With over 40,000 members, we promote the development of AI talent in a wide range of fields, such as the reskilling support service 'SHIFT AI for Biz' for corporations and 'SHIFT AI for School' for educational institutions. Furthermore, through the operation of our proprietary media 'SHIFT AI Times', information dissemination, training, and events, we support the growth of individuals and organizations and accelerate the utilization of AI throughout Japan. Company Name: SHIFT AI Inc. Location: Shibuya Scramble Square, 2-24-12 Shibuya, Shibuya-ku, Tokyo Representative: Shota Kiuchi, CEO Established: March 18, 2022 Capital: 83 million yen (including capital reserve) Business Activities: Consulting / Community Operation / YouTube Channel Operation / School Operation URL: https://shift-ai.co.jp/ *No. 1 number of users Survey by GMO Research & AI, Inc. ■Survey Item / Cumulative number of registered users at the time of the survey (February 2025) ■Survey Target / Community services operated by companies that provide lectures on AI utilization for business purposes, such as AI utilization case studies and practical know-how. Internet communities that do not provide lectures, communities operated by individuals, and community services for non-business purposes are excluded. [About CEO Shota Kiuchi] Shota Kiuchi CEO of SHIFT AI Inc. / Councilor of the Generative AI Utilization Promotion Association / AI Utilization Advisor for GMO AI & Web3, Inc. / AI Utilization Advisor for GMO AI & Robotics Shoji 150,000 followers (As of May 2026) Disseminates information on generative AI under the theme of 'Making Japan an AI advanced country'. URL: https://x.com/shota7180