Students face AI-free tests for AI-shaped jobs—what did 23 interviews reveal?
Engineering students and recent graduates in India described using AI to prepare for closed placement screens, clear coursework and meet AI-first workplace expectations. The Chennai-heavy qualitative sample exposes a tension, not a national prevalence estimate.
The short version
- What happened
- Researchers interviewed 23 AI-using students and 2026 graduates from Indian engineering and computing programmes: 13 students and 10 graduates, with two women in the sample.
- Who it affects
- Students may spend scarce time mastering tests that poorly match AI-assisted work, while employers receive weaker signals about independent reasoning and responsible tool use. Clearer policies and mixed AI-free and AI-permitted assessment could make expectations visible.
- What remains unproven
- This is an unreviewed qualitative study of 23 AI-using participants, not a representative survey of Indian engineering education.
Keep in mind
This is an unreviewed qualitative study of 23 AI-using participants, not a representative survey of Indian engineering education.
The sample clustered in established institutions, mostly in Chennai, included only two women and excluded several less-resourced settings.
Accounts were self-reported; graduates recalled earlier college experiences, and Tamil-English translation may have lost nuance.
Faculty, administrators, recruiters and managers were not interviewed, and no funding or competing-interest statement is reported.
Source analysis · 1 linked source. Primary analysed source date: 21 October 2026.
Read the full report · 3 min The evidence, methods, implications and limitations
Students described a mismatch between selection and work
The study interviewed 23 current students and recent graduates between September and October 2026. Thirteen were still studying and 10 had graduated; two were women. Participants came from engineering and computing programmes spanning several institution types, but most studied in Chennai. Seventeen interviews were in English and six partly or mainly in Tamil, usually lasting 30 to 45 minutes.
Across a reflexive thematic analysis, participants said campus placement gates—CGPA cut-offs, aptitude tests, coding screens and interviews—usually expected unaided performance. Yet graduates and interns described workplaces where AI use was normal or pressed by tighter deadlines. Many therefore used AI as a coach to prepare for tests that did not allow it, while remaining uncertain which foundational skills employers actually expected them to retain.[1]
The interviews illuminate a problem; they do not count it
Participants also distinguished passing or 'clearing' coursework from learning. AI helped explain weak teaching and fill gaps, but it could also produce assignments, rehearse predictable exam answers or become the first response to a task. Faculty rules were often unstated, and students described learning acceptable use from peers. Some graduates worried that work they could not fully claim as their own left them less able to defend a project in an interview or solve a problem unaided.
These are self-reported accounts, interpreted by the researchers rather than frequency estimates. Purposive and referral recruitment excluded rural, less-resourced and regional-language-medium institutions; only two participants were women, and the team did not interview faculty, recruiters or managers. The first author attended one represented institution and works as an AI consultant, a position addressed through reflexive notes and co-author consultation. Broader surveys and trials of AI-permitted assessments would test how far the pattern travels.[1]
The distinction that matters
The three gates participants described
AI use changed across the path from coursework to employment.
Coursework
Use AI to explain, draft or clear assignments
Grades may mix learning with assistance
Placement
Prepare with AI, then sit mostly AI-free screens
Selection tests unaided performance
Work
Use AI under delivery pressure
Employers may reward assisted speed