The future of competitive exams in an AI world

For decades, competitive examinations such as JEE and NEET have shaped the ambitions of Indian families. For many students, these examinations are an important academic goal and a potential route to higher education and socio-economic mobility. However, the skills they reward were largely defined in a world where solving difficult problems quickly was a distinctive ability.

That world is changing. AI models can now solve a growing number of physics, chemistry and mathematics problems similar to those found in entrance examinations, often within seconds and with step-by-step explanations. Questions that once required considerable individual effort can increasingly be approached using digital tools. This development has prompted educators and examination bodies to consider what high-stakes assessments should measure.

The response need not be to make examinations harder in the same way. Longer calculations and more complex variations of familiar problems may encourage additional practice without necessarily measuring deeper understanding. A broader approach could place greater emphasis on reasoning, conceptual clarity and the application of knowledge.

Reasoning involves asking students why, not only what. An examination could present a solution and ask students to identify a flawed assumption, or offer two approaches and ask which remains valid under changed conditions. Conceptual understanding involves connecting ideas across chapters and subjects, such as relating a chemistry concept to a biological process rather than recalling a formula in isolation. Application involves presenting unfamiliar situations in which students must understand and frame the problem before choosing a method.

Judgement is also important. As students increasingly encounter AI tools in their learning, they need to verify outputs, identify errors and recognise the limitations of automated answers. Future assessments could include tasks that ask students to evaluate an AI-generated explanation or identify mistakes in a worked solution.

Technology may also influence how examinations are prepared and conducted. AI-assisted question generation, subject to expert review and difficulty calibration, could help examination bodies develop larger pools of comparable questions. These pools may support the preparation of different question papers for multi-session examinations, including formats such as JEE Main. However, maintaining comparability across papers would require rigorous validation, transparent scoring methods, examination security and clear rules governing the use of technology.

Another approach is computer-adaptive testing, in which a candidate’s responses influence the difficulty or selection of subsequent questions. The GMAT has used adaptive testing in candidate assessment. Such systems require a carefully developed question bank, reliable scoring methods and appropriate validation to ensure that results are meaningful and comparable across test takers.

Multiple examination attempts may also give students additional opportunities to demonstrate their preparation, reducing the influence of a single difficult examination day. Some higher education institutions that accept SAT scores consider the highest section scores from different attempts under their score-combination policies. Any comparable approach would need clear eligibility criteria, consistent scoring and safeguards against misuse.

However, examination reform alone will not be sufficient.

Students generally prepare according to what assessments reward. If entrance examinations place greater emphasis on reasoning and application, schools and coaching institutions may need to adapt their teaching methods. This could include concept-based learning, regular formative assessments and practice designed to identify learning gaps early, rather than relying primarily on repeated exercises from previous papers. School boards could also incorporate assessment frameworks that develop these skills before Class 12.

Equity must remain part of the discussion. Students in Tier II and Tier III towns may have different levels of access to specialised coaching and learning resources. Assessments that value understanding and application could broaden the ways in which students demonstrate their abilities. However, this would depend on equitable access to quality teaching, suitable learning materials and adequate preparation opportunities.

Teachers will remain central to this process. AI can assist with question development and performance analysis, but teachers help students understand concepts, develop curiosity and work through unfamiliar problems. Professional development can help teachers incorporate reasoning-led activities into classroom discussions and assessments.

(The author is Founder of Teevra EduTech)

ABOUT THE AUTHOR

Vivek Varshney