A vast syllabus spanning the breadth of human civilization, a selection process that stretches across an entire year, and over ten lakh hopefuls competing for barely a thousand posts already — these make the UPSC Civil Service Examination one of the most challenging and unpredictable tests in the world.
This year's Prelims, held on May 24, only deepened that reputation, with aspirants and educators alike calling it among the most brutally difficult papers in recent memory.
And yet, an up-and-coming AI learning platform claims to have seen it coming.
UPSC — not so unpredictable?
EdMe, a Thiruvananthapuram-based AI learning platform, announced in a recent blog post that it predicted topic areas of roughly 85 per cent of the current affairs question topics, and around 63 per cent of question topics overall in the UPSC CSE 2026, weeks before the exam was held.
The platform, supported by the Keralam Start Up Mission, has already tied up with the Keralam State Civil Service Academy to help aspirants prepare better.
Speaking to EdexLive, Simeon Samuel, one of the company’s co-founders, explains that the findings were a result of the platform’s model. “We trained EdMe’s algorithm with the last 15 years’ question papers and current affairs information from guidebooks and news reports from 2020.”
Weeks ahead of the UPSC CSE Prelims exam, the start-up laid out the platform's methodology for identifying structural patterns in UPSC question-setting in a provocatively-titled blog post, “UPSC Is Not Random”.
It claimed to have identified structural patterns in UPSC question-setting, patterns that the company argues persist even as the Commission shifts formats, difficulty levels, and topic emphasis from year to year.
“We identified topics based on predictability, and organised them as three tiers: high priority, medium priority, low priority,” Samuel says.
What the numbers say
On May 29, EdMe released its formal predictions for the 2026 Prelims to users of its web application, in another blog post titled, “Prelims 2026 was surprising, yet predictable.”
The company claimed that its model correctly anticipated the 63 out of 74 current affairs question topics in the paper — approximately 85 per cent. In the overall exam, the model predicted question topics with 63 per cent accuracy.
“We found subjects like Indian Economy, Science & Technology, and Environment, which are based on deep structural and institutional policy developments, were more predictable,” Samuel notes.
With repeated training, the algorithm helped spot trends that only seasoned teaching faculty and mentors could, he says.
Thinking like a topper
Through this platform, Samuel hopes to help aspirants make their preparation more focused and prevent them feeling overwhelmed by the exhaustive syllabus. “People spend years trying to crack the CSE, but often miss out just by some percentage. We want to help them narrow that gap by helping them prepare smarter,” he says.
Through tools like visual aids, flowcharts, annotations, detailed explanation notes and predictability, an aspirant preparing using EdMe would be able to “think like a topper”, he adds.
He further warns that the platform does not always produce guaranteed predictions. “UPSC CSE still rests on conceptual understanding of the topics, so we can’t yet say that there’s a direct translation of topics. At this point, we can identify the topics that are likely to carry more weightage,” he clarifies.
With 5,000 active users, Samuel states that the company is trying to expand to other centres, and expand EdMe’s capabilities to include UPSC CSE Mains, JEE, NEET, and State Public Service Commission exams.