Is UPSC Prelims Getting Harder?
Short answer: yes and no. From 2015 to 2016, easy-question share fell from 43% to 35% — the paper genuinely got harder. But 2025 rebounded to 43% easy. Hard-question share has stayed in a narrow 2–10% band throughout. Full year-by-year below.
The actual trajectory
| Year | Easy | Moderate | Hard | Total | Easy % | Hard % |
|---|---|---|---|---|---|---|
| 2015 | 38 | 45 | 6 | 89 | 43% | 7% |
| 2016 | 45 | 40 | 9 | 94 | 48% | 10% |
| 2017 | 42 | 45 | 6 | 93 | 45% | 6% |
| 2018 | 38 | 46 | 8 | 92 | 41% | 9% |
| 2019 | 33 | 48 | 7 | 88 | 38% | 8% |
| 2020 | 39 | 44 | 2 | 85 | 46% | 2% |
| 2021 | 36 | 48 | 6 | 90 | 40% | 7% |
| 2022 | 33 | 51 | 7 | 91 | 36% | 8% |
| 2023 | 32 | 51 | 8 | 91 | 35% | 9% |
| 2024 | 41 | 44 | 9 | 94 | 44% | 10% |
| 2025 | 40 | 46 | 7 | 93 | 43% | 8% |
The common narrative — "UPSC keeps getting harder every year" — is half right. The data shows a clear hardening trend from 2015 through about 2016, then a slight easing in 2025. Not a straight line.
Easy-question share peaked at 48% in 2016, bottomed at 35% in 2023, and recovered to 43% in 2025. Net 2015→2025: a 0-point increase in easy share.
What actually changed
The real shift isn't in hard questions — hard share has hovered in a narrow 2–10% band every year. What shifted is the moderate tier: questions that aren't outright hard but require careful elimination, trap-awareness, and second-order thinking. Moderate climbed from 51% in 2015 to 56% at its peak.
In practical terms: UPSC hasn't been adding obscure facts. It's been adding traps. The well-prepared aspirant who could coast through 40+ easy questions in 2015 is now navigating 50+ moderate questions where option elimination is the differentiator — and where our trap analysis actually matters.
What this means for your prep
Don't over-index on impossible questions.Hard share is small and structurally stable. If you can't crack 7questions a year, you're not alone — most aspirants can't. Leave them.
Do over-index on the moderate tier.That's where Prelims is actually decided. 50 questions a year, each with a trap designed to catch aspirants who have memorised the fact but not practised elimination. This is exactly what subject-wise PYQ practice with trap analysis is built for.
What this doesn't tell you
Difficulty is not the same as predictability. A year can have more "moderate" questions yet still be more predictable if the topic mix is conventional. For the topic-mix cut, see most-repeated topics. For mock calibration against these distributions, see our mocks — each labelled with its intended difficulty spread.
Difficulty labels (easy / moderate / hard) are reviewed by UPSC aspirants and the PYQLabs community. They reflect expected accuracy for a well-prepared aspirant, not raw fact obscurity.