The shock started with CSAT 2022 — longer passages, denser arithmetic, no "sitter" questions. CSAT 2023 went further: only 14 reasoning questions, calculation-heavy maths, lengthy statement-based questions caused a record number of GS-qualifiers to fail CSAT. CSAT 2024 eased slightly but kept the calculation-heavy spirit. CSAT 2025 stayed moderate-tough.
CSAT failure rate — what the data shows
The pattern shift from 2022 onwards triggered a measurable surge in candidates failing the qualifying threshold despite clearing GS Paper-I. Approximate failure rates compiled from coaching post-mortems and aspirant survey data:
| Year | CSAT difficulty | Estimated CSAT failure among GS-qualifiers | Comment |
|---|---|---|---|
| 2019 | Easy | <15% | Comprehension alone cleared 66 |
| 2020 | Moderate | ~20% | 40 maths questions caused first wave of concern |
| 2021 | Moderate | ~25% | Pattern stabilising |
| 2022 | Tough | ~45% | Sharp jump — long passages, dense arithmetic |
| 2023 | Very tough | ~60–62% | Record number of GS qualifiers eliminated by CSAT |
| 2024 | Moderate | ~30% | Calculations heavy but cleaner numbers |
| 2025 | Moderate-tough | ~35% | Number System dominated; lengthy RC |
UPSC does not publish CSAT-specific qualifying counts, but Delhi HC pleadings in 2024–2025 referenced Expert Committee findings that confirmed the rising difficulty was "within Class X bounds" despite the visible failure spike.
What changed in 2022
Until 2021, most aspirants could clear 66 marks on comprehension alone. CSAT 2022 broke that pattern:
- 28 RC questions across 15 passages — including 9 questions from Level-3 inferential passages that hadn't appeared since 2018.
- Arithmetic questions used unfriendly numbers (no clean integers).
- Almost every section had multi-step questions instead of formula-plug-in. Coaching analysts counted fewer than 5 "sitter" maths questions in the entire paper.
- Statement-based questions in reasoning required two readings.
Many 2022 GS-qualifiers — including serious veterans — failed CSAT and lost the year.
What 2023 did
CSAT 2023 was widely called the toughest CSAT ever:
- Reasoning questions dropped to just 14.
- Maths questions were lengthy, calculation-intensive, and trap-laden.
- Comprehension passages became analytical, often requiring two readings.
- The paper triggered the Delhi HC writ petition (CW 4354/2025) challenging certain Paper-II questions as out-of-syllabus. The Expert Committee report upheld UPSC, finding all maths questions within Class X bounds. Petition dismissed.
- Among 5.83 lakh aspirants who appeared, only 14,624 cleared Prelims — a sharp ratio drop, much of which was attributed to CSAT.
What 2024 did
CSAT 2024 was noticeably easier than 2023 but still harder than the 2017–2021 era. Maths questions hit 35–36 (the highest count in years), but they used cleaner numbers. Comprehension passages were short and direct. Most candidates with steady prep cleared 66 comfortably.
What 2025 did
CSAT 2025 returned to moderate-tough. Number System alone yielded 25 questions; lengthy comprehension passages reappeared; a good attempt was pegged at 55–65 questions with high accuracy. The 2023-level shock did not repeat, but the 2019-level ease did not return either.
Why UPSC seems to be doing this
Three likely reasons, triangulated from official replies in court and Expert Committee notes:
- Filtering rote learners — UPSC wants policy-implementation talent, not exam-cramming machines.
- Aligning with mains demands — Mains rewards analytical comprehension; CSAT now demands the same.
- Levelling the field — a tougher CSAT actually helps strong readers, regardless of academic background. Engineers do not get the easy ride they once did.
CSE 2026 outlook
The CSE 2026 notification confirms no pattern change. Expect a paper closer to 2023–2025 difficulty than 2019. Plan for the harder scenario.
Mentor's takeaway
Don't prepare for CSAT 2026 assuming a 2019-level paper. Prepare for a 2023-level paper and you'll cruise through anything UPSC throws. The single biggest CSAT mistake of the last four years has been under-estimation.
BharatNotes