How Ujiyari Actually Performed Against UPSC 2026: A Self-Audit With the Numbers, Including the Bad Ones

Every current-affairs site claims hits after an exam. It is the easiest claim in the industry to make and the hardest to check, because the claim is almost always made backwards: the paper comes out, somebody searches the archive, finds something adjacent, and posts a screenshot.

This post does the opposite. It audits Ujiyari, which is my own site, against UPSC CSE Prelims 2026 and Mains 2026, using a test that can fail. The results are mixed. Ujiyari did well against Mains and poorly against Prelims, and one structural fact explains both.

1. The method, stated before the results

Three rules, fixed in advance.

A link only counts if it was published before the exam. Ujiyari's URLs carry their own dates: a daily note lives at /daily/2026/05/14/... and an editorial at /editorials/2026/08/.... For month-only editorial URLs, where the URL alone cannot settle whether a piece came before or after an exam in the same month, I read the lastmod date from Ujiyari's own sitemap rather than assuming. Anything published on or after the paper is worth nothing to a candidate who has already written it, and is counted as a miss.

Topic coverage is not the same as a hit. Having written about the National Quantum Mission does not mean a reader could answer a specific Prelims question about it. The strongest honest claim is "the material was available in advance", not "we predicted the question".

Where the method is weak, the weakness is stated. Two places matter, and both cut against the flattering reading. They are set out in section 5.

The snapshot used throughout is Ujiyari's sitemap as of 24 August 2026, covering 4,652 published URLs.

2. The fact that explains everything: the archive begins on 1 January 2026

Before any comparison, one thing has to be on the table.

Ujiyari has no content published before 1 January 2026. The daily archive runs from /daily/2026/01/01/ onwards. The editorials archive starts the same month. The yearly compilation exists for 2026 and no earlier year. Every monthly compilation is a 2026 month.

That is not a criticism of the site. It is simply how old it is. But it decides the entire result below, because the two exams sit on opposite sides of it.

• Prelims 2026 was written on 24 May 2026. Ujiyari's archive covered roughly 4.7 months before that date. UPSC's Prelims current-affairs window is conventionally taken as twelve to eighteen months. Most of the window pre-dates the site's existence.
• Mains 2026 was written from 21 to 23 August 2026. By then the archive covered nearly eight months, and Mains rewards recent, dated developments far more heavily than Prelims does.

Hold that in mind. The rest is largely a consequence of it.

3. Prelims 2026: weak, and honestly so

I took the current-affairs-anchored questions from our own Prelims 2026 GS1 answer key, extracted the named entity in each, and searched the archive for material published before 24 May 2026.

Result: 5 of 24 named items had dated pre-exam coverage. That is about 21 per cent.

What was covered in advance:

What was not covered at all: UNESCO Global Geoparks, Madhav Tiger Reserve and the Sakhya Sagar Ramsar site, REDD+ and Plan Vivo certification, LT-LEDS and BUR-4, Rainfed Area Development under NMSA, INTERPOL notice categories, the India-Germany visit, the Nobel Prize in Physics 2025, the Bharat Forecast System, the BAFTA result, and the Revamped Rashtriya Gram Swaraj Abhiyan.

Look at that list and the pattern is obvious. The Bharat Forecast System was launched in May 2025. The Nobel Physics prize was announced in October 2025. The Genome India Project ran from 2020. Deep Ocean Mission and the National Quantum Mission were sanctioned years earlier. A site whose first article is dated 1 January 2026 was never going to have covered them as news, because they were not news in its lifetime.

There is a real lesson in this for anyone using a young current-affairs site for Prelims: recency is not the constraint that binds in Prelims, coverage depth is. A daily site is at its weakest precisely where UPSC reaches back eighteen months.

Caveat, and it cuts in our favour, so treat it carefully. This test matches on URL slugs. If a topic was discussed inside an article whose slug does not name it, the test records a miss that was not one. So 21 per cent should be read as a floor, not a measurement. I am not going to guess how much higher the true figure is, because guessing is the thing this post exists to avoid.

4. Mains 2026: strong, and the dates hold up

The Mains comparison is much better evidenced, because the five solved-paper pages on BharatNotes already carry a hand-picked Ujiyari link under each question where one exists. That gives a per-question record to date-check.

Across the five papers, 113 Ujiyari links are attached to 70 of the 87 questions and parts. Checked against each paper's own exam date:

Questions with at least one link provably published before the exam: 57 of 87, or 66 per cent.

By paper:

The distribution is as informative as the total. GS1, GS2 and GS3 sit at 80 per cent each. GS4 and Essay are low, and that is correct rather than a failure: thirteen of GS4's items are ethics theory, paternalism and beneficence, duty against charity, Gandhian trusteeship, a line from Henry IV, and six of the eight Essay topics are metaphors. There is nothing dated for a current-affairs site to have covered. A site that claimed hits there would be manufacturing them.

One near miss deserves naming, because it is the sort of thing that usually gets quietly reclassified as a hit. GS2 asked a question focused specifically on IPMDA, the Indo-Pacific Partnership for Maritime Domain Awareness. Ujiyari published an editorial on IPMDA and maritime domain awareness dated 22 August 2026, which is the day GS2 was written. It is the single link in the whole set that is not provably pre-exam. It goes in the miss column.

5. What this does not prove, stated plainly

Two weaknesses, both of which make the Mains number look better than a blind test would.

Selection bias. I built those links by reading each Mains question and then searching Ujiyari for something relevant. That measures how often a relevant piece could be found when someone went looking. It does not measure what a candidate reading Ujiyari daily would have absorbed without knowing the questions. A blind test, where the links were fixed before the paper, would be the real trial, and it is not available retrospectively.

The date check is the part that survives this objection. I could bias which links were found; I could not bias when they were published. That 88 of 89 dated links pre-date their exam says the material genuinely sat in the archive beforehand.

Coverage is not marks. Nothing here shows that a reader scored better. Between reading a piece and writing a good answer sit recall, structure, word discipline and practice, none of which a daily note supplies. Any site telling you its coverage produced ranks is selling you something.

6. Where Ujiyari is concretely weak

Compiled from the same audit, not from impressions.

• No pre-2026 backfill. The single largest gap, and the direct cause of the Prelims result. There is no retrospective 2024 or 2025 archive.
• Ethics has almost no purchase. GS4 theory is 42 per cent linked and most of those links are adjacent rather than on point. Nothing exists on medical ethics and consent, which is where the 2026 case study on an unconsented hysterectomy lived.
• Institutional gaps. No coverage of the NCSC or NCST, both of which GS2 asked about directly, and nothing on separation of powers.
• Monthly compilations lag. On the 24 August snapshot, monthly compilations exist for January to June 2026. July has not appeared, and the April compilation has a Prelims section but no Mains or interview section, where other months have all three.
• Prelims-shaped output is thin. 1,873 daily notes and 1,469 editorials against 9 quiz pages and 10 PYQ pages. The site is built for the reader who writes answers, not the reader drilling MCQs, which is consistent with it doing better on Mains.

7. The 2026 interview: the stage that best fits what Ujiyari is

The calendar first, since prediction without a calendar is fantasy. Mains 2026 was written between 21 and 30 August 2026. On the ordinary pattern, results follow in roughly three months, personality tests run January to April 2027, most commonly February and March, and the final result lands around April 2027. UPSC publishes the schedule in phases with e-summon letters; nothing below should be taken as a substitute for that.

Here is the structural argument, and it is the strongest genuinely positive thing in this audit.

The interview's current-affairs window is the narrowest and most recent of the three stages. A board asks about what has happened lately, about the candidate's home state and DAF, and about issues on which an administrator would need a view. The relevant window for a February 2027 board is roughly September 2026 to January 2027.

That window falls entirely inside Ujiyari's archive. The very thing that crippled it for Prelims 2026, having no history, is irrelevant here, because the interview does not reach back. On this reasoning I expect Ujiyari to be at its most useful of any stage so far, and the site already publishes an interview-oriented section in each monthly compilation.

I am stating that as an expectation with a reason, not a result. The honest test is one I cannot run: it would need candidates to report what they were asked, and boards are not a published paper.

What would have to be true for it to work, and what I will therefore watch:

• monthly compilations must actually appear for September 2026 onwards, given that July's is still missing;
• the interview sections need state-wise and DAF-shaped material, not a rehash of the daily notes;
• the September-to-January window needs the same density the January-to-August window had, roughly 250 daily notes a month.

8. The 2027 cycle: the archive gap closes

UPSC released its 2027 calendar on 20 May 2026. CSE 2027 notification: 13 January 2027. Applications close 2 February 2027. Prelims: 23 May 2027. Mains: from 20 August 2027, over five days.

The single most consequential change is arithmetic rather than editorial. By 23 May 2027, Ujiyari's archive will run from January 2026 to May 2027, roughly seventeen months. That covers the conventional Prelims window in full, for the first time. The structural handicap that produced this year's 21 per cent simply expires.

That is a prediction with a real mechanism behind it, and it is falsifiable: if the 2027 Prelims audit, run by this same method, does not land far above 21 per cent, then the problem was never the archive's age and this analysis was wrong.

What is already on the calendar and therefore near-certain to be examinable, all of it verified rather than guessed:

• Census 2027. Houselisting ran April to September 2026; population enumeration is in February 2027 with a reference moment of 00:00 on 1 March 2027. It carries caste enumeration, the first since 1931. This is the largest single administrative event of the cycle and it sits three months before Prelims.
• Delimitation and women's reservation. The Constitution (131st Amendment) Bill was defeated in the Lok Sabha in April 2026, and the Delimitation Bill 2026 lapsed with it. The commencement of the 106th Amendment therefore remains tied to a census and a delimitation, and Census 2027 revives the whole question.
• India's BRICS chairmanship, with the summit in New Delhi in September 2026, produces outcome documents inside the 2027 window.
• India holds the FATF vice-presidency from July 2026 to June 2027.
• The Kalpakkam PFBR, critical on 6 April 2026, was projected for commercial generation around September 2026. Grid connection and power ascension fall in the window.
• The Solid Waste Management Rules 2026 require every legacy dumpsite to be mapped and assessed by 31 October 2026, a dated compliance milestone.

Predicting the themes of 2027 from the 2026 papers is a weaker exercise and I will keep it short and label it as judgement rather than fact. The 2026 papers rewarded three things: a dated fact attached to a settled framework, a directive verb answered as asked, and precision about what a thing is and how it works. Nothing about the 2026 papers suggests any of that is about to change, which is a mild claim and the only one the evidence supports.

9. What I am committing to, so this is checkable next year

• Run this same audit after Prelims 2027 and Mains 2027, by the same method, and publish it whatever it says.
• Treat 21 per cent as the number to beat on Prelims, and 66 per cent as the number to beat on Mains, with the selection-bias caveat carried forward and, if possible, a set of links fixed before the paper so the next Mains test can be blind.
• Close the named gaps: medical ethics and consent, the NCSC and NCST, separation of powers.
• Publish the missing monthly compilations and keep them current, since the interview argument in section 7 depends on it.

10. How to check me

Everything above can be reproduced. Ujiyari's URLs carry their publication dates, its sitemap carries lastmod for editorials, and the exam dates are public. Take any claim of coverage on any site, including this one, and ask the only question that matters: was it published before the paper? If a site cannot answer that from its own URLs, the claim is not checkable, and an unfalsifiable claim of success is worth exactly nothing.

The finding I would keep from this exercise is the least flattering one. A young daily current-affairs site is structurally strong for Mains and for the interview, and structurally weak for Prelims, and no amount of editorial effort in the present fixes a missing past. The only cure is time, which is the one input nobody can buy.

Method note: figures drawn from Ujiyari's sitemap snapshot of 24 August 2026 (4,652 URLs) and from the Mains 2026 solved-paper data on BharatNotes. Prelims matching is slug-level and therefore a floor rather than a measurement. If any figure here is wrong, it will be corrected on this page with a dated note saying what was wrong and what is right.