Ishvaram Research
What 351,000 Conversations With an AI Astrologer Reveal About India's Real Worries
The headline findings
Between 2026-06-02 and 2026-07-20 — 7 complete weeks — 3,75,731people used Ishvaram's free Vedic tools, started 3,51,591 conversations with its AI Jyotishi, and computed 3,06,171birth charts. The platform's intent classifier could categorize 1,34,834 of the first questions seekers asked. Four patterns stand out, and each contradicts a common assumption about how India uses astrology.
1. Career beats marriage — by a wide margin
The stereotype says Indians consult astrologers about marriage. The classified questions say the modern seeker's first anxiety is work: career questions outnumber marriage questions roughly three to two, and doshas — the fear that something in the chart itself is blocking life — run a close third.
| What the first question was about | Questions | Share |
|---|---|---|
| Career and work | 52,830 | 39.2% |
| Marriage and relationships | 32,531 | 24.1% |
| Doshas and remedies | 29,207 | 21.7% |
| Full kundali reading | 5,341 | 4% |
| Muhurat / timing (when should I…) | 5,194 | 3.9% |
| Health | 3,808 | 2.8% |
| Money and finance | 2,996 | 2.2% |
| Just wanted to talk now | 2,927 | 2.2% |
2. India consults the stars at night
Conversation starts and chart computations peak between 22:00–00:00 IST, at ~5× the volume of the 04:00–05:00 IST trough. There is a secondary plateau across 11:00–16:00 IST, but the modal moment a seeker asks about their future is after the household has gone quiet — not the morning puja slot the stereotype would predict.
3. Panchang is a daytime, Saturday habit
Timing lookups — panchang, choghadiya, muhurat, and rahu kaal — invert the night-time pattern: they peak at midday, and their weekly rhythm follows the traditional planetary week. 20% of this timing traffic already happens on Hindi-language pages.
| Rank | Day | Note |
|---|---|---|
| #1 | Saturday | Shani's day |
| #2 | Sunday | — |
| #3 | Thursday | Guru's day |
| #7 | Friday | quietest |
4. Mangal Dosh is India's biggest dosha worry
Across 6,78,592 concern-carrying consultation events, six classical doshas account for 99.2% of all dosha anxiety. Mangal Dosh leads, but the strength of Pitra Dosh — ancestral karma — in second place is the least-reported finding in this table.
| Dosha | Consultation events | Share |
|---|---|---|
| Mangal Dosh | 1,68,561 | 25% |
| Pitra Dosh | 1,58,173 | 23% |
| Shani Dosh / Sade Sati | 1,26,269 | 19% |
| Rahu Dosh | 89,492 | 13% |
| Kaal Sarp Dosh | 70,696 | 10% |
| Ketu Dosh | 60,142 | 9% |
What each number counts
Definitions matter when a figure is quoted away from its table, so each counted quantity is stated here exactly once. A unique visitor is one distinct browser reaching the platform inside the window (3,75,731). A conversation started is one chat session opened with the AI Jyotishi (3,51,591) — a person may open several. A kundali computed is one birth chart the engine actually calculated (3,06,171), which is why it trails conversations: not every visitor supplies birth details. A classified first question is the opening question of a conversation that the production classifier could categorise (1,34,834); every intent percentage on this page divides by that figure, never by conversations.
One caveat about the window itself, since it bounds every claim above: these 7weeks are a single seasonal slice, deliberately chosen as a clean, uninterrupted serving period rather than a rolling series. India's concern mix moves with the festival calendar, exam and results cycles, and the wedding season, so treat the ordering of intents as a finding about this period rather than a permanent ranking. We expect to repeat the measurement rather than extrapolate from it.
Methodology and privacy
All counts come from Ishvaram's first-party, consent-scoped product analytics for the complete IST window 2026-06-02 to 2026-07-20, aggregated on 2026-07-30. Question categories are assigned by the platform's production intent classifier at conversation time; first questions the classifier could not categorize are excluded from percentage bases, so shares are of classified questions. Dosha shares count consultation events rather than unique people. All times are Indian Standard Time. No conversation text was accessed by any person or model in preparing these aggregates, and nothing in this study can identify an individual. Limitations worth stating plainly: this is one platform's self-selected audience, and the classifier categorized about a third of first questions — the mix of the remainder is unknown.
How the underlying platform computes its astrology — the Swiss Ephemeris engine, Lahiri Ayanamsa, and the review process — is documented on our methodology page. For additional cuts of this data, interview requests, or corrections, reach us via ishvaram.com; corrections are applied to this page directly with a dated note.
How to cite this study
Journalists, researchers and encyclopaedia editors are welcome to reproduce any figure here without asking, provided the source is attributed and the window is named — a percentage quoted without its 7-week period invites misreading. The suggested reference is: Ishvaram Research, “What 351,000 Conversations With an AI Astrologer Reveal About India's Real Worries”, published 2026-07-30, covering 2026-06-02 to 2026-07-20.
Worth noting what kind of evidence this is, because it is easy to conflate with polling. Surveys record what respondents say they would ask; these are observational aggregates of what people actually typed unprompted, before anyone offered them categories to choose between. That removes recall bias and social desirability — nobody is performing an answer for an interviewer — while introducing the trade-off named above: the audience selected itself by arriving here. Neither instrument is a substitute for the other, and the honest reading treats them as complementary evidence about the same underlying demand.
On data availability: everything released here is aggregate. Conversation-level records are never shared, licensed or sold, and no extract exists that could be re-identified, so requests are answered with further aggregate breakdowns rather than underlying rows. This URL is the canonical version and supersedes any mirror or screenshot; where a figure is later revised, the change is logged inline with its date rather than silently overwritten, so a citation made today remains checkable against what it originally said.
About this study
How was this study measured?
All figures come from Ishvaram's first-party product analytics between 2 June and 20 July 2026 (IST). Question categories were assigned by the platform's production intent classifier at conversation time; percentages are calculated over the 134,834 first questions the classifier could categorize. No conversation text was read by any person or model to produce these aggregates.
Can journalists or researchers cite or republish these figures?
Yes. Cite 'Ishvaram Research, 2026' and link this page. For the underlying breakdowns, additional cuts of the data, or founder commentary, contact us through ishvaram.com — we can usually turn requests around within a day.
Does this data identify anyone?
No. Every number on this page is an aggregate count of machine-assigned categories across hundreds of thousands of sessions. The study contains no names, no birth details, no locations at individual level, and no conversation text.