Offbeat research
India's AI Paradox: 4-Source Triangulation
India's AI Paradox · Methodology Deep-Dive
Analysis by
Shashwat Ghosh
"Anthropic shows what Indians do with AI. OpenAI shows who does it and at what scale. Google/Ipsos shows what they think and feel. Microsoft shows where India sits on a normalised global basis. Together, they're the most complete picture of any country's AI adoption ever assembled."
Shashwat Ghosh, Co-Founder and Fractional CMO, Helix GTM Consulting
Why 4-Source Triangulation?
No single platform captures the full picture of a country's AI adoption
Source 1: Anthropic Economic Index, India Brief
What It Measures
- • Actual task-level usage patterns from Claude.ai conversations
- • Tasks mapped to O*NET occupation codes
- • Occupation concentration, autonomy delegation, productivity multipliers
- • Geographic distribution by Indian state
Data Parameters
- • Sample: ~58,000 Claude.ai conversations from India
- • Period: November 2025
- • Published: February 16, 2026
- • Context: Released at India AI Impact Summit
Caveats: Claude's user base skews technical and English-speaking. India-specific brief uses different methodology than global report (no randomised sample). Self-selected users, not representative of general population. O*NET mapping designed for US occupations: may not perfectly capture Indian job categories.
Source 2: OpenAI Signals, India
What It Measures
- • ChatGPT platform usage data (WAU, message volume)
- • Age demographics of Indian users
- • Feature-specific engagement (Codex, data analysis, voice)
- • Free vs paid tier patterns
Data Parameters
- • Scale: 100M+ weekly active users in India
- • Rank: #2 global market
- • Published: February 20, 2026
- • Context: Alongside Tata 100MW data centre, Mumbai/Bengaluru offices
Caveats: Data selectively presented alongside major commercial announcements (Tata deal, new offices, PM Modi meeting). No per-capita data provided. "Signals" is a PR-infused data release, not a peer-reviewed study. Free tier dominates India's 100M WAU: paid conversion unknown.
Source 3: Google/Ipsos "Our Life With AI" Survey
What It Measures
- • Self-reported AI usage, knowledge, and attitudes
- • Trust in AI accuracy, workplace impact, learning intent
- • Sentiment (promising vs concerning scales)
- • Cross-country comparison (21 countries)
Data Parameters
- • Sample: ~21,000 adults (18+), ~1,000 per country
- • Countries: 21 (including India)
- • Period: Sep 22 to Oct 10, 2025
- • Published: January 15, 2026
Caveats: Online sampling methodology inherently skews towards urban, educated, digitally connected respondents, especially significant in India where 500M+ lack regular internet access. ~1,000 respondents cannot represent 1.4 billion people. Self-reported usage often over-estimates actual behaviour. Google-sponsored survey may have framing bias towards positive AI narratives.
Source 4: Microsoft AI Diffusion Report (H2 2025)
What It Measures
- • Cross-platform AI penetration normalised per capita
- • Adjusted for OS/device market share, internet penetration, population
- • Not limited to Microsoft products
- • Country-level rankings across 100+ nations
Data Parameters
- • Coverage: 100+ countries
- • Method: Aggregated, anonymised telemetry
- • Published: January 8, 2026
- • India rank: #64 per-capita AI adoption
Caveats: Most methodologically rigorous for cross-country comparison, but telemetry still depends on Microsoft ecosystem footprint. Adjustments for OS market share and internet penetration are modelled, not directly observed. "AI adoption" definition may differ from other sources.
Source Comparison Matrix
| Dimension | Anthropic | OpenAI | Google/Ipsos | Microsoft |
|---|---|---|---|---|
| Data Type | Behavioural | Platform | Attitudinal | Telemetry |
| India Sample | ~58K convos | 100M+ WAU | ~1,000 | 100+ countries |
| Per-Capita Data | Yes (#101) | No | No | Yes (#64) |
| Task-Level Data | Yes | Partial | No | No |
| Age Demographics | No | Yes | No | No |
| Attitudes/Trust | Partial | No | Yes | No |
| Cross-Platform | No | No | Partial | Yes |
| State-Level Data | Yes | Yes | No | No |
The Convergence Principle
When 3-4 independent sources, using different methodologies, measuring different things, with different biases, point in the same direction, the signal is strong. We found 9 such convergences across these 4 platforms, documented in each section of this report.
Shashwat Ghosh
Co-Founder and Fractional CMO
4 Data Sources
Built by Shashwat Ghoshhttps://gtmhelix.com/ai-trendwatch/india-ai-4-platform-method.html