What this research found
Did the arrival of generative AI after 2022 widen the research productivity gap between India and China? Counts of AI-related publications drawn from OpenAlex for 2019 to 2024 show both countries growing strongly in the post-2022 period — India by 40.3% and China by 42.6% — leaving a differential of just 2.3 percentage points. An earlier pass built on download-capped samples had appeared to show India's output shrinking; re-querying OpenAlex for true annual totals reversed that conclusion.
- Measured on full population counts, India published 291,014 AI-related papers across the 2019–2021 baseline and 408,336 across 2022–2024, a rise of 40.3%. China went from 851,375 to 1,214,463 over the same periods, a rise of 42.6%, so the growth differential is only 2.3 percentage points.
- The scale gap, not the growth gap, is what separates the two. China's annual output stayed at roughly three times India's throughout, moving only from 3.10:1 in 2019 to 3.02:1 in 2024, with period ratios of 2.93:1 and 2.97:1.
- OpenAlex caps any single download at 10,000 records, and that cap understated true annual counts by factors ranging from 30x to 590x. Working from the capped sample produced the opposite headline — India down 1.3% against China up 10.3%, an 11.7 point gap — which vanished once counts were re-fetched using a year-grouped query that bypasses the limit.
- Within the 40,000-paper sample, papers with Chinese affiliations averaged 2.2 to 2.6 times the citations of Indian ones (126.2 against 56.2 in the later period), and the two countries moved in opposite directions on partnership: India's share of internationally co-authored papers rose 9.5 points to 48.2% while China's fell 7.3 points to 46.4%.
- The cited literature puts India's R&D spending at 0.63–0.65% of GDP against China's 2.68%, a 4.25-fold gap, and India's AI-specific investment at $1.25B against more than $180B — the structural asymmetry that the stable 3:1 output ratio reflects.
How it was done
AI-related publications were pulled from OpenAlex using eight artificial-intelligence concept identifiers, assigned to India or China by institutional affiliation, and split into a 2019–2021 baseline and a 2022–2024 generative-AI era. A first pass downloaded 40,000 records, 10,000 per country per period, and supported comparisons of citations, h-index, team size and international collaboration using t-tests, Mann-Whitney U tests and Cohen's d. Because the download cap distorted publication counts, annual totals were re-fetched with OpenAlex's year-grouping query, which returns exact population figures, and the trend charts and growth rates were recomputed from those. The partial 2025 year was dropped as incompletely indexed, and the work was written up as a 28-page manuscript with seven figures and more than 50 references.
Data sources
- OpenAlex — AI-related publications with Indian or Chinese affiliations, 2019–2024, across eight AI concept identifiers; exact annual counts plus a 40,000-record sample
- Stanford HAI AI Index Report 2025
- NITI Aayog National AI Strategy (2023) and Carnegie's 2025 analysis of gaps in India's AI capacity
- NSF 2022 R&D trends and international comparisons
- Krishna et al., The Rise of Chinese Universities (2025) and Prathap, JASIST (2017)
Limitations
The citation, h-index and collaboration comparisons still rest on the download-capped 40,000-record sample rather than full populations, so they describe the sampled papers rather than each country's entire output. Counts are unadjusted for field composition, publication lag or citation age, and 2025 was excluded because indexing for that year was incomplete.
Figures from this analysis
How this research was produced
K-Dense Web planned and ran this scientometrics investigation end to end — gathering the sources, carrying out the analysis, producing the figures, and drafting the report. The full session transcript, including every intermediate step, is available to view.
