An analysis of the top 100 cited AI papers per year from 2020 to 2022 shows US-based AI research as the most impactful and Google the dominant organization
Who Is publishing the most Impactful AI research right now? With the breakneck pace of innovation in AI, it is crucial to pick up some signal as soon as possible. Tweets: @zetavector , @zetavector , @jungwooha2 , @markpeak , @zetavector , @tobywalsh , @alexeyab84 , @zetavector , @mmalex , and @ylecun Tweets: @zetavector : The 100 most cited AI papers for 2022. A detailed analysis of the most cited papers for the last three years allows good insights into the organisations and countries publishing the most impactful AI research right now. Read here: https://www.zeta-alpha.com/... A thread 🧵 https://twitter.com/... @zetavector : In the recent discussion about the impact of AI R&D from various Big Tech labs, we can now contrast volume and impact. Volume top-5: @GoogleAI @Tsinghua_Uni @MSFTResearch @CarnegieMellon @MIT Impact top-5: @GoogleAI @MetaAI @MSFTResearch @berkeley_ai @DeepMind https://twitter.com/... Jung-Woo Ha / @jungwooha2 : Report on Top-100 impact papers in 2022 from Zeta-Alpha. #NAVER is placed at Top-5 conversion rate (Top impact papers / Total published papers)!! I am so proud of the #NAVER_Cloud AI members including #NAVER_AI_Lab! https://www.zeta-alpha.com/... https://twitter.com/... @markpeak : “OpenAI will kill Google” https://www.zeta-alpha.com/... https://twitter.com/... @zetavector : The US and Google are by far the most prolific country and organization, but OpenAI is on a league of its own in terms of ratio of publication to blockbuster papers. Read about it in our latest analysis of the most impactful research since 2020👇 https://www.zeta-alpha.com/... https://twitter.com/... Toby Walsh / @tobywalsh : Australia easily in top 10 .... https://www.zeta-alpha.com/... Alexey Bochkovskiy / @alexeyab84 : Proud to be among the authors of the top 3 papers for 2020: 1. ViT (Google) - Alexey Dosovitskiy et al. 2. GPT-3 (OpenAI) - Tom Brown et al. 3. YOLOv4 (Academia Sinica) - Alexey Bochkovskiy et al https://www.zeta-alpha.com/... https://twitter.com/... https://twitter.com/... @zetavector : Recent rumours of China overtaking the US seem largely exaggerated, when we look at the top-100 cited papers in AI in 2022. The dominance of US-based AI Research becomes even more prominent when we look at the citation count itself: https://www.zeta-alpha.com/... https://twitter.com/... @mmalex : instant ngp clocking in at #7 most cited ai paper in 2022! 🤯 i get a massive thrill whenever i see brilliant work (that i couldn't do!) that says ‘...we build on a hashgrid’ or ‘...use ingp base...’ and it really reinforces how science progresses by building on each others' ideas https://twitter.com/... Yann LeCun / @ylecun : At @MetaAI we favor publication quality over quantity. That's why among the 100 most cited AI papers in 2022, @MetaAI has authored (or co-authored) 16, ranking 2nd just behind Google with 22. Our research is having a large impact on the community. (and NYU ranks nicely, too). https://twitter.com/...
Context & Ripple Effects
This analysis lands mid-debate over what 'winning' at AI research means. Earlier work established the volume case: an [[a:939503|Allen Institute study found Chinese scholars out-published US scholars every year since 2005]], and the Nikkei study claimed China also led on output and cited-paper counts in 2021, producing roughly twice as many papers as the US.
Zeta Alpha's cut through the same period argues that at the very top of the citation distribution the picture reverses: US-affiliated work dominates the annual top 100, Google alone authored 22 of them in 2022 with MetaAI next at 16, and OpenAI stands out on blockbusters per paper published — GPT-3 was one of the three most cited papers of 2020, alongside Google's ViT.
First-order effects
- Big Tech labs are confirmed as the center of gravity for impactful AI research: Google and MetaAI together account for nearly two-fifths of the 2022 top 100, while OpenAI's high blockbuster-to-volume ratio shows a small lab can out-punch its publication count.
Second-order effects
- National AI-leadership claims now split along metric lines — China-leaning studies cite volume and total citations, this analysis cites elite-paper share — so governments and funders get dueling evidence bases depending on which ranking they quote.
Third-order effects
- The pattern does not hold cleanly forward: the later [[a:864420|CSET analysis put the Chinese Academy of Sciences ahead of Google in highly cited paper production]], suggesting the US impact edge narrows as Chinese institutions climb the quality curve, and shrinking US-China co-authorship means these rankings become harder to attribute cleanly to any one country.
The trend: Assessments of national AI strength are shifting from raw publication volume toward top-end citation impact, with corporate labs like Google and OpenAI capturing an outsized share of the field's most influential work.