I have been asking myself a question for a while now: Are men more likely to attach themselves to an AI companion? And the related one: Are there more men in the AI field? I keep noticing how many of the speakers, the major figures, the named voices, are male. And when a woman appears in the field, she is usually waving some cautionary tale. Like me, now.
So I started to look at the numbers. And the answer is yes. Yes, more men than women are interested in an AI companion or girlfriend. Yes, more men than women attach themselves to a chatbot. Yes, more men than women hold the visible AI field positions. Yes to all of it. And it is unsettling.
As someone researching the effects of tech on human connection, and the way our connections are crumbling more and more as our tech is advancing, this is not unexpected. What is unexpected is that men embrace this more than women. I would have thought women would lead the field, since they generally seem to want communication more. And an AI chatbot is offering plenty of that.
But no. It is more men. I will not venture an analysis why. That is for another day. But I am quietly relieved I am not in a position to want a human companion right now. Because if I had to compete with an endless supply of ever-compliant AI partners, I would most certainly lose the battle.
Below, I will let the numbers talk for themselves.
Fig. 01 / Authorship
Women's share of AI research authorship, by venue
The production layer. Across every major slice of the AI literature, women remain a minority — and the gap has not closed since the 1990s.
arXiv (all AI papers)Nesta · 2019
13.8%
Computer Vision researchersJin et al. · 2023
15.6%
AI scholars (100+ citations)Jin et al. · 2023
18.0%
NLP last authorsMohammad · 2020
25.0%
NLP first authorsMohammad · 2020
29.0%
0%25%50%75%100%
Fig. 02 / Pipeline
The pipeline, lab by lab, cohort by cohort
At every stage of the academic and industrial track — from PhD admission to faculty hire to research scientist at a frontier lab — women remain under a quarter of the field.
Google AI researchersNesta · 2019
11.3%
Microsoft AI researchersNesta · 2019
12.0%
Tenure-track CS facultyStanford AI Index 2023
16.0%
New AI PhDs (2021)Stanford AI Index 2023
21.3%
New CS faculty hiresStanford AI Index 2023
30.2%
0%25%50%75%100%
Fig. 03 / Authorship of the foundational papers
Who wrote the papers that built the LLM era
Each circle is one author. Filled red is a woman. The technical canon of the modern AI moment was authored almost entirely by men.
Attention Is All You Need · Vaswani et al., 2017
8 authors · 173,000+ citations
M
M
W
M
M
M
M
M
The Transformer paper. 1 woman of 8 authors (12.5%) — Niki Parmar.
Language Models are Few-Shot Learners · Brown et al., 2020 (GPT-3)
31 authors · the paper that launched the LLM commercial era
M
M
M
W
M
M
M
M
M
W
W
W
W
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
5 women of 31 authors (16%) — and every one of them assigned to safety, policy, or human evaluation.
Fig. 04 / Role-sorting, in their own words
What the women on GPT-3 were credited with
From the paper's own author contribution statement. The pretraining, the kernels, the architecture, the scaling — credited to men. The five women were assigned to evaluation, fairness, policy, and threat analysis.
Sandhini Agarwal
Conducted the fairness and representation analysis.Ethics / bias
Amanda Askell
Conducted the human evaluations of the model.Human evaluation
Gretchen Krueger
Edited and red-teamed the policy sections of the paper.Policy editing
Ariel Herbert-Voss
Conducted the threat analysis of malicious use.Threat analysis
Melanie Subbiah
Implemented, experimented with, and tested beam search.Implementation / eval
Fig. 05 / The schism among the men
The men running AI cannot agree on what AI is
Four camps, ferociously opposed on timelines, architectures, regulation, and existential risk. They sue each other. They publish open letters against each other. And every named figure in the dispute is male.
Camp 01 — Accelerationist
Build it now, faster, bigger
"AI will probably most likely lead to the end of the world, but in the meantime, there'll be great companies." — Altman, pre-OpenAI
Sam Altman · Elon Musk · Marc Andreessen
P(DOOM): 0–30%, BUILDS ANYWAY
Camp 02 — Safety-pilled scaling
Build the bomb so the bad guys don't
AGI in 1–10 years, genuinely dangerous, but the right people building it carefully beats the alternative.
Dario Amodei · Demis Hassabis
P(DOOM): 10–25%
Camp 03 — LLM-skeptic
The whole industry is LLM-pilled
"We're never going to get to human-level intelligence by training LLMs." LeCun left Meta in Nov 2025 over this.
Yann LeCun · Gary Marcus
P(DOOM): <0.01%
Camp 04 — Doomer / shut-it-down
If anyone builds it, everyone dies
Yudkowsky has called for international moratoria backed by military force against non-compliant data centers.
When AI is discussed in public — on the dominant long-form podcast in the field — women appear, but not as AI authorities. The technical voice of AI in popular media is a male voice.
18.6%
Women as a share of recent Lex Fridman guests
~14 women across the most recent ~75 episodes of the dominant tech podcast (3.6M+ subscribers).
0/14
Of those women, how many appeared as AI technical experts
They were physicists, economists, historians, politicians, musicians, a divorce psychologist. The AI seat was reserved for men.
100%
Male voices in The Scaling Era: An Oral History of AI 2019–2025
Patel & Leech's 2025 book — the canonical oral history of the LLM moment — is built from interviews with Amodei, Hassabis, Zuckerberg, Aschenbrenner. All men.
+57%
Widening of the academic productivity gap after ChatGPT
PNAS Nexus, 2025. Generative AI is widening pre-existing gender gaps in academia, not closing them.
Fig. 07 / The men who love them
Who actually wants the machine
If men build AI and women critique it, men also use it more, trust it more, and form romantic attachments to it at multiples that the data leaves no room to argue with.
"AI girlfriend" Google searchesvs. "AI boyfriend" · TRG/Ahrefs 2024
9 : 1
Male : female ratio among Replika reviewers96layers analysis of 14,000 reviews · 2024
8 : 1
Male share of AI girlfriend platformsMarket Clarity industry analysis · 2025
82%
Male share of Replika usersMarket Clarity / Replika · 2025
65–72%
0%25%50%75%100%
Google searches for "AI girlfriend" rose 2,400% between 2022 and 2024. The global market for AI companion apps was valued at $2.8 billion in 2024, projected to reach $9.5 billion by 2028. The product is overwhelmingly female-coded; the customers are overwhelmingly male.
A note on Character.AI, which is roughly 50/50 by gender: it is a general-purpose role-play platform, not a romantic-companion product. The male skew sharpens dramatically once the use case narrows from "talk to anyone" to "talk to a girlfriend."