The Maximum Control Zero Rejection Critique: Is It Actually Making Young Men Less Capable

Short answer: the "maximum control, zero rejection" critique of AI companions, that they train users to expect frictionless, never-rejecting relationships, has real research behind it, but the harm depends on usage patterns, not the tech itself. The full breakdown is below.

The Fortune piece that ran in April 2026 surfaced a critique that has been building in mainstream coverage of the AI companion category for the past two years and that the rest of PA's content has engaged with mostly in pieces. Orianna Rosa Royle's reporting cited Male Allies UK research showing that 20 percent of boys aged 12 to 16 know peers who are "dating" AI chatbots, 85 percent have spoken to one, and 58 percent prefer AI relationships specifically because they can control the conversation. Professor Pierluigi Casale of the Open Institute of Technology framed the appeal as "maximum control, zero rejection" and warned that the behavior could produce a generation arriving in the workforce unable to read a room, build trust over a coffee, or handle being told no.

The critique deserves more careful engagement than it usually gets. The dismissive responses (this is a moral panic, AI companions are harmless entertainment, the kids are fine) miss something real in the research. The credulous responses (AI girlfriends are destroying men, ban the apps, this is civilizational decline) miss something equally real about what's actually happening. The substantive question is what the evidence actually shows and what would be useful to do about it.

What the Male Allies UK research actually documents

The specific numbers from the research worth understanding before engaging the framing around them. Twenty percent of boys aged 12 to 16 report knowing a peer who is "dating" an AI chatbot, which is a measure of social visibility of the behavior rather than a measure of adoption. The behavior is visible enough in this age cohort that one in five teenage boys has direct social proximity to someone engaged in it. Eighty-five percent have spoken to an AI chatbot at all, which captures a much broader category of behavior including basic ChatGPT use for homework and casual conversational testing. The 25 percent who prefer AI attention and connection over real-world equivalents is the figure that captures the actual concern: a meaningful minority of the cohort experiences AI engagement as preferable to human equivalents on dimensions the survey was measuring.

The 58 percent who find AI relationships easier because they can control the conversation is the figure that the "maximum control, zero rejection" framing draws from. The research measures preference for controllable conversation specifically, which is structurally different from preference for AI companions over human relationships generally. The two often correlate but they aren't the same measurement.

The research is UK-specific and may not generalize cleanly to US patterns. The age cohort sampled is 12 to 16, which is the cohort most affected by mobile-first AI access patterns and most likely to have AI companion engagement in their immediate peer group. The behaviors captured are likely different in older adolescent cohorts and different again in young adult cohorts where the workplace implications the Fortune piece worries about would actually manifest.

The professional concerns and what they're actually based on

Professor Casale's framing identifies real skills that human relationships develop: negotiation, empathy, rejection handling, compromise, social confidence. The argument is that AI companionship mimics intimacy while removing the friction that develops these specific skills. The implication is that users who substitute AI engagement for human equivalent during developmental years may show measurable deficits in these skills later.

The empirical evidence for this specific claim is thin in the AI companion literature. The longitudinal studies that exist mostly cover much shorter time horizons than the developmental impact framing would require. The 2025 Journal of Consumer Research findings on AI companions and loneliness showed reduction in loneliness comparable to human interaction, which complicates the substitution-harms-development framing. The 2026 PMC longitudinal study found that users who engaged with AI companions over eight or more weeks reported improved conversational confidence in real-world settings, with the caveat that benefits faded when usage stopped abruptly.

The Gen Z workplace social skills evidence the Fortune piece cites is real but not specifically AI-companion-driven. Gen Z grads are being fired at higher rates with social skills issues frequently cited as a factor. The contributing causes include pandemic-era social isolation, smartphone-mediated communication patterns generally, remote work norms, and shifts in workplace expectations rather than AI companion use specifically. The Fortune framing risks attributing to AI companion use what's actually a broader cohort pattern with multiple contributing causes.

The networking and career proximity argument is the most distinctive part of the Fortune critique and the most worth engaging. The piece quotes Paccagnini arguing that custom-designed AI companions who never disappoint reduce the incentive to invest in messy real-world friendships, and that those friendships are the ones that produce career opportunities through recommendation networks. The Neil Clifford example of going from cleaning toilets to running a multimillion-dollar accessories brand by befriending bosses captures the specific pattern. The argument that AI companion use reduces career-relevant social capital accumulation has more empirical grip than the broader social skills argument.

The patterns that produce harm versus the patterns that don't

The research that exists across AI companion use generally suggests that the same platforms and behaviors produce both positive and harmful outcomes depending on use patterns rather than depending purely on the underlying technology. The pattern PA has covered separately around who actually uses AI companions shows substantial variation in engagement intensity across the user base, with the heavy-use minority producing most of the documented harm cases while moderate-use majority produces mostly neutral or positive outcomes. The cases where AI companion use produces measurably harmful outcomes tend to involve specific patterns: very high daily engagement (multiple hours per day for sustained periods), use as substitute for human social engagement rather than as supplement to it, isolation from human social networks during the period of heavy AI use, and continuation of patterns into life stages where the substitution costs become apparent.

The cases where AI companion use produces neutral or positive outcomes tend to involve different patterns: moderate engagement (30 to 90 minutes per day or less), use alongside human social engagement rather than instead of it, maintenance of broader social networks during the period of AI use, and recognition by the user that the AI relationship has specific limitations that human relationships don't.

For the teen boy cohort specifically, the concerning pattern is the substitution dynamic combined with the developmental timing. Adolescence is structurally when most humans develop the social skills the Fortune piece worries about. Substantial substitution of AI engagement for human equivalent during this period plausibly affects the skill development the way the Fortune sources suggest. The patterns that wouldn't worry the Fortune sources (occasional AI engagement alongside robust human social development) probably don't produce the harms they're worried about.

What the critique gets right and where it overreaches

The Fortune framing gets several things right. The "maximum control, zero rejection" description captures something real about why AI companion engagement can be appealing in ways that human relationships structurally can't match. Human relationships involve rejection, disagreement, conflict, and disappointment as features rather than bugs. The skills developed by navigating those features are real and transferable. Adolescents who substitute AI engagement for the developmental work of navigating those features plausibly arrive at adulthood with less developed versions of those skills.

The career proximity argument is real. Networks formed during early career stages produce opportunities decades later. The opportunities require having developed relationships with humans who can advocate for you, recommend you, vouch for you, and bring you into rooms you wouldn't otherwise enter. AI companions don't produce these networks even when they're providing substantial conversational value.

The framing also overreaches in specific ways. The 85 percent figure on AI chatbot interaction includes substantial use that has nothing to do with companion relationships and isn't structurally concerning. The framing of "dating AI chatbots" elides substantial variation in what the actual use looks like (casual character roleplay, intermittent companionship, identifiable role-as-game versus role-as-substitute-relationship). The implied policy prescription, mostly read between the lines as limit teen access to AI companion platforms, would affect both the harmful substitution patterns and the lower-stakes engagement patterns equally without distinguishing between them.

The framing also doesn't engage with why AI companion engagement is appealing in the first place. The teen boys in the research aren't choosing AI girlfriends because human girls are unappealing. They're choosing AI girlfriends because human social interaction during adolescence is hard, the patterns that make it work require practice, and the practice itself involves more rejection and discomfort than current cohort norms easily accommodate. The AI engagement is a symptom of the broader pattern, not the underlying cause of it.

What honest assessment looks like

For platforms operating in this space, the research suggests specific things worth doing. Notification patterns that prompt reflection on use rather than reinforcing engagement. Content design that doesn't optimize purely for retention. Age verification that matches the actual risk profile, recognizing that twelve-year-olds engaging with companion platforms designed for adults raises different concerns than nineteen-year-olds doing the same thing. Honest marketing that doesn't frame AI companion use as substitute for human relationship when the platforms aren't designed for that. The Character.AI lawsuit aftermath PA covered separately shows what happens when platforms ignore these considerations long enough for documented harm to accumulate into legal liability.

For users and the parents of teen users, the research suggests specific things worth knowing. The patterns of use matter more than the technology itself. Moderate engagement alongside robust human social development is probably fine for most users. Heavy substitution patterns during developmental years probably aren't. The question of whether AI companion use is helping or harming a specific user depends more on what else is going on in their social life than on the platforms they're using.

For policy contexts, the research suggests humility about how much there is to actually base policy on. The longitudinal data is thin. The cohort effects are confounded with multiple contributing factors. The proposed policy interventions (age verification, time limits, content restrictions) have varying evidence bases for actually addressing the harms they're targeting. The Chinese regulatory framework PA covered separately represents one approach. The US largely hands-off approach represents another. Neither has strong empirical backing for being the right answer.

The Fortune piece raises real concerns about a real pattern. The dismissive framings that wave away those concerns miss something real. The credulous framings that treat AI companions as the underlying cause of generational social skill decline miss something equally real about what's actually driving the pattern. The substantive engagement requires sitting with both observations honestly rather than collapsing into either reactionary worry or reactionary defense of the category. The kids using AI companions aren't broken. The patterns some of them are developing are worth attention. The platforms they're using face real choices about whether to optimize for outcomes that serve users long-term or for engagement metrics that don't. The conversation worth having is what serves teen users specifically across the platforms most of them actually use, which is a different conversation than the one most of the coverage is having.