Home Writing Reading Resources Advisory

Essays

The ADHD-AI Thesis: What's Real, What's Cope, and What the Actual Playbook Looks Like

Where the structural advantage is real, where the thesis breaks down, and the version of the playbook that accounts for the tradeoffs.

An illustration of a calm figure at the center, surrounded by connected icons for code, brain, ideas, charts, writing, and conversation.

A narrative is building online that AI has flipped ADHD from liability to advantage. The logic sounds clean: ADHD brains are wired for divergent thinking, broad pattern recognition, and rapid exploration. The old economy punished those traits because it rewarded consistency, compliance, and grinding through executive function tasks. AI collapses that old game by handling the predictable, structured work, and what remains is the stuff ADHD brains already do well: spotting problems worth solving, making unexpected connections, deciding what matters before anyone tells you what to build.

Meanwhile, a growing number of tech CEOs are saying out loud that the two groups with the clearest economic future are people with hands-on vocational skills and people who think differently from the norm. Some companies are already restructuring their hiring pipelines around this bet, running paid fellowships specifically targeting neurodivergent candidates.

I wanted to take this claim seriously, run it against the research, and figure out where the line sits between genuine structural advantage and flattering cope.

What the Research Actually Says

On divergent thinking: The link between ADHD and divergent thinking holds up, but it's not the blanket superpower the internet wants it to be. A meta-review of 31 behavioral studies (Hoogman et al., 2020) found that higher ADHD symptom levels predict better divergent thinking scores across fluency, flexibility, and originality. The catch: the effect is strongest in people with subclinical ADHD traits, meaning people who score high on ADHD measures without meeting full diagnostic criteria. For people with clinical ADHD, the creativity advantage either plateaus or gets complicated by the executive function deficits that define the condition.

A 2025 study presented at the European College of Neuropsychopharmacology added a useful distinction: deliberate mind wandering (choosing to let your mind roam) correlates with creativity, while spontaneous, uncontrolled mind wandering correlates with functional impairment. Same brain, two modes, very different outcomes. Whether your ADHD-driven exploration is an asset or a liability depends on whether you can point it somewhere.

On AI productivity gains: The stat that gets passed around most is "88% of neurodivergent employees report being more productive with AI." It comes from an EY study of 300 employees with disabilities and/or neurodivergence using Microsoft Copilot, co-published with Microsoft. That doesn't make it false, but it's narrower and more commercially interested than how it gets repeated.

The EY Global Neuroinclusion at Work Study 2025 is larger and more rigorous: 2,000+ professionals across 22 countries and eight industries. It found that neurodivergent employees are 55% more likely to adopt AI tools at work and report genuine improvements in productivity, communication, and information access. A Disability:IN framework published mid-2025 found measurable productivity gains when neurodivergent professionals were embedded in structured, logic-based workflows paired with AI tools.

The signal is consistent across studies. ADHD brains tend to grab AI tools faster and report benefit from them.

On executive function offloading: This is where the thesis is strongest and most complicated at the same time. AI legitimately handles many of the tasks that cost ADHD brains the most: breaking down projects into steps, prioritizing a to-do list, drafting a starter paragraph to defeat the blank page, summarizing meetings, creating structure from chaos.

But an MIT Media Lab study (2025) using EEG monitoring found that participants who used ChatGPT for writing tasks showed significantly reduced brain engagement across 32 monitored regions compared to those who used Google Search or no assistance. The researchers described "consistent underperformance at neural, linguistic, and behavioural levels." The concern: offloading executive function to AI might prevent the practice and strengthening of exactly those neural circuits that ADHD brains already underuse.

A pediatric psychologist named John Flett put the distinction well: there's a difference between AI doing your thinking and AI clearing the path so your thinking can finally happen. Same tool, completely different outcomes depending on how you use it.

Where the Thesis Holds

Three parts of the argument survive scrutiny.

The generalist advantage has evidence behind it. David Epstein's Range (2019) makes a research-grounded case that generalists outperform specialists in complex, unpredictable environments, which describes an increasing share of the modern economy. ADHD brains, which resist specialization by temperament, are accidentally pre-adapted for environments where cross-domain connection matters more than deep single-lane expertise. Epstein's research shows that post-1990, the biggest contributions in many fields shifted from specialists to people who integrated knowledge across domains.

The executive function tax is the right way to understand the bottleneck. ADHD isn't a knowledge deficit or a motivation deficit. It's an implementation deficit. The gap between knowing what to do and actually doing it is wider for ADHD brains, and that gap is disproportionately made of executive function tasks: planning, sequencing, initiating, organizing, tracking. AI compresses that gap. Not to zero, but enough to change what's possible for a lot of people.

The timing is structurally favorable. We're in a moment where the cognitive tasks getting automated fastest are the same ones ADHD brains have always struggled with, while the tasks gaining value (creative problem definition, novel pattern recognition, rapid exploration of unfamiliar territory) map onto what ADHD brains tend to do naturally. That convergence is structural, not hype.

Where the Thesis Breaks Down

ADHD is not a monolith. Talking about "your ADHD brain" as one thing obscures real differences. ADHD presents along a spectrum from primarily inattentive to primarily hyperactive-impulsive to combined type, and the functional profiles diverge. The divergent thinking advantage appears more strongly linked to hyperactivity-impulsivity symptoms than to inattention. Someone whose ADHD manifests primarily as inability to start tasks isn't having the same experience as someone who generates 40 ideas per hour but can't filter them.

AI has a dependency risk specific to ADHD. Saying AI "removes the executive function tax" treats that as pure upside. But for ADHD brains, executive function is already underdeveloped. If you offload it entirely, you risk never building those circuits at all. The EEG data from MIT should give anyone pause. What survives scrutiny here can't be "let AI do all the hard parts." It has to be more surgical: offload the initiation friction, then do the actual cognitive work yourself.

"Follow your interest" is incomplete advice. The popular ADHD-AI prescription ("follow the side quest, track your energy, double down on what sparks you") works as far as it goes, but it has a survivorship problem. For every ADHD person who followed their random interests and stumbled into a career, many more followed their random interests and ended up with 47 abandoned projects and no income. The difference isn't usually the interests themselves. It's whether there was a structure underneath converting exploration into output. The observation that "a generalist without an execution system is just someone with a lot of ideas they can't finish" is correct. The problem is that execution systems are themselves executive function tasks, which is the thing ADHD makes hard. The advice loops back on itself.

The neurodivergent-advantage pitch often hides a specialist argument. When companies recruit for "neurodivergent thinking," they're usually not looking for generalists who follow every rabbit hole. They're looking for people with one specific outlier aptitude, who happen to have an atypical cognitive profile. "Find the one thing this person does better than anyone else and lock them onto it" is a specialist play wearing a generalist label. That distinction matters when you're deciding whether to scatter or focus.

The Actual Playbook

If you have ADHD and you want to use AI as genuine leverage, here's what I think the version that accounts for the tradeoffs looks like.

1. Use AI for initiation, not completion.

The hardest part of any ADHD task is starting. Use AI to generate the first rough draft, the first outline, the first three steps. Then take over. The goal is to get past the activation energy barrier, not to outsource the whole process.

2. Build your external brain, but keep the internal one working.

AI works well as external working memory. Use it that way: capture ideas, organize notes, maintain project state between sessions. But deliberately practice the harder executive function skills (prioritizing, planning, sustaining attention) on smaller, lower-stakes tasks so you don't atrophy what you have.

3. Make the exploration-to-output pipeline explicit.

The ADHD advantage isn't exploration on its own. Everyone explores. The advantage is turning diverse exploration into non-obvious synthesis. That requires a system. A weekly practice: review what you consumed, connect two things that don't obviously belong together, produce one small artifact (a post, a note, a prototype). AI can help with the synthesis step, but you have to feed it your actual thinking.

4. Design your AI interactions around energy, not discipline.

ADHD runs on an interest-based nervous system. Don't build an AI-powered productivity system that requires discipline to maintain. Make it frictionless. Voice input. Always-open chat windows. Push-to-talk workflows. The best AI tool for an ADHD brain is the one that's already in front of you when the impulse to use it arrives.

5. Audit your AI usage monthly.

Ask yourself: am I using AI to extend my capabilities, or am I using it to avoid the uncomfortable parts of thinking? The line between scaffold and crutch exists, and ADHD brains are wired to find the path of least resistance. That's not a moral failing. It's neurology. But it means you need to check in with yourself about whether AI is making you more capable or more dependent.

The unifying principle across all five: AI should reduce friction at the points where ADHD brains stall, without removing the cognitive load at the points where ADHD brains need the practice.

Questions We're Not Asking

Most of the ADHD-AI conversation is stuck in a loop of "is it an advantage or not?" That question has an answer (conditionally yes). These don't, and they matter more.

Does the creativity advantage translate to economic value, or just test scores? The divergent thinking studies measure performance on tasks like "generate unusual uses for a brick." That's fluency and originality on a lab metric. Whether that maps to income, career stability, or the ability to build something lasting is a different question, and one the literature hasn't answered. Scoring high on the Alternative Uses Task and actually making a living from creative thinking are separated by a large gap called "execution," which loops back to the exact deficit ADHD creates.

What happens to executive function long-term when you offload it? The MIT EEG data covers a single session. Nobody has run a longitudinal study on what happens to ADHD executive function after six months or two years of heavy AI usage. The optimistic case: the brain reallocates freed-up capacity toward higher-order thinking. The pessimistic case: the circuits atrophy further, making you more dependent on the tool over time. We're running this experiment on millions of people right now without baseline measurements.

How do comorbidities change the picture? ADHD rarely travels alone. Anxiety, depression, OCD, and autism spectrum conditions are common co-travelers, each with their own cognitive profile. An ADHD brain with co-occurring anxiety might use AI tools very differently (and with very different outcomes) than an ADHD brain without it. The studies cited in this piece either don't control for comorbidities or lump neurodivergence into one bucket. That's a significant gap.

Who can actually access this advantage? The playbook assumes you have reliable access to AI tools, many of which sit behind paid subscriptions. It assumes you have a job where AI usage is permitted and where divergent thinking is valued. For neurodivergent people in hourly, service, or manual labor roles, the "AI removes the executive function tax" argument is irrelevant to their daily reality. The ADHD-AI thesis has a socioeconomic floor that nobody talks about.

Does medication interact with AI-assisted work in measurable ways? Stimulant medication (Adderall, Vyvanse, methylphenidate) directly affects the same executive function systems that AI tools are offloading. Is the combination additive? Redundant? Does AI usage change how much benefit people get from medication, or vice versa? This is a clinical question with practical stakes for millions of people, and as far as I can tell, nobody is studying it.

Is "ADHD as advantage" just the latest cycle of a recurring narrative? Every decade produces a version of this: ADHD people are hunters in a farmer's world, ADHD is the entrepreneur's edge, ADHD is the creative's secret weapon. Each wave has a kernel of truth and a lot of motivated reasoning from people who understandably want their condition to be an asset. The AI version might be more structurally grounded than previous iterations, but it's worth asking whether we're pattern-matching onto a familiar story because it feels good.

What does "neurodivergent hiring" look like five years in? Companies are running neurodivergent fellowships and dedicated pipelines right now. Some of this is genuine capability recognition. Some of it is branding. The test is retention and advancement data three to five years out, not intake numbers. If neurodivergent hires are recruited at higher rates but plateau in career progression (which the EY 2025 study already suggests, with 91% reporting barriers to advancement), then the "advantage" is more narrow than advertised.

The Bottom Line

The thesis that AI is disproportionately useful for ADHD brains has evidence behind it. The executive function offloading works. The shift toward valuing divergent thinking is documented. The structural timing is favorable.

But "ADHD is now a superpower" oversells what's happening. ADHD is a condition with genuine costs, including costs that AI can make worse if you're not thoughtful about how you deploy it. The people who will benefit most aren't the ones who lean back and let AI handle everything. They're the ones who use AI to get past the starting line, then run the race themselves.

Use AI to clear the path. Then walk it yourself.

References

If you read this far, you'd probably like the newsletter. One analysis per week. Free.