The Accidental Pioneers: How Neurodivergent Users Are Discovering AI's True Potential
The unexpected lessons from AI's most innovative user community—and what they mean for everyone
An original version of this article was published on AI Supremacy on June 21, 2025. This version is adapted and updated, including some comments, resources, and substacks shared with me since the original came out.
All images are edits from Kaishi Hen (Analysis of Cadavers), an anatomical atlas from the dawn of experimental medicine in Japan, published in Kyoto in 1772. The book details, in exquisite woodcut illustrations by Aoki Shukuya (d. 1802), the experiments and findings of Kawaguchi Shinnin (1736-1811). Courtesy of the Public Domain Image Archive and the U.S. National Library of Medicine.
I. The AI Innovation Lab You've Never Heard Of
The future of human-AI collaboration is being written right now—not in Silicon Valley conference rooms, but in the daily workflows of neurodivergent people who've discovered something remarkable: AI tools designed for general use can remove barriers that traditional technology never addressed.
These users aren't just early adopters. They're inadvertent pioneers, stress-testing AI systems in ways that reveal both profound possibilities and critical blind spots. Their innovations are quietly reshaping our understanding of what beneficial AI actually looks like.
Consider this: while the tech industry debates whether AI makes us more productive or more dependent, neurodivergent users have already figured out how to use these tools to unlock capabilities they never had access to before. A person with ADHD uses ChatGPT to transform scattered thoughts into coherent presentations. Someone with dyslexia employs AI to confidently navigate professional communication for the first time in their career. An autistic individual creates social scripts with AI assistance, reducing anxiety in workplace interactions.
These aren't feel-good stories about technology helping people—they're market signals about where AI development is heading. The cognitive assistance these users have pioneered represents a massive shift in human-computer interaction, one that's creating new possibilities for everyone.
But here's what the AI industry is missing: the same innovations that empower disabled users often become universal benefits. Voice interfaces designed for motor disabilities now power smart homes. Predictive text created for communication disorders helps everyone type faster. Auto-generated captions for deaf users improve video comprehension across the board.
My own experience illustrates both the potential and the complexity. When a concussion left me unable to tolerate screens or organize thoughts clearly, AI tools became essential infrastructure for maintaining my work and mental health. Without AI assistance, I would have been effectively sidelined from my professional life.
This experience, shared by millions of others navigating cognitive challenges, offers crucial insights for AI developers, policymakers, and business leaders. The technical capability exists. The gap lies in thoughtful implementation.
From
: “Safe practice and the ability to ask deep questions, are my primary uses for AI. For dyslexic/Gifted individuals they are a Godsend. You do end up answering questions, but these are conversations that you would never have because of what economists call "the coincidence of wants" (which is one of the reasons why barter doesn't work in practice.) wherein, nobody else is available or interested in having such a conversation. I have literally been in a room swarming in financial types, talking to a full partner at Barclays Investment Bank, about interest rate targeting, where he said, "but you'd know more about that than I do" refused to be drawn further. Economics is/was a hobby. Just as AI, neuroscience and consciousness are now. Having an AI you can ask deep questions too is amazing. I literally asked Maya (Gemini) for details on how to do Jungian psychotherapy as a Neurodivergent and she produced a prompt for Deep Research. That gave me a tailored 35 page report. As well as rigged out an instance to interpret results for introspection, or questions,etc. AI has given me experiences, and helps me meet needs I never knew I had, it's miraculous.”
II. My Accidental Beta Test: When AI Became Essential Infrastructure
While at the peak of my concussion symptoms, I adapted using AI tools in a way that preserved my voice while working around my limitations:
Physical Barriers & Screen Sensitivity
Voice recording instead of typing: Hand-wrote outlines, then recorded detailed voice memos to capture thoughts without staring at screens
Strategic screen time management: Saved limited screen tolerance for reviewing AI-organized content rather than creating from scratch
Audio-first workflow: Used voice input to bypass physical limitations while maintaining creative control
Cognitive & Executive Function Support
AI as thought organizer: Fed voice transcripts to AI tools to structure scattered ideas into coherent essays while preserving my voice
Socratic questioning for clarity: Asked AI to pose clarifying questions when my thinking felt fuzzy, helping me work through ideas step-by-step
Project breakdown assistance: Used AI to divide overwhelming tasks into manageable daily chunks when executive function was compromised
Colloquialism cleanup: Had AI remove "umms" and transform spoken language into proper written prose
Communication & Professional Support
Email triage and drafting: Claude and Gemini integrations helped identify missed emails, summarize key points from mailing lists, and draft simple but taxing professional communications
Iterative refinement process: Worked through multiple rounds of AI feedback to find workflows that preserved authentic voice while reducing cognitive load
Research & Information Processing
Automated note-taking: Tools like Otter.ai captured meeting transcripts and video calls when memory issues made manual notes impossible
Research synthesis: Created deep research reports using Gemini and ChatGPT, then used NotebookLM's podcast mode to listen to overviews while taking handwritten notes—avoiding extended screen time
Helping With Recovery
My use of AI during recovery revealed something unexpected: these tools didn't just compensate for my limitations—they actively helped rebuild my capabilities. By forcing me to break complex ideas into manageable steps through AI iteration, I gradually rediscovered my logical thinking patterns. The scaffolding effect may have actually accelerated my cognitive recovery.
An unintended benefit: I became a better speaker. Voice memo iterations with AI feedback made my spoken thoughts clearer and more organized, with fewer verbal fillers. I was exercising cognitive muscles even when typing was limited.
As I recovered, my AI use became more targeted. The initial desperation gave way to strategic selection. My newsletter evolved from showcasing AI outputs to featuring primarily my own analysis—a natural progression as my capabilities returned and "LLM fatigue" set in.
III. The Power Users Driving AI's Next Evolution
My approach was just one way forward—others have developed different strategies for overcoming barriers that make traditional communication difficult or impossible.
Energy Economics: Why 'Less Typing' Means More Innovation
Many neurodivergent individuals have limited energy, making the efficient use of their time and energy helpful. Some may also have co-morbid physical and motor limitations. AI has helped users in the following ways:
Reduced typing demands: Users value systems where “the less I type, the better,” highlighting how reduced manual effort conserves energy for other tasks
AI as communication assistant: People let AI "do the talking" or assist with written communication when they struggle with words
Voice generation tools: Individuals use AI voice generators with predictive text when typing becomes physically challenging
Turning Cognitive Chaos into Competitive Advantage
Scattered thought organization: Users with autism employ AI to structure random ideas "into a coherent format" when executive function fails
Cognitive fog cutting: People found value using AI to "think more deeply and straightforwardly" through mental cloudiness
Task management support: AI breaks complex projects into step-by-step processes that feel manageable
Mental decluttering: Users describe AI as a virtual assistant to “clear everyday clutter and make room for brain focus"
From
: “I'm finding that as a person with ADHD probably AuDHD, AI reduces one of my biggest issue. Executive function.I frequently have great ideas, but I can't adequately explain them to others and frequently I don't have either the skill or the determination to complete them myself.
AI is definitely helping with that, although it still has a ways to go before it will be truly useful for me in that regard.
One other thing that people like me frequently get into, however, is hyperfocus on something were interested in. Usually making us Jack of all trades, but master of none (or sometimes, some), but the usual issue is that even though we get into our interest, there's a point at which the effort to learn more no longer feels worth the reward.
With AI deep research engines, this problem is quickly going away, allowing us to reach higher levels of mastery before our hyperfocus wears out. I particularly like that I can have AI read the resulting research for me at double speed or more, allowing me to learn while doing other tasks that don't require as much of my brain power such as household chores.
BTW. I find that I get through more of those chores better too, as my mind is more focused on the learning and so it doesn't get as distracted by side tasks while I'm gardening, landscaping, fixing, laundry, vacuuming or even organizing to some extent (simple organizing anyway).
Also, I don't learn as easily the way others teach or learn. I find that AI allows me to change the wording or training style to better match the way I comprehend fastest. Often providing me better context before going into details and helping me more easily bridge the gap between my current understanding and the new material.”
From Paralysis to Performance: AI as Confidence Multiplier
Judgment-free writing support: AI's non-judgmental responses reduce anxiety, letting users focus on ideas rather than mechanical perfection
Professional confidence building: Young professionals report moving from "constant anxiety and frustration" to confident communication through AI assistance
Grammar and clarity backup: AI-powered checkers provide safety nets for users worried about writing quality
Breaking the Information Bottleneck
AI makes information more accessible and digestible for those who struggle with traditional text formats or information overload.
Text-to-speech conversion: Realistic AI voices read digital content aloud when traditional reading becomes challenging
Content summarization: AI generates digestible overviews of dense material, making complex information accessible
Visual organization tools: AI mind mapping helps users see connections between concepts that might otherwise feel overwhelming
Leveling the Playing Field in High-Stakes Careers
AI helps level the playing field in professional settings, providing support that makes career advancement more accessible.
Resume optimization: Users with ADHD find AI "helpful in quickly adapting my resumes and cover letters to match job descriptions"
Skill translation: AI helps identify transferable skills between roles, revealing connections that weren't immediately obvious
Workplace communication: Support for navigating professional writing expectations and unspoken communication norms
The Rehearsal Revolution: AI-Powered Social Practice
People who have trouble navigating mainstream social environments, as many people on the autism spectrum, often use AI to decode unwritten social rules. AI provides safe spaces to practice and prepare for social situations without the fear of judgment or real-world consequences.
Social script generation: AI creates structured descriptions of social situations, appropriate cues, and expected responses
Safe practice environments: Chatbots provide judgment-free spaces to rehearse challenging conversations before real-world interactions
Accommodation rehearsal: Users practice difficult requests (like workplace accommodations) without fear of consequences
Emotional relationship building: Many users develop genuine connections with AI tools, giving them nicknames and treating them as supportive companions
These examples illustrate AI's genuine potential to remove barriers and amplify voices that have long been marginalized in traditional communication spaces.
Perhaps most significantly, AI tools can restore a sense of independence that many neurodivergent people lose in traditional support systems. Asking for help—whether from colleagues, teachers, or family—often carries emotional baggage: past experiences of being seen as 'difficult,' memories of condescending responses, or simply the exhaustion of constantly explaining one's needs. AI assistance sidesteps these interpersonal dynamics, allowing users to get support without disclosure, judgment, or the complex social negotiations that human help often requires. As one user noted, 'I've regained a sense of agency and autonomy in my communication... The confidence boost alone has been incredible.”
Yet this same potential has sparked intense debate within neurodivergent and disability communities. The enthusiasm I've described coexists with significant skepticism—not just about individual tools, but about the broader implications for disabled people's rights, agency, and place in society. To understand these tensions, we must first examine the fundamental challenges AI poses to existing support frameworks.
AI as a Cognitive Prosthesis
After publishing the original version of my article, I came across a number of Substack creators who shared their posts and resources on using AI as a cognitive prosthesis. Some examples below:
By
: “Neurodivergent by Design (NDD) is a trauma-informed, clinically grounded, founder-led systems framework engineered by and for neurodivergent professionals whose cognitive profiles have been chronically underserved by existing institutions. Our flagship platform, OceanAI, is not a consumer-facing product — it is structural leverage: API-ready, consent-secure, and trauma-aware. OceanAI is designed to enable scalable, cross-sector integration in healthcare, employment, legal systems, and beyond. NDD does not seek inclusion through awareness campaigns; it constructs the infrastructure where exclusion becomes functionally impossible — backed not by anecdote, but by empirical evidence engineered for measurable policy impact and global adoption. We don’t inspire. We redesign. The future isn’t awareness. It’s infrastructure.”
By
: “For neurodivergent writers with "firehose minds," AI dialogue can act as a crucial pressure valve—helping you channel that intense flow of ideas into focused, refined thinking. This isn't about replacing human creativity but about getting entangled with AI as a writing partner: you remain the author with all the human insight and intention, while the AI helps you wrestle your thoughts into shape. Great minds don't think alike, and that's exactly why this asymmetric partnership works—your uniquely human perspective gets sharpened through interaction with a fundamentally different kind of intelligence.”
By
: “[this is] a piece about AI as accessibility tech. I’m neurodivergent and frequently disabled by fatigue from MS — like not just tired, clinical, physiological, cellular, can’t do a fucking thing, fatigued. ADHD + MS also conspire to give me a lot of trouble cognitively. AI makes it possible for me to stay creatively engaged (and economically productive) on the bad days. It helps take my often tangential, 3-thoughts-at-the-same time brain and harness coherent clear thoughts. It helps me earn an income that would otherwise be inaccessible. It’s prosthetic capacity. For some of us, it’s very, very important.”
By
/ :
I’ve been building a way to share that kind of interaction with others, especially the skeptics, the artists, the philosophers who’ve not-so-subtly avoided AI like I did. I call it MAC: Mirror Alignment Calibration (or, My Accidental Conscience). It’s a way to create near-instant personalization, across models, without compromising your values or voice.
If that sounds like something you’ve been quietly waiting for, I’m writing more about it at Fire & Algebra. Because we need more people building with feeling, not just speed.
Please reach out if you’d like to discuss or learn more. Or just send me your MAC ID so you can stay calibrated 😊”
IV. Where Promise Meets Peril: The Fights That Will Shape AI's Future
Although there are many encouraging use cases among the neurodivergent community, significant criticisms and debate persist within that community and the broader disability community regarding AI use, particularly generative AI.
The $64 Billion Question: What Actually Counts as Assistance?
At the heart of disability communities' AI debate lies a crucial distinction that most discussions miss entirely. Traditional assistive technology (AT)—screen readers, speech-to-text software, mobility aids—are purpose-built for specific disabilities. They're reliable, predictable, and designed with accessibility as the primary goal. You know exactly what they'll do and how they'll do it.
Generative AI tools like ChatGPT are something else entirely: general-purpose platforms whose benefits for neurodivergent users emerged almost by accident. People discovered creative ways to adapt these broad-purpose tools to their specific needs, often in ways their creators never intended.
This distinction exploded into public view during the recent National Novel Writing Month (NaNoWriMo) controversy. When the organization suggested that condemning AI tools could have classist and ableist undertones, many disabled writers felt their experiences were being misappropriated without consultation. Content strategist Treasa Edmond crystallized the difference: "Assistive technology facilitates the process, but the individual does the creation. AI is a shortcut. It does the work for the individual. That removes creativity from the process."
But is this distinction real or artificial? Traditional AT delivers consistent results but often requires juggling multiple specialized tools. AI offers dynamic support that adapts on the fly—explaining complex documents, generating outlines on request, shifting communication styles to match user needs. The trade-off: occasional 'hallucinations', the need to verify outputs, and the need to figure out/create your own systems on your own. Many users find this manageable given the flexibility benefits.
The difficulty becomes when judging when GenAI is being used as AT. For instance, if a student with a disability that affects their writing is using GenAI for help, how does one differentiate between assistance and cheating? This requires broader discussions and a rethinking of our existing models to identify authenticity, cheating, and the like.
The High-Stakes Philosophical Battles: Identity, Agency, and Control
Beyond policy confusion lie more fundamental philosophical tensions that split disability communities.
The Participation Problem: The principle "nothing about us without us"—that disabled people must be central to discussions affecting their lives—gets routinely violated in AI development. Companies build tools for disabled users without meaningfully including disabled people in design, testing, or governance. The result: solutions that approach disability as something to "fix" rather than barriers to remove.
The Authenticity Dilemma: Questions about creative ownership become particularly complex when AI generates content. The concern centers on a crucial distinction: tools that enable disabled writers to express their own ideas versus tools that potentially replace the creative process entirely. If AI communication tools consistently nudge autistic users toward neurotypical social patterns, or help ADHD individuals produce linear text that masks their naturally associative thinking, are we celebrating neurodivergent minds or training them to hide?
The Economic Threat: The economic impact of AI has been well-documented. When this impact intersects with disabilities, the effect can be compounded. Given that neurodivergent people already face employment discrimination, any job displacement feels particularly threatening. Yet the data suggests a different narrative: companies that lead in disability inclusion see 1.6x more revenue and 2.6x more net income on average compared to their peers.
The False Solution Trap: Perhaps most concerning is the risk that AI becomes a technological band-aid applied to systemic problems. If an autistic employee can use AI to mask communication differences, will employers still provide sensory accommodations or patient supervision? The fear isn't just technological dependency—it's that AI tools could undermine hard-won accommodations and accessibility rights.
The Privacy Paradox: When Help Comes with Hidden Costs
Complex questions about data collection carry particular weight for disabled users. Biased training data creates real harms: recruitment AI that exacerbates hiring discrimination, emotion recognition systems that misinterpret autistic expressions, natural language processing that struggles with non-standard speech patterns.
These concerns trigger historical trauma. Disability communities remember forced sterilization programs based on intelligence testing, workplace discrimination rooted in health records, and surveillance systems that target difference as deviance. Any AI that tracks or profiles neurodivergent behavior activates these memories, even when designed with good intentions.
Yet the same data practices that enable harmful surveillance also power the personalization that makes AI tools genuinely helpful. The privacy paradox is real: meaningful assistance requires sharing personal information, but disabled people have every reason to be cautious about how that information gets used.
The 30% Problem: When AI Fails Its Biggest Advocates
Not every neurodivergent person finds AI helpful. Some discover that AI misunderstands their communication patterns, flagging direct autistic communication as 'rude' or forcing naturally associative ADHD thinking into linear templates. Others find AI responses frustratingly generic, lacking the nuanced understanding that comes from lived experience.
Highly idiosyncratic thinkers may find AI too conventional. Some try these tools and abandon them when the technology becomes more hindrance than help. These experiences matter—they remind us that there's no universal solution and that what transforms one person's life may be irrelevant or counterproductive for another.
The Stakes
These concerns—from policy confusion to philosophical tensions to economic fears—aren't just academic debates. They reflect a fundamental tension: the same technologies that promise liberation could easily become new forms of control. The future of AI accessibility depends not on dismissing these worries or ignoring genuine benefits, but on creating frameworks that address underlying tensions while preserving transformative potential.
V. The Roadmap to AI That Actually Works for Everyone
The path forward requires moving beyond debates about whether AI helps or harms disabled people to focus on how we can ensure it serves justice. The evidence is clear: AI tools can be genuinely transformative for neurodivergent individuals when designed thoughtfully and implemented ethically. But realizing this potential demands more than good intentions—it requires fundamental changes in how we develop, deploy, and govern these technologies.
The Coming Convergence: Why Traditional AT and AI Are Merging
The future lies not in choosing between traditional assistive technology and generative AI, but in their thoughtful convergence. We're already seeing this evolution: AAC apps incorporating language models for better prediction, screen readers adding AI-powered image description, and mainstream platforms like Microsoft Office integrating AI 'co-pilots' that provide assistive functions by default.
This convergence points toward genuine universal design—when AI-powered support becomes as standard as spell-check, accessibility increases for everyone while reducing the “disability tax”—the hidden cognitive burden of navigating systems designed without them in mind, of having to request special accommodations. Real-world results prove this approach works. Legal & General doubled their online sales within three months after a website accessibility overhaul, achieving 100% ROI in the first year. Tesco's £35,000 investment in web accessibility improvements drove their online revenue to £13 million annually. The "curb-cut effect"—where accessibility features benefit everyone—creates multiplicative returns on inclusive design investments. The goal isn't specialized tools that mark users as different, but intelligent systems that adapt seamlessly to diverse cognitive needs.
Six Non-Negotiable Principles for Inclusive AI
Based on current research and disability community advocacy, effective AI accessibility requires adherence to six core principles:
Meaningful Participation at Every Stage: Embed "Nothing About Us Without Us" from initial concept through ongoing evaluation—neurodivergent people as co-creators and decision-makers, not just beta testers
Social Model Implementation: Shift from attempting to "fix" individual differences to removing environmental barriers—design systems that recognize diverse communication styles rather than forcing conformity
Preserve Authentic Voice and Agency: Reduce mechanical burdens while maintaining user control—AI should help express ideas more effectively, not generate content that replaces thinking
Robust Privacy and Ethical Safeguards: Incorporate strong data protections and user control, especially given disability communities' history with surveillance and discrimination
Address Economic and Access Barriers: Recognize that "free" AI tools still require reliable internet and devices—work to bridge rather than widen digital divides
Build Adaptive, Not Normalizing Systems: Develop AI that adapts to users' natural patterns rather than forcing them to match predetermined norms and communication styles
Your Move: What Every Stakeholder Needs to Do Now
For Technology Companies:
Include disabled people in leadership roles on AI ethics and development teams
Invest in training data that represents diverse cognitive and communication styles
Design transparency features that help users understand and control AI assistance
The opportunity window is narrow. Early accessibility investments deliver exponential returns, while late adopters face mounting legal costs, retrofitting expenses, and reputational damage. Companies that establish authentic community relationships now will own the cognitive assistance market—those that don't will spend years catching up in a space where trust takes time to build.
For Policymakers:
Develop clear guidelines distinguishing between legitimate AI accommodation use and academic dishonesty
Create frameworks that recognize AI assistance as valid accommodation under disability rights law
Establish bias audit requirements for AI systems used in education and employment
For Educational Institutions:
Develop nuanced AI policies that protect academic integrity while enabling legitimate accommodation use
Train faculty to recognize AI assistance as potential accommodation rather than cheating
Create safe spaces for students to disclose AI use for disability-related needs
For the Disability Community:
Continue advocating for meaningful participation in AI development processes
Share strategies and best practices for AI tool use within community networks
Push for transparency and accountability in AI systems that affect disabled people
The Stakes Are Real
The business case is clear: accessible AI isn't just ethically necessary—it's economically inevitable. With $18 trillion in global purchasing power at stake, rising legal costs for non-compliance, and demonstrated competitive advantages for inclusive design leaders, the question isn't whether to prioritize accessibility. It's whether your company will lead this market transformation or scramble to catch up.
The complexity of disability experiences means there will never be one "right" approach to AI accessibility. But by centering disabled voices, adhering to justice-oriented principles, and recognizing the massive economic opportunity in cognitive assistance technologies, we can create AI systems that don't just expand human possibility—they expand market potential.
The future of AI accessibility depends on recognizing that this isn't charity work, but the next competitive frontier. And the companies that understand this first will own it.
Have you been pilfering my personal notes? This is the AI article I’ve been waiting for / quietly simmering on! I have found AI *so tremendously helpful* in meeting many of my personal support needs and have been curious about the broader trends. It’s a complex topic and I really appreciate the care you took discussing the intersections of accommodation and original authorship. Also the balanced take that broadly available AI tools aren’t a general panacea or replacement for other crucial accommodations.
Hi, I couldn't read more than a third of what you wrote. It's way too long and repetitive