Tokenization is how AI systems break down language into countable units for processing and billing, and longer or more complex text (like neurodivergent learning materials with examples, definitions, and multi-layered explanations) requires more tokens and therefore costs more. Understanding this helps you negotiate between comprehensiveness and cost, and explains why a dense 500-word learning guide might cost more per use than straightforward text.
Tokens are the unit by which language models process text. One token ≈ four characters or 0.75 words. When you use paid AI APIs (like ChatGPT's API, Claude's API, or Gemini's API), you pay per token: input tokens cost slightly less, output tokens cost slightly more. Understanding tokenization is particularly important for neurodivergent learners because accessibility often increases token count, which increases cost.
Here's the tension: accessible formatting for neurodivergent learners—bullet points instead of paragraphs, extra white space, repetition for emphasis, chunked information—requires more tokens than dense, efficient text. If you're using a paid API and requesting accessible formatting, you're paying a "neurodiversity tax" unless you're strategic about it.
Original text (efficient, hard to parse for many neurodivergent readers):
"Photosynthesis is the biochemical process by which plants convert light energy into chemical energy, occurring primarily in chloroplasts through two main stages: the light-dependent reactions and the light-independent Calvin cycle, both essential for glucose production and oxygen release."
(22 tokens)
Accessible version (bullet points, chunked):
"Photosynthesis: Light to Energy
• Definition: Plants convert light energy into chemical energy
• Location: Happens in chloroplasts
• Two stages: Light reactions + Calvin cycle
• Output: Glucose (food) + Oxygen
• Why it matters: Foundation of food chains; essential for life"
(48 tokens)
The accessible version is 2.2x the token cost, but for someone with ADHD or dyslexia, it's dramatically more usable. The dense original might require three read-throughs; the bullet version is processed in one pass with lower cognitive load.
If you're a student requesting AI help across hundreds of explanations weekly, this adds up. One semester of ChatGPT API usage requesting all explanations in accessible bullet-point format could cost 2-3x more than requesting dense text.
This is a real equity issue. Neurotypical users get efficient formatting by default. Neurodivergent users requesting accessibility pay a literal premium. Awareness of this dynamic is the first step toward mitigation.
Strategy 1: Compress accessibility. Request bullet points but ask the AI to minimize whitespace: "Use bullet points (to aid ADHD processing) but combine related points to reduce token count." This trades some visual separation for lower cost.
Strategy 2: Batch requests by format. Request dense text first, then ask for an "accessible summarization." The summary (expensive) is smaller because you're condensing established content rather than expanding from scratch. Token math: dense explanation (50 tokens) + accessible summary request (80 tokens) = 130 total. If you'd requested accessible from the start: 110 tokens. Not huge savings, but real.
Strategy 3: Use free or inclusive-tier tools. ChatGPT Plus ($20/month for unlimited usage) eliminates per-token cost, making accessibility decisions purely about cognition, not budget. Claude has a free tier with daily message limits. Notion AI charges per workspace, not per request. If your institution provides free access to Claude or ChatGPT, use it without token guilt.
Strategy 4: Pre-structure your requests to maximize token efficiency. Rather than "Explain this concept," ask: "Explain this concept in 5 bullet points, maximum 15 words per bullet." You're specifying compression upfront, which paradoxically increases accessibility (constraints force clarity) while reducing tokens.
Strategy 5: Export and reuse. When an AI provides a particularly accessible explanation of something you'll reference repeatedly, save it. Don't re-request; retrieve your local copy. This is especially valuable for frequently-accessed information like ADHD strategies, dyslexia accommodations, or core academic content.
AI pricing models generally penalize accessibility. This isn't accidental—it's baked into how language models work. Dense text is cheaper to generate than formatted text. Understanding this helps you make informed choices: sometimes the $0.10 cost of fully accessible formatting is worth it; sometimes you compromise slightly to reduce cost.
The most important realization: your neurodiversity-related accessibility needs are valid even if they cost more. If you're paying out of pocket and need to choose between accessibility and cost, seek free or institutional resources. Don't deprioritize your learning access to save money.
Try this: Take a prompt you commonly use. Write two versions: one efficient/dense, one in your preferred accessible format. Count tokens in both using a token counter (OpenAI has one; Claude's website displays token counts). Calculate the cost difference if you used the accessible version daily for a semester. Decide consciously whether that cost is justified for your learning, and explore free/inclusive-tier tools if budget is a barrier.
Peri can explain this concept, give practical examples, help you decide whether it applies to your situation, or recommend a journey if appropriate.
Explore related journeys or tell Peri what you're working through.