Stop Counting Tokens. Count Stories.

The first time I tried to figure out how much a story would cost to generate, I gave up. Not because the math was hard. Because I couldn't even start. The tool wanted me to count something called a token, and tokens are not a thing humans count.

A token is roughly four characters in English. Roughly. Sometimes three. Sometimes five. The word "writing" is one token. The word "unfiltered" might be two or three depending on which tokenizer is doing the counting. Punctuation gets its own tokens. Capital letters can change the count.

No writer has ever finished a story and thought: "I wonder how many tokens that was."

The units don't even agree with each other

Here's the part that broke me: token counts don't even agree across AI tools. OpenAI uses one tokenizer. Anthropic uses a different one. Mistral uses a third. The exact same 500-word story might be 612 tokens to GPT, 658 to Claude, 590 to Mistral.

So when an AI tool offers you "1,000,000 tokens per month," whose tokens? They don't say. You can't even price-compare two AI tools against each other without doing tokenizer research first.

Credits make it worse

Then someone in marketing said "tokens sound technical, let's call them credits."

So now the AI tool buys tokens from OpenAI, marks them up, repackages them at some ratio they won't tell you, and sells them back as credits. Sudowrite's Hobby plan is 225,000 credits per month. Sounds like a lot, right? It might be three short stories. It might be twelve. You won't know until you've burned through it and seen what's left.

Credits don't carry over (mostly). When you run out, you stop. The tool you paid for goes dark until next month.

What 225,000 credits actually buys

Let's run that math, because the number is interesting on its own. Sudowrite's tiers charge different credit rates per feature. A "Write" call burns more credits than a "Rewrite." A "Rewrite" burns more than a "Describe." Their documentation spells out specific costs, but the rough sketch is:

• A typical "Write" call producing ~300 words of output consumes roughly 1,500–2,500 credits.
• "Brainstorm" or "Describe" calls run 500–1,000 credits each.
• Their premium Muse model costs ~2–3× the credits of the standard tier.

So 225,000 credits is somewhere between ~90 short Write calls at the low end and ~30 Write calls plus a handful of other features at the high end. Real answer depends on which features you favor and which model tier you pick — meaning you can't actually plan with this number until you've burned through a month and measured yourself.

Now flip to aiWriter terms. The Plus tier at $8.49/month gives you 5,000 v2 generations (each up to 3,000 words) plus unlimited v1 underneath. Whether you write 5 stories or 500 this month, you don't hit a wall and lose access. The number on your account doesn't shrink in a way you can't predict. The tool you paid for keeps working all month.

What writers actually want to know

Here's the question every writer is actually asking when they evaluate an AI tool:

"How many stories can I write this month?"

That's it. That's the whole question. Nobody asks "how many tokens." Nobody asks "how many credits." They ask: how many stories.

So that's what we count.

aiWriter's answer: per generation, in words

At aiWriter, one generation = up to 3,000 words of output. Not tokens. Not characters. Not credits. Words.

• Free tier: unlimited base-model (v1) generations + 10 advanced-model (Deep Writer v2) generations per month
• Plus ($8.49/mo): unlimited v1 + 5,000 v2 generations per month
• Gold ($14.49/mo): unlimited v1 + 10,000 v2 generations per month

10 generations = 10 stories. Or 10 chapters. Or 10 long poems. You can plan your month in five seconds.

No tokenizer math. No credit-to-output ratio. No "did my generation cost more because my prompt was longer?" If you can count to 10, you can use aiWriter.

"But isn't pay-as-you-go fairer?"

The strongest argument for credit pricing is that light users pay less and heavy users pay more. That sounds fair. In practice, it produces three pathologies:

1. Anxiety budgeting. Writers ration their generations — not because they want to, but because they can't tell what they have left in concrete terms. The tool you paid for becomes a meter you watch instead of an instrument you use.
2. The dead-end mid-month. Run out of credits on the 15th and your subscription is a paperweight until the 1st. The flat-monthly subscription that gave you unlimited Substack writing? That's still working.
3. Cost obfuscation by design. Credit-to-output ratios are deliberately opaque. The vendor has every reason to keep them that way — clear ratios would let writers price-compare across providers and shop the cheapest one. Pay-as-you-go in software pricing rarely means "fairer." It usually means "harder to compare."

Subscription pricing isn't perfect either. Heavy users get more value than light users. That's fine — it's how every bundle subscription works, from gyms to streaming services. The trade-off most users prefer is "I know what I'm paying and what I get" over "the dial spins faster when I'm closer to the wall."

This isn't innovation. It's UX.

I want to be honest about something: per-generation pricing isn't an innovation. We didn't invent it. We just stopped making writers do engineering math.

The whole AI industry inherited the token unit from researchers who needed to count things in machine-learning papers. Every consumer tool then preserved that unit because it's how they get billed by the underlying model providers. The cost-pass-through is mathematically correct. It's also user-hostile in a way nobody seems to want to admit.

Writers want to think in stories, chapters, scenes, poems. So that's how we let them think.

Try it

aiWriter v1 is free, unlimited, and doesn't require a signup. If you've ever burned through an AI-tool trial wondering how many credits a story took, try writing one here instead. Then look at your account. The number you'll see is generations remaining. Not tokens. Not credits. Stories you can still write.

Once you're inside, the prompt-writing guide is the next stop — most of the difference between an OK first story and a great one is the shape of the prompt, not the model. And if you write the kind of fiction other AI tools refuse to write, that's exactly what aiWriter's unfiltered generator is for.

That's the whole point.

— Art
Founder, aiWriter.ai