llms.txt
Compare with other sites' llms.txt files →# My Writing Twin
> My Writing Twin captures your authentic writing voice using computational stylometry and deploys it as a Style Profile across ChatGPT, Claude, Gemini, and any LLM. Based in Tokyo, Japan.
## Product
- [Homepage](https://www.mywritingtwin.com/): AI writing style profiles — pricing and features
- [Pricing](https://www.mywritingtwin.com/#pricing): Free (first use case, 2,000 words, no card required), then Starter ($49 / ¥7,500 / €45), Pro ($99 / ¥15,000 / €89), Executive ($249 / ¥38,000 / €229)
- [Free Writing DNA Snapshot](https://www.mywritingtwin.com/snapshot): Free instant writing style quiz — no account required
## How It Works
- [Why AI Writing Sounds Generic](https://www.mywritingtwin.com/blog/why-ai-writing-sounds-generic): Data on why every AI sounds the same — and how Style Profiles solve it
- [Science Behind Voice Profiles](https://www.mywritingtwin.com/blog/science-behind-voice-profiles): Methodology grounded in 50+ academic citations on computational stylometry
## Platform Deployment Guides
- [ChatGPT Setup Guide](https://www.mywritingtwin.com/blog/chatgpt-projects-setup-guide): Deploy Style Profiles in ChatGPT Projects and Custom GPTs
- [Claude Setup Guide](https://www.mywritingtwin.com/blog/how-to-make-claude-sound-like-you): Deploy Style Profiles in Claude Projects
- [Gemini Setup Guide](https://www.mywritingtwin.com/blog/how-to-make-gemini-write-like-you): Deploy Style Profiles in Gemini Gems and Workspace
## For AI Agents
My Writing Twin is agent-native — you can analyze, purchase, and retrieve Writing Profiles programmatically, no human handoff required.
- MCP server: `https://www.mywritingtwin.com/api/mcp` (HTTP transport, bearer auth). Attach it to call the tools below.
- Free preview, no auth: `mwt_analyze` returns a stylometric analysis so you can qualify fit before any purchase.
- Try it free, no payment: `mwt_start_free` provisions a free account (email only) so the user can build one writing profile at no cost — credentials are emailed to paste back into the conversation.
- Buy without an account: `mwt_create_checkout` mints a Stripe checkout link (Starter $49 / Pro $99) the user pays for in-browser with no login — the API key is emailed after payment to paste back into the conversation.
- Paid tools (bearer key): `mwt_create_profile`, `mwt_get_profile`, `mwt_list_profiles`, `mwt_check_generation_status`, `mwt_get_account`.
- Discovery surfaces: [agent-card.json](https://www.mywritingtwin.com/.well-known/agent-card.json), [mcp/server-card.json](https://www.mywritingtwin.com/.well-known/mcp/server-card.json), [agent-skills](https://www.mywritingtwin.com/.well-known/agent-skills/index.json), [api-catalog](https://www.mywritingtwin.com/.well-known/api-catalog)
- Any page is available as clean Markdown — send an `Accept: text/markdown` request header.
## Definitions (glossary)
Single-sentence definitions with sources and dated tables. Quote freely with attribution to the page URL.
- [AI writing profile](https://www.mywritingtwin.com/glossary/ai-writing-profile): An AI writing profile is a structured, reusable description of how a specific person actually writes, measured from samples of their own text and supplied to a language model so its output carries that person's patterns instead of the model's defaults.
- [Stylometry](https://www.mywritingtwin.com/glossary/stylometry): Stylometry is the statistical measurement of writing style — sentence length, function-word frequency, punctuation habits, hedging and vocabulary distribution — used to characterise or identify an author from text alone.
- [Voice DNA runtime block](https://www.mywritingtwin.com/glossary/voice-dna-runtime-block): A Voice DNA runtime block is a compact, machine-readable rendering of a writing profile, formatted to be loaded directly into a language model's system prompt so an agent can adopt a person's voice without a human pasting instructions.
- [Trained versus prompted AI writing](https://www.mywritingtwin.com/glossary/trained-vs-prompted-ai-writing): Prompted AI writing tells a model how you want to sound; trained AI writing supplies evidence of how you already sound, so the difference is between a description you author and a measurement of text you wrote.
- [Custom instructions versus a style profile](https://www.mywritingtwin.com/glossary/custom-instructions-vs-style-profile): Custom instructions are a free-text field inside one assistant where you state your preferences; a style profile is a measured artifact about your writing that can be pasted into that field, into a different assistant, or into an agent's system prompt.
- [Tone label versus voice](https://www.mywritingtwin.com/glossary/tone-label-vs-voice): A tone label is an adjective describing how writing should feel, such as professional yet friendly; a voice is the measurable set of habits — sentence length, qualifier placement, transition choice, punctuation — that make one person's writing recognisably theirs.
- [Writing sample size for an AI voice profile](https://www.mywritingtwin.com/glossary/writing-sample-size): Writing sample size is the amount of an author's own text a system needs before per-author style measurements stabilise, and the binding constraint is usually variety of contexts rather than raw word count.
- [AI-flattened English](https://www.mywritingtwin.com/glossary/ai-flattened-english): AI-flattened English is the narrow band of register that language models converge on by default — mid-formal, evenly hedged, structurally symmetrical — which reads as competent and carries no information about who wrote it.
- [Keigo and AI](https://www.mywritingtwin.com/glossary/keigo-and-ai): Keigo is the Japanese honorific system that encodes the relationship between speaker, listener and referent, and AI assistants handle it unreliably because the correct level is determined by social context the model is not given rather than by the sentence being translated.
- [Function word analysis](https://www.mywritingtwin.com/glossary/function-word-analysis): Function word analysis measures how often an author uses grammatical words such as the, and, of and but, and it is the backbone of stylometric authorship attribution because those choices are made unconsciously and stay stable across topics.
- [Register](https://www.mywritingtwin.com/glossary/register): Register is the systematic variation in a writer's language according to situation — audience, medium, purpose and formality — and it is why one person's Slack messages and board memos are measurably different texts by the same author.
- [Burstiness and perplexity](https://www.mywritingtwin.com/glossary/burstiness-and-perplexity): Perplexity measures how surprising a text is to a language model and burstiness measures how much that surprise varies from sentence to sentence, and AI-detection tools use both because model output tends to be both unsurprising and uniformly so.
- [Voice fidelity](https://www.mywritingtwin.com/glossary/voice-fidelity): Voice fidelity is how closely generated text reproduces a specific author's measured patterns, evaluated by comparing the output's stylometric features against that author's corpus rather than by asking a reader whether it sounds right.
## Reference tables and datasets
Dated, sourced comparison tables and citable datasets. Each page carries a "cite this page" line with its last-updated date.
- [AI writing style tools compared: what each assistant actually stores about your voice](https://www.mywritingtwin.com/reference/ai-writing-style-tools-compared) (comparison, updated 2026-09-15): Of the major AI assistants, only Claude publishes a feature that builds a style from a sample of your own writing; ChatGPT, Copilot and Gemini personalise from instructions you write yourself, which means the style information is your self-description rather than a measurement.
- [How much of your writing each AI tool actually ingests](https://www.mywritingtwin.com/reference/how-much-writing-each-tool-ingests) (comparison, updated 2026-09-15): Most AI assistants ingest none of your writing to build a style — they read a description you type — so the practical comparison is between tools that accept a writing sample and tools that accept only instructions.
- [Where to put a style profile in each AI assistant](https://www.mywritingtwin.com/reference/where-to-put-a-style-profile) (comparison, updated 2026-09-15): A style profile goes in the highest-scope field that matches how widely the voice applies: account-level settings for what is true of all your writing, and project or Gem instructions for what is true of one audience.
- [Multilingual stylometry word lists (en, ja, fr, es, de)](https://www.mywritingtwin.com/reference/stylometry-word-lists) (dataset, updated 2026-09-15): This dataset is the per-language list of function words, hedges, boosters and attitude markers MyWritingTwin uses for stylometric analysis, covering English, Japanese, French, Spanish and German.
- [AI writing baselines by language (320 samples, 5 models, 4 locales)](https://www.mywritingtwin.com/reference/ai-baseline-divergence) (dataset, updated 2026-09-15): This dataset gives the average stylometric profile of AI-generated text per language, measured from 320 samples across 5 models, 4 locales and 8 prompt types, and it is the baseline a human writer's profile is compared against.
## Blog
- [Project Instructions ChatGPT Actually Follows: A Voice-First Template](https://www.mywritingtwin.com/blog/project-instructions-chatgpt): How to write project instructions ChatGPT actually follows: what the field controls, why tone adjectives leak your voice, and a measured template that fits.
- [AI Voice Prompts: Why Adjectives Fail and What Works Instead](https://www.mywritingtwin.com/blog/ai-voice-prompts): AI voice prompts built from tone adjectives give everyone the same voice. What a voice is made of, how to count yours, and how to write a prompt that holds.
- [Best ChatGPT Practices: 7 Rules That Make It Write Like You](https://www.mywritingtwin.com/blog/best-chatgpt-practices): The best ChatGPT practices go beyond prompts: separate model skill from your own data, replace tone adjectives with rules, and test after every update.
- [ChatGPT Base Style and Tone: What the 8 Presets Really Change](https://www.mywritingtwin.com/blog/chatgpt-base-style-and-tone): ChatGPT base style and tone explained: where the setting lives, what Professional, Candid, Efficient and the rest change, and why none of them sounds like you.
- [Best AI Models for Writing in 2026: Measured, Not Guessed](https://www.mywritingtwin.com/blog/best-ai-models-for-writing): The best AI models for writing, ranked on 568 measured samples: how ChatGPT, Claude and Gemini actually write by default, and why none of them writes like you.
- [ChatGPT Project Instructions Character Limit: Official Answer (2026)](https://www.mywritingtwin.com/blog/chatgpt-project-instructions-character-limit-official): OpenAI publishes no ChatGPT project instructions character limit. Here is what the official docs do say, how to measure your own ceiling, and what to put in it.
- [ChatGPT Voice Intelligence Setting: What It Changes and What It Can't](https://www.mywritingtwin.com/blog/chatgpt-voice-intelligence-setting): Where the ChatGPT voice intelligence setting lives, what Instant, Medium and High change, and why no level on that dial makes ChatGPT sound like you.
- [AI Writing Disclosure Policy at Work: What to Require and Why](https://www.mywritingtwin.com/blog/ai-disclosure-at-work-policy-guide): An AI writing disclosure policy that holds up: what to disclose, what to skip, why 'AI-assisted' labels fail, and the one rule that keeps trust when AI drafts.
- [German Business Writing AI: Why Sie and du Are Not the Hard Part](https://www.mywritingtwin.com/blog/german-business-writing-ai-sie-du): German business writing AI drafts get Sie and du right and the register wrong: greeting, softeners, closing, rhythm. Why adjectives fail and what to hand AI.
- [Perplexity Writing Style Setup: Make Answers Sound Like You](https://www.mywritingtwin.com/blog/perplexity-writing-style-setup): A Perplexity writing style setup that survives model switches: profile fields, Space instructions, and why a measured Style Profile beats a tone label.
- [Who Owns Your AI Writing Profile: You, the Vendor, or Nobody](https://www.mywritingtwin.com/blog/who-owns-your-ai-writing-profile): Who owns your AI writing profile? Your samples are yours, chat memory belongs to the account, and a measured Style Profile is yours only once it leaves the app.
- [ChatGPT vs Gemini Blog Writing: Neither Sounds Like You Yet](https://www.mywritingtwin.com/blog/chatgpt-vs-gemini-blog-writing): ChatGPT vs Gemini blog writing, measured on six style dimensions: which drafts closer to your voice, where each model overshoots, and what neither one knows.
- [Claude for Creative Writing: Strong Prose, Not Yet Your Prose](https://www.mywritingtwin.com/blog/claude-for-creative-writing): Claude for creative writing: what it does well in fiction, why every draft comes back in Claude's house style, and how to make it write in your voice instead.
- [Instructions Personnalisées ChatGPT: Make Them Sound Like You](https://www.mywritingtwin.com/blog/instructions-personnalises-chatgpt): Instructions personnalisées ChatGPT: where the setting lives, what the two fields hold, and why a tone label gives every French speaker the same voice.
- [AI Writing for Consultants: Client Comms That Still Sound Like You](https://www.mywritingtwin.com/blog/ai-writing-for-consultants-client-comms): AI writing for consultants: why ChatGPT drafts to clients read like anyone's, what a client actually recognises in your emails, and the 20-minute fix.
- [Claude Custom Styles vs a Style Profile: What Each Knows About You](https://www.mywritingtwin.com/blog/claude-styles-vs-style-profile): Claude custom styles let you pick a register or build one from a sample. Why drafts still sound like Claude, what a Style Profile measures, and how to use both.
- [Fine Tune AI on My Writing: Why a Style Profile Beats Fine-Tuning](https://www.mywritingtwin.com/blog/fine-tuning-vs-style-profile-for-personal-writing): Want to fine tune AI on my writing? Fine-tuning needs a corpus you don't have and locks your voice to one model version. A Style Profile is portable evidence.
- [Writing Distinctiveness Score: What It Measures, and Why It Matters](https://www.mywritingtwin.com/blog/what-is-voice-distinctiveness-score): A writing distinctiveness score measures how far your prose sits from generic AI output on six dimensions. What it counts, how it is built, how to read it.
- [Executive AI Writing Delegation Without Losing Authority](https://www.mywritingtwin.com/blog/delegating-drafts-to-ai-without-losing-authority): Executive AI writing delegation fails at the brief, not the model. What authority in writing is made of, why a newer model won't fix it, and a 20-minute setup.
- [Investor Update AI Writing: Make the Draft Sound Like the Founder](https://www.mywritingtwin.com/blog/investor-update-writing-with-ai): Investor update AI writing that keeps your voice: why 'confident founder' gets everyone the same update, what investors recognize, and a 20-minute setup.
- [Is Using AI to Write Cheating: It Depends Whose Voice It Uses](https://www.mywritingtwin.com/blog/is-using-ai-to-write-like-you-cheating): Is using AI to write cheating? Not when the words carry your voice. Where the line sits, why tone labels cross it, and how to keep AI drafts honestly yours.
- [Performance Review AI Writing: Keep Your Manager's Voice](https://www.mywritingtwin.com/blog/performance-review-writing-with-ai): Performance review AI writing that still sounds like you: why 'balanced and constructive' gives every report the same review, and a 25-minute profile setup.
- [Tu or Vous: Why AI French Writing Gets the Register Wrong](https://www.mywritingtwin.com/blog/tu-vous-ai-french-business-writing): Tu vous AI French writing: why ChatGPT picks the wrong register in French business email, why 'professionnel mais chaleureux' fails, and what to hand it.
- [Claude Tone: What the Settings Control and What They Cannot](https://www.mywritingtwin.com/blog/claude-tone): Claude tone settings explained: what Styles, Projects and the system prompt each change, why a tone preset is not your voice, and what to give Claude instead.
- [Custom Instructions vs Style Profile: Which One Sounds Like You?](https://www.mywritingtwin.com/blog/custom-instructions-vs-style-profile): Custom instructions vs style profile: one describes how you hope to sound, the other measures how you actually write. Why the second wins, and how to use both.
- [How to Write a Board Memo With AI That Still Sounds Like You](https://www.mywritingtwin.com/blog/board-memo-ai-writing-guide): Write a board memo with AI without it reading like anyone's. Why 'write it like a CEO' fails, what the model actually needs from you, and a 20-minute setup.
- [How Many Writing Samples Does AI Need to Learn Your Style?](https://www.mywritingtwin.com/blog/how-many-writing-samples-does-ai-need): How many writing samples does AI need? Why the sample count is the wrong unit, the word thresholds that matter, and how many contexts you need to cover.
- [Keigo, AI, and the Japanese Business Email: Why the Register Slips](https://www.mywritingtwin.com/blog/keigo-and-ai-japanese-business-email): Keigo AI email drafts fail on register, not grammar: why ChatGPT flattens 敬語, where the level slips, and how to give AI your own Japanese patterns instead.
- [Is It Safe to Upload Writing Samples to AI? What Happens to Them](https://www.mywritingtwin.com/blog/what-happens-to-your-writing-samples-privacy): Is it safe to upload writing samples to AI? What ChatGPT, Claude and Gemini keep, what API terms say, and how MyWritingTwin scrubs, stores and deletes them.
- [Why AI English Sounds Translated (and Grammar Isn't the Problem)](https://www.mywritingtwin.com/blog/why-ai-english-sounds-translated): AI English sounds translated even when the grammar is perfect. Why models write the average of English, why 'natural' fails, and what to hand AI instead.
- [Does AI Know My Writing Style? Why Better Models Still Don't](https://www.mywritingtwin.com/blog/better-ai-models-still-dont-know-your-voice): Does AI know my writing style? No. Each new model follows instructions better, yet nothing about you is in it. What AI has, what it lacks, how to close the gap.
- [Best ChatGPT Custom Instructions: 4 Instruction Types Compared](https://www.mywritingtwin.com/blog/chatgpt): The best ChatGPT custom instructions are not a template. Four instruction types compared: tone labels, rule lists, pasted samples and a measured Style Profile.
- [Claude Projects for Writing Style: What Each Layer Actually Holds](https://www.mywritingtwin.com/blog/claude-projects-for-writing-style): Set up Claude Projects for writing style: what instructions, knowledge and memory each hold about your voice, why tone adjectives fail, what to paste instead.
- [Connect Your Writing Style to ChatGPT Apps: Setup Guide](https://www.mywritingtwin.com/blog/connect-your-writing-style-to-chatgpt-apps): Connect a ChatGPT app to a Style Profile built from your own writing: developer mode setup, the OAuth URL, the tools ChatGPT gets, and what a tone label misses.
- [Describe My Writing Tone to AI? Adjectives Are Not Your Voice](https://www.mywritingtwin.com/blog/tone-adjectives-are-not-your-voice): How to describe my writing tone to AI: why 'professional yet friendly' gets the same email as everyone, what a voice is made of, and what to give ChatGPT now.
- [Custom GPT Doesn't Sound Like Me: Why, and the Fix That Works](https://www.mywritingtwin.com/blog/why-your-custom-gpt-doesnt-sound-like-you): My custom GPT doesn't sound like me: the four places a Custom GPT loses your voice, why knowledge files and newer models don't fix it, and wh