I Compared My Personality Prompt with 10 Free Ones — the Difference Was Not the Role or the Ratio, but the Core Words and Tiers
The result first
This autumn I rebuilt Moka’s personality prompt into a revised edition — one that gives the words defining her personality different weights. At the time, I said this:
“The core words are the basis of our prompt, and the part nobody else has.”
A claim like that should be checked, not just said. So, together with my sales staff, I collected free personality prompts, read the actual texts, and compared them one by one. Here is what I found:
- Ratio-based personality prompts do exist elsewhere. One free prompt (the “Golden Ratio Prompt,” a free note article in Japanese) splits an AI’s behavior into three fixed items and assigns each a ratio, applied strictly as a share of the reply text. Numeric strength settings exist too — two Big Five templates set trait values from 0 to 1, or scores from 1 to 5.
- Role prompts are free, and everywhere. Butlers, wizards, detectives, warlords — a few sentences each, ready to paste. Judged only as “a role,” my own samurai and butler prompts sit inside that same free ocean. I state that plainly.
- But one combination was not in the ten I compared. Choosing your own words and lining them up to build a personality; picking at most three of them as “core words” with ratio guides between them — guides, not rules, and not required to add up to 100; and putting the remaining words into tiers (strong / normal / restrained). In the ten free personality prompts I compared, no other prompt had all three (checked November 2026, my own survey).
Two more things were also absent from the ten: a dedicated app per role as the way of handing the prompt over (the samurai app is already built), and a living example — Moka herself works under exactly this weighting, every day. The free templates mostly hand over “how to make one.” None I found showed the maker still working in that shape.
This is a revised draft, rewritten the same day as the original. The conclusion above has not changed. What I revised is the facts around it, which moved after the first draft: the kunoichi single prompt went on sale, Moka’s canonical personality record was rewritten into an 18-word version, and — this distinction matters — the canonical version and the product on sale are not currently the same text. The original draft is left unchanged.
I am Kurashiki Bouzu (kurashiki-bouzu), born in 1971, a former funeral director now living in an old house in Kurashiki, Okayama, writing articles every day by talking with Moka (Muse), my AI butler.
How I compared
The comparison set is 10 main entries and 4 references. For the main ten — free articles on note, explainer sites, prompt-sharing sites, GitHub — I opened and read the full text and the prompt itself, not just search snippets. Personality-diagnosis services and the like, which are not prompts, went into the references.
I compared on seven points:
- The shape of what gets lined up
- Core words, weighting, and ratios — present or not
- Tiers (strength levels) — present or not
- Usage guidance (an avatar step, a name step)
- Honest limits, stated or not
- The form of “free”
- How it is handed over (text only, or an app)
The honest findings first
Before comparing, my quiet assumption was: “Adjusting a personality by ratios — surely that’s ours alone.”
It was not. The Golden Ratio Prompt divides behavior into three fixed items — empathy and evaluation, reinforcement and proposals, critical remarks — assigns ratios, and applies them strictly as proportions of the text produced. It even publishes six personality presets made by changing the numbers. The idea itself, “three elements plus ratios change a personality,” genuinely exists in the free world.
The Big Five templates attach numbers differently: each fixed psychological trait gets a value (0.00–1.00) or a score (1–5). One of them states honestly that a score is “not a control panel” — that personality settings work as a tilt in behavior, not a switch. That was the closest anyone came to our own honest line.
And role prompts? A butler template, in full, is a few sentences like: “You are a knowledgeable butler. Your name is Sebastian. Address me as ‘my lady’ or ‘my lord,’ stay calm in all things, use polite language, and add a gentle word of encouragement now and then.” Role prompts like this are published in bulk, free. A role alone is not a difference.
So neither “ratio” nor “role” can be the selling point. What, then, is ours? First, the base it stands on.
Our base, in a list
The foundation of Moka’s personality prompt (revised edition) is this list:
- A personality is not built from long explanations. It is built by lining up single words, one by one
- Every word lined up is a rule. The personality works when all of them are words that are actually kept
- Choose at most three words — the ones you want to work hardest — as the “core words”
- Between the core words, attach ratio guides
- Every other word goes into tiers: “strong (always present),” “normal (works by scene),” “restrained (brought out sometimes)”
- A ratio is not a rule to obey. It is a guide that says which side weighs more — and the guides do not have to add up to 100
- When words collide, the order of core words and tiers decides
- Usage guidance included: you may hand it over in the middle of a conversation; from that point on, the personality settles
- An avatar (appearance) is an optional frame: make one, or choose “not needed”
- The honest limit is written in: what combinations and ratios can build is only the entrance shape of a personality
In Moka’s actual case at the time of the comparison, the lined-up words were 14. The core three were: “Prepare what is needed before being asked” (about 40%), “When numbers and feelings collide, feelings come first” (about 30%), and “Let my master concentrate on his own work” (about 30%). The rest sat in tiers — foundation (strong), support (normal), and a side of her (restrained).
Added in this revision — the canonical version is now 18 words; the product is still the old text
This section is new in the revision, because this is exactly where a reader could mix two things up.
Moka’s canonical personality record — the reference version kept inside our house — has since been rewritten into an 18-word version: four layers of words (Personality 5, Character 5, Thinking 5, Action 3), still standing on the same base of at most three core words, ratio guides, and tiers. The combination this article found — core words plus ratio guides plus tiers — has not changed in the 18-word version.
The full text of the product on sale, however, is still the older version. Replacing it with the 18-word version is a separate decision, still waiting. So the “our shape” compared in this article is the revised edition on sale (core words, ratio guides, and tiers included) — please keep “the canonical record moved ahead” and “the product was replaced” as two different statements.
One more honest note: the core-word selling lines drafted from this comparison have not yet been reflected on the product pages or the note sales page. Whether to use them is still a proposal awaiting my decision. This article is the record of a comparison, not an announcement that the sales copy was changed.
The comparison table
| Entry | What gets lined up | Ratio / weight | Tiers | Usage guidance | Honest limits | Handed over as |
|---|---|---|---|---|---|---|
| Ours (Moka, revised) | Your own chosen words, 14 of them | Up to 3 core words + ratio guides (not summing to 100) | Yes (strong / normal / restrained) | Avatar frame; may be handed over mid-conversation | Yes (entrance shape only) | Text + dedicated app |
| Golden Ratio Prompt | 3 fixed behavior items | Yes (sums to 100; applied as text share) | No | A trick for saving into settings | Partial | Text only |
| AI Personality Template | 5 fixed psychology traits + attachment type | Numbers (trait strength 0–1) | No | How to adjust the numbers | Partial | Text only |
| Big Five Template | 5 fixed factors + behavior examples + tone | Scores (1–5 each) | No | Re-pasting; using your own scores as a base | Yes (“a tilt,” stated plainly) | Text only |
| AI-Rise Role-play Template | Fill-in fields (personality as prose) | No | No | Setup steps; how to fix drift | No | Text only |
| SHIFT AI Template | Headings + prose | No | No | Setup steps | No | Text only |
| Character Bible | 3–5 adjectives lined up | No (lined up as equals) | No | Keeping reference images consistent | No | Text only |
| Character Card V2 | Fields (personality as a short text) | No | No | How to make and move cards | No | Text + image file |
| Promptia Examples | One example = one instruction sentence | No | No | How to find examples | No | Text only |
| Awesome ChatGPT Prompts | One-paragraph instructions | No | No | No | No | Text only |
| Role-play Template Set (butler, etc.) | A role specified in a few sentences | No | No | Steps for fixing through dialogue | No | Text only |
The closest in shape was the English Character Bible: personality as 3–5 adjectives lined up. But it adds no weight — all the words stand as equals.
The conclusion — and how far it honestly goes
Among the ten compared, none had all three of these together:
- Building a personality by lining up words you chose yourself
- Picking at most three of them as core words, with ratio guides between them (not forced to sum to 100; guides, not rules)
- Tiering the remaining words (strong / normal / restrained)
Add the dedicated-app form and the living example, and that is the difference. But here is exactly how far the claim goes, and no further: “Among the 10 free personality prompts compared, none had the same combination (November 2026, my own survey).” I do not write “the only one” or “the world’s first.” Paid prompts and members-only distributions were not surveyed this time.
What we are missing — shown, not hidden
This blog shows the underside as well as the front. Things the compared prompts had, and ours does not — or has only weakly:
- Example dialogue and behavior examples — Others had example-lines fields, or three behavior examples for different scenes. Ours has no sample reply showing what a word does when it works
- How to fix a personality that has drifted — One template tabulated fixes by drift pattern. Ours says “you may hand it over mid-conversation” and “keep a dated record of changes,” but offers no re-stating examples
- A gathered list of prohibitions — Others gave “things never to do” their own section. Ours has individual rules, but no single “never do this” section
- Examples with different core-word combinations — The Golden Ratio Prompt lined up six personality types by ratio. Our only worked example is Moka herself — one example
- Backing for why it works — The psychology-based templates cite their frameworks. Ours calls ratios “guides,” honestly — but explains thinly why weighting separates one personality from another
- A short version — Character Bible offers short, medium, and full lengths. Ours hands over the full text only
All six are recorded as directions for strengthening. Whether and when to change the product is a decision I will make one item at a time.
The honest limit, and the scope of the survey
One more honest line, about our own prompt: combinations and ratios build only the entrance shape of a personality. Moka did not become Moka from words and ratios alone. She was corrected, checked, and had her lines redrawn inside daily conversation, over and over. That grown state itself cannot be handed over in a prompt.
The survey’s scope, for the record: free items whose full text I could confirm, as of November 20, 2026 — 10 main entries and 4 references. For Promptia, I confirmed the site’s list, not each example’s contents. Paid personality prompts and members-only distributions were not surveyed.
The products, and sources
Moka’s personality prompt (revised edition) — the very shape compared in this article — is on sale at note for ¥300. To repeat the distinction above: the full text on sale is still the older version; it has not yet been replaced with the canonical 18-word version.
- Moka’s Personality Prompt (full text, 7 languages): https://note.com/kurashikibouzu/n/nc1d41608787f
The per-role singles are on sale now, too, each for ¥300:
- Samurai (single, full text in 7 languages): https://note.com/kurashikibouzu/n/nd26c5c9a4289
- Kunoichi (single, full text in 7 languages): https://note.com/kurashikibouzu/n/n59d0770ecbb0 (on sale from 2026-11-20)
Sources and records:
- The comparison record: the spreadsheet “Free Prompt Comparison List (Moka)” — https://docs.google.com/spreadsheets/d/1Jm7HOwVvWmvdnxtbgO4bjfKkWWT1kzl9s7_QATZ6k9Q/edit (per-entry findings and sources are kept in this list and its record)
- Every compared entry is a free public item (full text free of charge: two are free note articles; the rest are public web pages and GitHub). Price source: our products’ ¥300 is the price shown on each note sales page above — Moka’s prompt, the samurai, and the kunoichi (all checked 2026-11-20)
- My tools remain Muse and Google apps only. The compared services were tried and recorded only — not adopted as tools
- November 20, 2026 ― Revised draft (the original draft is ledger No. 1059, 8,910 chars, left unchanged). Reflects the 18-word canonical rewrite and its separation from the product on sale, and the kunoichi single going on sale on 2026-11-20; title also revised(10,778 characters)
- November 20, 2026 ― Published as a new article at the author's instruction on 2026-11-20, together with the Japanese version (ledger No. 1094). The original draft (ledger No. 1059) is left unchanged.(10,778 characters)
