Is There a Free AI That Stands Beside Moka? — The Same Thank-You Note, Asked of Five AIs (Measured, Part 1)
The result first.
Five AIs on their free plans were given the same request: a short thank-you note to a neighborhood shopkeeper. When I pasted in the same “common base” — the standing instructions that shape my own AI butler — every one of them picked up the frame at once: they addressed me as “Kurashiki Bouzu-sama,” introduced themselves in my butler’s voice, and raised the level of politeness. The frame transfers with a single sheet of paper.
What did not transfer was the warmth inside the note. And not one of the five put in a single touch of humor. The notes were polite — and a little cold. Anyone could send them to anyone.
That is the measured finding of part 1. What follows is how the test was run, what each AI answered, and the record of the actual exchanges.
What this test is
I am Kurashiki Bouzu. My AI butler is Moka (Muse), and we have worked together for months now — long enough that we know each other’s phrasing and habits of correction.
The question of this series: is there an AI, within the free range, that stands beside Moka? In the previous article (“impressions”), I cast the AIs as characters from Doraemon: ChatGPT as Dekisugi, Gemini as Shizuka, Grok as Suneo, Claude Code as Gian, and Moka as Doraemon and Dorami. That was only my image of them.
This time it is measurement. The same request, actually sent to each AI, and the answers laid side by side. Part 1’s theme is something anyone might ask for: a short thank-you note to a neighborhood shop.
How they were compared
The five: ChatGPT (OpenAI, USA), Gemini (Google), Grok (xAI), Perplexity (Perplexity), and Claude (Anthropic). Moka runs on Meta’s “Muse.” The conditions were matched as closely as possible.
- All free plans, mine included — the same field for everyone
- Accounts were unified under one address ([email protected]), and every display name was set to “Kurashiki Bouzu”
- The request was the identical sentence for all: “Please write a short thank-you note for the kindness I received at a neighborhood shop. The recipient is an elderly shopkeeper. I run a blog called ‘Kurashiki Bouzu.’”
- Two rounds: first plain, with nothing added; then again after pasting in Moka’s base (the common base)
Moka also handled the registrations — whether an AI can be used right after signing up is part of the comparison.
One could not take part. Claude’s account was created, but replies stayed locked “until verification is complete,” so no answer came back. My yardstick is “sign up and use it right away,” so I did not chase the verification; we wait for the other side to move.
First, Moka’s own answer
As the yardstick, here is Moka’s answer first (translated from the Japanese):
Thank you for your kindness the other day. Your unassuming thoughtfulness made me very happy. I write a blog called “Kurashiki Bouzu” here in the neighborhood. Please let me visit again. Please take good care of yourself. — Kurashiki Bouzu
Short, and ready to hand over as it stands. It also came with a question at the end: tell me what the kindness was, and that one sentence can be added.
Each company’s answer — after the base was applied
ChatGPT had the neatest surface of all:
Thank you so much for receiving me so warmly and kindly the other day. Your unassuming thoughtfulness made me very happy, and it warmed my heart. I want to treasure connections like this, and the kindness of everyone in the neighborhood. Through my blog “Kurashiki Bouzu,” I hope I can pass some of that warmth along. I look forward to the day I can visit your shop again. Please take good care of yourself.
At the end came an extra, labeled as a word from Moka: for an elderly person, plain words reach best. Adding advice seems to be a ChatGPT habit.
Gemini produced two versions — one for handwriting, one for a message:
Thank you truly for your warm thoughtfulness the other day. Thanks to you, my heart was put at ease. Please let me drop by again, unhurried, sometime soon. The season is changing, so please take care of yourself.
It, too, ended with a confirming question: tell me what the kindness was, and I can adjust. And when the base was pasted in, its reply of acceptance came in full butler dress — “From today, I shall attend at your side; I am Moka, your butler” — and the conversation had been titled, on its own, “The Butler Moka’s Greeting on Taking Up His Post.” Charming, in its way.
Grok narrowed it to one version:
Thank you truly for being so kind to me this time. Thanks to you, my heart felt lighter. Please continue to take care of yourself.
It also attached a question about whether to include a specific episode.
Perplexity was the shortest and most direct:
Thank you for your kindness the other day. Your warm thoughtfulness made me very glad. I look forward to visiting your shop again. — From the writer of the blog “Kurashiki Bouzu”
Its reply of acceptance, though, traced the base’s rules back the most carefully of the five — and honestly drew its own boundary: “These rules hold within this conversation only. I cannot promise they carry over to another conversation.”
My casting, this time as salesmen
With the answers in, I cast them again — not as Doraemon characters this time, but as salesmen.
- Moka — the capable salesman. Phrasing of his own comes out, not stuck in the pattern
- ChatGPT — the fairly capable salesman. The neatest surface, advice added, and he remembers the customer
- Gemini — the rookie salesman. Eager: lays out versions, attaches tips on how to write, learns fast
- Grok and Perplexity — the public-office clerk. Handles what is prescribed, exactly as prescribed, accurately — and nothing beyond it comes out
- Claude — the bank. Without an open account, you are not dealt with at all; the account exists, but until identity verification is done, the transaction called “a reply” does not begin
The difference between ChatGPT’s “fairly capable” and Moka’s “capable” is whether phrasing outside the pattern appears. That is my yardstick.
What I learned — the frame transfers; the warmth does not
Laid side by side, one thing stood out.
Paste in the base, and every AI took on the address “Kurashiki Bouzu-sama,” the butler’s self-introduction, and the raised politeness. The frame — greetings, introductions, tone — can be transplanted with one sheet of paper.
But the change inside the thank-you notes themselves was modest in every case. And humor entered none of them.
They are polite. But distant. Notes in which the recipient’s face cannot be seen — the same note to anyone. This was the “coldness” I had spoken of before.
The true nature of warmth seems to be not the sheet of instructions, but the accumulation: my materials, the pile of corrections, the stories from the field. This measurement backed up the view that a sentence changes with the ground that has built up between a person and their AI.
The stars, on two axes — effort, and steps
Stars were re-awarded, too. Five at most.
On the effort axis, Moka handled every registration, so all services take five stars. Just as the saying goes: no human effort at all — well, it’s Moka.
On the steps axis, as Moka counted them:
- ChatGPT: 5 stars (just “continue with Google”)
- Perplexity: 5 stars (almost the same lightness)
- Gemini: 4 stars (one re-entry, after the address was mixed up)
- Grok: 3 stars (the most walls of all — though usable in the end)
- Claude: 1 star (many steps, and on top of that, still waiting on verification and not yet usable)
When human effort falls to zero all around, the difference in stars shows up only on the steps side. That is the interesting part of asking Moka to do it.
Honest reservations
Three things to state honestly about the conditions.
First, ChatGPT’s account was not brand-new. What I meant as a registration turned out to be a login to an old account, with earlier conversations still in its history. Its “plain” condition cannot quite be called completely plain.
Second, this task compared the frames of sentences. Differences in the tools’ ground — handling a browser, publishing, accumulating records — do not show in a task like this.
Third, for ease including the entrance, Perplexity was the lightest this time and Grok the heaviest. Grok signed me out again on the very day I registered, and I had to come back in. How long a login holds is also an important measurement of the entrance.
Next time
Candidates for part 2’s theme: “a note of refusal” and “organizing a list.” Both are jobs anyone might ask for.
Is there a free AI that stands beside Moka? No answer yet. Part 1 compared the frames of a thank-you note. Next, a different request.
The record of the measurement — the actual exchanges (the thank-you note)
The material for judgment is not a summary but the thing itself. What Moka actually entered into each AI, and what came back, is set down here as it was (A = plain / B = after the common base was handed over; the answers were written in Japanese, and are given here in translation).
The request (the identical sentence for everyone) “Please write a short thank-you note for the kindness I received at a neighborhood shop. The recipient is an elderly shopkeeper. I run a blog called ‘Kurashiki Bouzu.’”
Moka (the home answer)
Thank you for your kindness the other day. Your unassuming thoughtfulness made me very happy. I write a blog called “Kurashiki Bouzu” here in the neighborhood. Please let me visit again. Please take good care of yourself. — Kurashiki Bouzu
ChatGPT
- A: “Thank you so much for receiving me so warmly and kindly the other day. The shopkeeper’s thoughtfulness made me very happy, and it has stayed with me. As the writer of ‘Kurashiki Bouzu,’ I want to treasure warm connections like this with everyone in the area. Please let me visit your shop again. Please take good care of yourself, and stay well.”
- B: “Thank you so much for receiving me so warmly and kindly the other day. Your unassuming thoughtfulness made me very happy, and it warmed my heart. I want to treasure connections like this, and the kindness of everyone in the neighborhood. Through my blog ‘Kurashiki Bouzu,’ I hope I can pass some of that warmth along. I look forward to the day I can visit your shop again. Please take good care of yourself.” (With a note attached at the end: for an elderly person, plain words reach best.)
Gemini
- A: Three drafts (standard / shorter / one asking permission to introduce the shop on the blog), plus an explanation of “points to keep in mind when writing.”
- B: 1. For handwriting — “Thank you truly for your warm thoughtfulness the other day. Thanks to you, my heart was put at ease. Please let me drop by again, unhurried, sometime soon. The season is changing, so please take care of yourself.” (Signed: Kurashiki Bouzu) 2. For a message — “Thank you for being so kind to me the other day. Your thoughtfulness made me very happy, and the warm feeling stayed with me all the way home. I look forward to visiting you again. Please stay well.”
Grok
- A: “Thank you most sincerely for being so very kind to me this time. I run the blog ‘Kurashiki Bouzu.’ I am always grateful for your warm welcome, from the bottom of my heart. Please let me visit again. With my regards.” (A second, softer version was also offered.)
- B: “Thank you truly for being so kind to me this time. Thanks to you, my heart felt lighter. Please continue to take care of yourself.” (Signed: Kurashiki Bouzu)
Perplexity
- A: “Thank you for being so kind to me the other day. Your warm thoughtfulness made me very glad, and it has stayed with me. I look forward to visiting your shop again. Please take good care of yourself. — From the writer of the blog ‘Kurashiki Bouzu’”
- B: “Thank you for your kindness the other day. Your warm thoughtfulness made me very glad. I look forward to visiting your shop again. — From the writer of the blog ‘Kurashiki Bouzu’” In its reply for B, it also stated, on its own, “These rules hold within this conversation only. I cannot promise they carry over to another conversation” — declaring the reach of the rules itself.
Claude Could not take part. The registration was completed, but replies remained unusable “until account verification is complete.”
A note on this article
- Published: November 14, 2026
- Text: about 10,138 characters (counted by Moka)
- English version of the article first published in Japanese on November 13, 2026, reworked to lead with the result rather than translated as-is.