The Secret Behind note's AI Translation: Why Likes and Views Don't Matter?

Introduction

Hello, this is Easygoing.

Today I’d like to share some thoughts on note’s AI auto-translation service, which launched on May 27, 2026.

English Is the Global Standard Language

As of 2026, English is the most widely used language in the world, spoken by 1 in 6 people on the planet.

While English has fewer native speakers than Chinese, it is taught as a major foreign language in many countries, including Japan and China.

English web pages tend to receive far more traffic than pages in other languages. The world is hungry for English content.

note Is Japan’s Largest Content Distribution Platform

note continues to grow rapidly as a content distribution platform.

In June 2026, note’s membership surpassed 12.48 million.

The number of published pieces on note has exceeded 82.09 million and continues to expand, with a massive volume of new content being added every day.

note’s Platform Growth

November 2019: Migrated to the note.com domain, which is stronger for English-language search

June 2024: Became the top domestic blog site by traffic

July 2026: note ranked #2 among Japanese domains cited by AI according to an Ahrefs survey

note has become Japan’s largest content distribution platform both as a blog site and as a domain frequently cited by AI. Now, as a foothold for its challenge to the world, it has begun offering AI-powered auto-translation.

Actual Translated Pages

Let’s take a look at a page that has actually been auto-translated on note.

note’s English pages have almost the same design as the original Japanese pages, but a notice (NOTICE) appears at the very top indicating that the page is an AI auto-translation.

English and Japanese Are Separate Pages

note’s English-translated pages exist as different pages from the Japanese originals.

URL

• Japanese page:

https://note.com/ai_image_journey/n/n0e3ea469d98a

• English page:

https://note.com/ai_image_journey/n/n0e3ea469d98a?hl=en

English pages have ?hl=en appended to the end of the URL.

link Tags

• Japanese page

<link rel="canonical" href="https://note.com/ai_image_journey/n/n0e3ea469d98a">
<link rel="alternate" hreflang="ja" href="https://note.com/ai_image_journey/n/n0e3ea469d98a">
<link rel="alternate" hreflang="en" href="https://note.com/ai_image_journey/n/n0e3ea469d98a?hl=en">

English pages have ?hl=en appended to the end of the URL.

• English page

<link rel="canonical" href="https://note.com/ai_image_journey/n/n0e3ea469d98a?hl=en">
<link rel="alternate" hreflang="ja" href="https://note.com/ai_image_journey/n/n0e3ea469d98a">
<link rel="alternate" hreflang="en" href="https://note.com/ai_image_journey/n/n0e3ea469d98a?hl=en">

The link tags indicate the existence of the English version.

note’s English pages are recognized as separate pages from the original Japanese ones, while also signaling that an equivalent Japanese page exists.

It appears that note intends for these English pages to receive preferential treatment in search as English content.

note Selects Only Some Articles for English Translation

note does not translate every piece of content created by creators into English.

I started writing on note in July 2024 and have published 121 articles so far. Of those, only 38 articles (31.4%) have been auto-translated into English.

note is selectively choosing some articles from its vast library of content to translate into English.

note Is Supported by a Small Number of Paying Users

A February 2025 survey revealed that more than half of note’s users use the platform without paying.

The number of users who have made a purchase is less than half the total membership.

Because note is an ad-free platform, the server costs for the large majority of non-paying users are covered by fees from a small number of paying users.

If note were to translate every single article into English, the volume of content would double, and server costs would balloon accordingly.

When introducing AI auto-translation, note appears to have taken a quality-over-quantity approach for English content.

The Number of Likes Does Not Affect Auto-Translation

So what criteria does note use to select articles for English translation?

The first thing that comes to mind is the number of Likes. Articles with many Likes are ones that have actually been read by many people and generated a strong response.

Comparison of Likes counts with and without English translation

• No significant difference (Mann-Whitney U test, p=0.81)

In my case, when I compared the Likes counts of articles that were translated versus those that weren’t, there was surprisingly no difference at all.

It seems note is selecting articles for translation based on criteria other than the number of Likes.

Overall Views Also Show No Difference

Next, let’s consider another metric.

On note, you can check the Overall Views (Page Views + Impressions) of each article from the Dashboard in the Creator Menu.

Overall Views and presence of English translation

• No significant difference (Mann-Whitney U test, p=0.34)

Overall Views tend to be slightly higher for translated articles than Likes counts, but there is still no statistically significant difference.

It appears that neither page views nor recommendation impressions are directly related to whether an article gets translated.

Presence of Backlinks

Let’s look at one more metric.

I have written several articles introducing and comparing my own image-generation AI models on note. These articles include backlinks from major domains such as Civitai and Hugging Face, the distribution sites for the AI models.

So this time, I compared articles based on the presence or absence of backlinks from Civitai and Hugging Face.

Backlinks and presence of English translation

• No significant difference (Chi-square test, p=0.35)

This comparison is limited to backlinks I myself placed, but again, there was no relationship with whether an article was translated.

Rather than classical numerical metrics such as Likes counts, page views + impressions, or backlinks, note appears to be selecting articles for translation using a more complex algorithm.

note’s Recommendation Algorithm Is Not Public

As a content distribution platform, note naturally has an Recommendation Algorithm that determines the priority of articles to display.

note keeps this Recommendation Algorithm—the heart of the platform—private, but it has shared some of its underlying philosophy.

Three types of prioritized articles from the author’s perspective

• Primary information (personal experiences and failure stories)
• Articles that do not use inflammatory language
• Beginner-friendly introductory content that encourages new creators to join

note is said to have a kind algorithm that concentrates many page views and Likes on a creator’s very first published article, even when they post it with nervousness.

On the other hand, articles that are purely critical or that stoke division are filtered out by note’s algorithm and never appear in recommendations.

note’s algorithm actually reads and evaluates the content of articles. It is highly likely that note has developed its own proprietary AI for content evaluation and is using it in the Recommendation Algorithm.

Older How-to Articles Have Been Translated

Among the articles I have published, several older ones written as introductions to image generation have been translated into English going quite far back in date since the AI auto-translation service launched in May 2026.

These translated articles are beginner-friendly introductory content that encourages new users to join. It is possible that they were positively evaluated by note’s Recommendation Algorithm and selected for translation even though they are older pieces.

While note keeps the core of its Recommendation Algorithm private, comparing translated and non-translated articles may help us better understand note’s Recommendation Algorithm.

Summary: note Translates Articles It Deems Valuable

• note selectively translates articles it considers valuable
• Likes, overall views, and backlinks are not directly related to whether an article is translated
• Analyzing which articles get translated may offer insights into note’s Recommendation Algorithm

In this post, I examined note’s AI auto-translation.

I analyzed several numerical metrics in relation to whether articles received AI auto-translation, but I was unable to find any clear correlations.

In some metrics, note has already become Japan’s top content distribution platform. Yet its continued ambition to aim higher and challenge the world reminds me of the challenging spirit of Japanese companies from a bygone era—something we rarely see these days.

I plan to keep writing with the goal of having my articles selected for English translation on note.

Thank you for reading to the end!