When I decided to ship this blog in four languages, I had a clear mental ranking. English would win on volume. Spanish would be runner-up because of the sheer speaker count. Japanese would stay steady because it's my native language. Portuguese, I figured, was the long tail. I added it mostly out of completism.

Twenty-two days later, the GA4 snapshot disagrees with every part of that ranking.

• PT: 748 pageviews, 709 sessions
• EN: 195 pageviews, 176 sessions
• JA: 27 pageviews, 29 sessions
• ES: 7 pageviews, 7 sessions

That is Portuguese pulling roughly 3.8× English, 28× Japanese, and 107× Spanish on the same blog, same publishing cadence, same author. One Portuguese article on its own (a post about a 24-hour security agent: 375 PV) got more pageviews than my entire English blog combined.

I wrote that article hoping Spanish would surprise me. Instead Portuguese surprised me, and Spanish quietly continued to not exist.

The setup, so you can discount my numbers properly

This is not a clean comparative experiment. It's a single blog, kenimoto.dev, running four language directories (/en/, /ja/, /pt/, /es/). Articles get translated through a cross-language LLM pipeline, then hand-edited for register and locale (BR Portuguese vs PT Portuguese, LatAm-neutral Spanish vs Spain Spanish). The window: 2026-04-30 to 2026-05-21, 22 daily snapshots.

EN has 26 articles. JA has 25. PT has 17. ES has 10. So PT has fewer articles than EN and still beats it almost 4 to 1.

If you stop reading here, take this one thing: language asymmetry can swallow article-count asymmetry whole. Adding articles in a saturated language is slower than adding articles in an underserved one.

Why Portuguese pulled ahead

I don't think the answer is "Portuguese readers like me more." I think three asymmetries are stacking on top of each other.

1. TabNews is a community door English doesn't have

TabNews is a Brazilian developer community where you can post a technical article and have it actually read by humans, the same day, without already having an audience. There's no clean equivalent in English. Hacker News exists, but the floor for getting noticed there as a no-name is much higher, and the topic surface is narrower.

When I cross-post the same article to TabNews (PT) and Dev.to (EN), TabNews delivers consistent referral traffic. Dev.to mostly delivers crickets unless I already have followers. That difference shows up directly in the GA4 numbers.

2. Portuguese AI-search results are thinner

English LLMO content is a saturated market. Thousands of decent articles compete for the same prompts in ChatGPT, Perplexity, Gemini. Your share of voice as a small site is correspondingly small.

In Portuguese, the field is thinner. Fewer technical blogs fight for the same prompts, so when an AI engine needs a Portuguese source for "spec-driven development com Claude Code," there are far fewer candidates to pick from. The first reasonable answer in PT wins. The first reasonable answer in EN gets buried.

There's a structural reason this matters more in 2026 than it used to. ChatGPT, Google AI Overviews, and Perplexity largely ignore hreflang when they assemble an answer. They pull from whichever language version of your content looks strongest and translate it on the fly. So your Portuguese article isn't competing only for Portuguese prompts; if it's the strongest source on a topic, an AI engine can surface it for adjacent queries and translate it. In a thin language, "strongest source" is a much easier bar to clear.

3. I'm an early-mover on /pt/llms.txt

Most major Brazilian developer sites don't ship an llms.txt yet. Some big LatAm Spanish sites don't either. By having /pt/llms.txt, /es/llms.txt, /ja/llms.txt, /en/llms.txt from day one, I give AI crawlers a clean menu in their target language. In English this is just hygiene; everyone has one. In Portuguese it's mildly differentiating.

This also lands at a moment when AI crawlers are doing far more of the reading. One 2026 analysis found OpenAI's ChatGPT-User crawler made about 3.6× more requests than Googlebot across the data it studied. If the bots are the ones reading first, the language they can read cleanly matters more, not less.

Why JA got 1/27th of PT (painful for me to type)

Japanese is my native language. I write the JA versions myself, not via translation, so the prose is the cleanest of the four. And the JA blog got 27 pageviews. Twenty. Seven.

The honest reason: Japanese developers mostly read Qiita and Zenn, not standalone blogs. When I post to my own domain in Japanese, I'm asking readers to leave their normal habitat. When I post the same article to Zenn instead, it gets dozens of reads on day one.

So the JA strategy has to change. The blog shouldn't try to compete with Qiita/Zenn for the JA audience; it should be the canonical archive that AI crawlers index, while the Qiita/Zenn versions do the human-traffic work. That's the opposite of how the PT side works, and that's fine. Different language, different distribution.

Why ES is at 7 pageviews and I mostly deserve it

ES has 10 articles. The translations are clean LatAm-neutral. The problem is distribution: I have no equivalent of TabNews to post to. Stack Overflow en español exists but isn't the same shape of community. Platzi and Código Facilito are great, but they aren't open posting platforms.

So ES sits in the weird middle: AI-search competition is also thinner than EN (a tailwind), but the community door is missing (a headwind). The result is single-digit pageviews. I don't have a clever fix yet; the next 30 days of ES experiments are about finding a posting hub that isn't a giant company's gated platform.

The multi-language LLMO checklist I wish I'd had on day one

If you're about to translate your blog into N languages, here's the playbook I'd give past-me:

1. For each target language, identify the community door first. Not the audience size. The door. Brazil has TabNews. Japan has Qiita/Zenn. English-speaking Hacker News exists but the bar is brutal. Spanish LatAm: still searching.
2. Ship /{lang}/llms.txt from day one. It's 15 minutes per language. Most non-English sites don't have one. This is the cheapest moat you'll ever build.
3. Set up GA4 with language-prefix filters before publishing. Otherwise you'll spend month two retrofitting analytics instead of writing.
4. Resist the urge to translate everything. Translate the 20% of articles most likely to land in the community door. The rest can wait until you've validated the channel.
5. Treat each language's AI-search share-of-voice as a separate KPI. Run the same brand-relevant prompts in ChatGPT, Perplexity, Claude.ai in each language, monthly. The asymmetries are huge and you can only manage what you measure.

What I'm doing next

• Doubling PT publishing cadence from 1/week to 2/week and measuring whether TabNews referral scales linearly or saturates.
• Reframing JA: blog as AI-crawler archive, Zenn/Qiita as the human-distribution surface.
• Finding the missing ES community door, even if it means experimenting in three LatAm hubs at once.
• Leaving EN cadence alone. The English market is saturated; my marginal article there is worth less than my marginal article in PT.

If you've been resisting multi-language because "I don't have time," consider this: the language with the highest ROI on your time may not be the one with the most speakers. It may be the one with the fewest competitors in the AI-search layer and the most welcoming open community.

For my blog, that was Portuguese. For yours, it might be Indonesian, or Korean, or Polish. The only way to find out is to ship one article in each, plug in GA4, and see which one the AI engines start citing first.

---

If you want the deeper playbook on measuring and improving your AI-search visibility across languages, I wrote a short book on exactly this: LLMO Quickstart. The multi-language chapter is the one I rewrote three times after the numbers above came in.