Open Source · AI Programming · Markdown

A Developer Builds a Personal Site When AI Reads Everything—and It's Not a Losing Bet

By king王一帅 · Oct 31, 2026

Read original on juejin.cn ↗ Google Translate ↗ Alt translation

AI citation is becoming a real distribution channel, but most sites aren't structured to be verifiable primary sources. This post lays out a concrete workflow—RSS, canonical, tool pages, and honest llms.txt use—that any developer can copy to keep a presence on the AI-driven web.

Summary

The post starts with a blunt admission: an independent blog in the AI era is a fool's errand if treated as a traffic business. Platforms have recommendation feeds, AI summarizes pages, and people rarely click unknown domains. But the author makes a different bet. He claims that as AI pushes content supply to infinity, the value of first-hand experience, real project records, and verifiable authorship increases.

He describes concrete steps to make a personal site citable by both humans and AI: publish first on the site, sync elsewhere, keep RSS feeds, add llms.txt only as a guide (not a strategy), and build structured tool pages for his open-source editor QuireKit. He built the entire site, from pages to comments, with his own QuireKit stack, and open-sourced the site code. He acknowledges the readership may be tiny, but he sees the site as a personal source of truth that belongs to him, not to platform algorithms or AI summaries.

Takeaways

— The author treats SEO-based traffic and ad revenue as dead for new independent sites; he doesn't expect a return to the 2010 blog-traffic model.

— He bets that AI's content flood makes first-hand experience, real project records, and verifiable signatures more valuable.

— He uses four discovery paths beyond direct visits: AI search citations, GitHub referrals, developer communities, and RSS subscriptions.

— He warns that being read by AI isn't the same as being cited or driven traffic; llms.txt is not a standard and many AIs ignore it.

— He structures the site for citable content: conclusion cards, fact tables, FAQ, version history, tool pages, and per-article .md endpoints.

— He writes first on yishuai.wang, then syncs to platforms, setting canonical links where supported and adding a first-published note otherwise.

Conclusions

The author's distinction between 'being read by AI' and 'being cited by AI' is a useful reality check; it separates the promise of llms.txt from its actual effect.

His claim that first-hand experience becomes scarcer as AI content mounts is plausible but unverified; he says so himself, which is refreshing in a field full of overclaiming.

He inverts the old SEO mindset by treating the site as a source of truth rather than a traffic play, which may be the only viable model for personal sites now.

Concepts & terms

llms.txt

A proposed text file for listing pages a site wants AI models to read. The author notes it's not a standard and many AIs don't use it, so it shouldn't be relied on as a distribution strategy.

Micropub

An open API standard that lets external clients publish posts to a website, enabling writing from other tools.

Canonical tag

An HTML element that tells search engines which URL is the preferred version of content, used here to point to the original article on the author's site.

From the discussion

Featured comments

king王一帅

Old Jue, actually I'm talking about you. I didn't mind at all, but I still have to thank you, because you made me think deeply. Actually, regarding publishing tools, only after experiencing them can you know the difference from others on the market. I haven't published, and I don't blame others for giving direct evaluations without experience. Actually, the content I posted also has problems. Next time I won't announce in advance [lightbulb].

See top comments, translated →

Source: juejin.cn ↗ Google Translate ↗ Backup ↗