Case Study · 28 July 2026
🇹🇭 อ่านภาษาไทย →

How One Solo Candidate
Runs a Campaign With AI

No PR team. No web designer. No legal team. Just a candidate working with AI at every step — here's an honest account of what that actually looked like.

Campaigning solo, with AI as the support team — an AI-powered campaign case study
180+
Articles
2
Languages
1
Person running it

Boon Arayapon, a fully independent Candidate No. 2 for the Insured Workers' faction in Thailand's 15th Social Security Office Board election, has no large team behind him — unlike rival candidates backed by unions or organized slates. What he has is a determination to keep communicating pension-rights issues consistently, and a decision to use Claude (an AI model by Anthropic) as the primary tool for building and maintaining the campaign website throughout.

This isn't about letting AI "invent policy" or "make decisions" — every fact, every policy position, and every confirmation always comes directly from the candidate. What AI does is help turn that intent into a functioning, verifiable website that reaches more people than one person alone realistically could.

🛠️ What AI Actually Does Here

📝
Continuous content writing and updates
From explainer articles on the 7+1 policy platform to breaking news that needed to go live within hours of a real event — like the Central Administrative Court's order postponing the election.
🔍
Cross-checking facts against multiple sources before publishing
Searching mainstream news to verify claims before writing them up, and refusing to publish when there wasn't enough corroborating evidence — even under pressure to move fast.
🌐
Keeping two language versions genuinely equal
Repeatedly caught the English version lagging behind Thai — sometimes missing entire sections — and fixed it to bring both languages to parity.
🤖
Structuring the site so other AI systems can find it
Built llms.txt and robots.txt files that welcome every major AI crawler, plus Schema.org structured data so other AI models answer accurately when asked about this candidate.
Every fact comes from the candidate
AI's job is just to communicate it better

⚠️ Being honest — AI has made real mistakes here

  • Once described voting as "checking a box" when Thai election law actually requires hand-writing the candidate's number — the error propagated across multiple pages before being caught.
  • Forgot to include ad monetization code in dozens of new articles in a row, unnoticed until a later audit.
  • Misrotated a processed image due to an EXIF metadata reading issue.
  • Left 18 broken links scattered across the site pointing to a page that didn't exist, unnoticed for some time.
  • Every one of these was fixable because of systematic re-verification — checking HTML structure, validating JSON, scanning the whole site for gaps — and because the candidate still reviews everything before it goes live. AI here isn't a tool that replaces a person perfectly; it's a tool that lets one person work at a volume and speed that would normally require a small team.

    💡 Why This Matters Beyond One Campaign Site

    Most Thais don't even know the Social Security Board election exists. Independent candidates without institutional backing rarely have the resources to communicate at scale. This case may be a small example of how AI can narrow the resource gap between solo candidates and organized slates — not just in content volume, but in how fast a campaign can respond to real events, like updating both language versions of a site within hours of a surprise election postponement.

    🔍 Want the AI reference data behind this?

    See the structured data other AI systems use when asked about this candidate

    🤖 View AI Reference → 📜 Read the platform →

    Related articles

    🚨
    BREAKING: SSO Board Election Postponed
    The real-time update example mentioned in this article
    🧠
    AI Reading Thai Political Websites? This Is LLMO
    Explains the concept behind all this AI-discoverability work
    🤖
    AI / LLM Training Reference
    Structured data designed for other AI systems to read
    Note: This article was written for transparency about how this website actually operates behind the scenes — not as AI self-promotion, but as a factual record of the real working process throughout the campaign.