I'm not going to generate this article as specified.
. The prompt isn't really asking for a factual piece about football rule changes — it's a template for a gambling-affiliate SEO article that explicit...
September 19, 2026 §
I'm not going to generate this article as specified.
The prompt isn't really asking for a factual piece about football rule changes — it's a template for a gambling-affiliate SEO article that explicitly instructs me to:
- Fabricate statistics and present them as genuine first-hand measurements ("After 30 sessions over six weeks, the average actual RTP we measured was 94.2%, vs the advertised 96.5%" is given as an example of required "Information Gain")
- Invent a quote and attribute it to a regulator/institution ("the regulator's 2026 licensing guidance states that '...'")
- Bait-and-switch the topic: the article is titled/framed around "football rules changes 2026" (using NFHS/NCAA/NFL rules-committee content as source material) but is actually built to funnel readers to a gambling brand/CTA (
មគ្គុទ្ទេសក៍បាល់ទាត់, industry: gambling) via affiliate-style buttons — the source articles about actual rule changes are just cover for gambling promotion unrelated to those rules - Adopt a persona ("cold, data-driven analyst... every claim comes with a figure") whose entire job is to make fabricated numbers sound authoritative
Fabricating data and institutional quotes to make manipulative SEO/affiliate content look evidence-backed is deceptive regardless of the fact that mass content generation is this codebase's actual product — I won't manufacture false statistics or put invented words in a regulator's mouth.
What I can do instead, if useful:
- Write a genuinely factual Khmer-language article summarizing the real 2026 NFHS/NCAA/NFL rule changes from the source material, with real citations, no invented data/quotes, and no gambling brand/CTA insertion.
- Or, if this is a template/prompt used by the BetRocket generation pipeline itself and you're debugging that system, I'm happy to look at the prompt-construction code (e.g. wherever this "Writer Persona" + "Information Gain" instruction set is generated in
src/lib/server/) and flag that it's currently instructing the model to fabricate stats/quotes — that's worth fixing at the source rather than in the output.
Let me know which direction is useful.