Launched 2 days ago: self-hosted AI detector tested against 50+ LLMs. Here's what I built and why.

Hey IH,

I've spend 25 years doing SEO (more than that), I thought I'd seen every content quality problem there was. Then two things happened simultaneously that genuinely made me angry, so much frustration recently with lazy article writers and contractors:

1. AI-generated content got sophisticated enough to be deliberately engineered to fool detectors — Maverick Journalist personas, model-hopping, adversarial prompting. I watched this slip past Turnitin, Originality.ai without triggering a flag.

2. All those "top" detectors started throwing false positives on clean, human-written text content. Writers were losing contracts. Agencies were hemorrhaging trust with clients.

Both problems. Same broken tools.

I tested every major detector on the market. The results were embarrassing. So I stopped complaining and built iWordCraft.

What makes it different:

- Not an API wrapper. Runs entirely on self-hosted infrastructure (Docker, ML + custom Math Engine)

- Dual-Layer Detection: local ML flags generation patterns, mathematical heuristics catch aggressive spinning

- Zero text transmitted to external providers — ever

- Tested against 50+ LLMs using real evasion techniques. Detection rate: 100%. False positives: zero. Two of the most popular AI detectors failed all tests!

That last line took a long time to be able to write honestly.

I'm at the stage where the engine is solid and I want real users stress-testing it — not synthetic tests (already did that), but actual workflows. Messy client content. Aggressively spun articles. Edge cases I haven't thought of.

Free accounts for anyone willing to genuinely try to break it and tell me what they find. Drop a comment or DM me directly.

Happy to answer anything about the architecture or the detection methodology.