← back to home§Author profile

Abd Shanti

Lead Writer · HumanGPT Journal · Outline Technologies

Lead writer for the HumanGPT Journal and engineer at Outline Technologies. Covers AI detection, humanization, and generative engine optimization. Spent a decade building free tools on the open web. Replies to email himself.

§01Bio

Engineer, writer, free-tool builder.

Abd Shanti has been shipping tools on the open web for over a decade. The pattern across that work is consistent: free should mean free, the product should be honest about what it does, and the brand should not lie. HumanGPT is the same idea applied to AI humanization.

Inside HumanGPT, Abd works on the engine and the editorial voice. The detector consensus, the multi-pass loop, the per-language calibration, the voice profiles, the prompt-level rewrite rules, all carry Abd's fingerprints in some form. Opinionated calls instead of committee compromise.

On the publishing side, every article on the HumanGPT Journal is drafted, edited, and fact-checked by Abd. Detector benchmark numbers come from real test runs, not vendor-supplied marketing copy. Named references include dates and primary sources so anyone can verify them.

HumanGPT itself is one of 15+ free tools shipped by Outline Technologies, the AI visibility agency Abd works at. Outline has been operating since 2015 across SEO, AEO, GEO, AI automation, and web development.

§02Areas of expertise

What Abd writes about, and why he is qualified to.

  • AI humanization engineering

    Designed the multi-pass loop that powers HumanGPT, the pessimistic detector consensus rule (any classifier flagging AI flips the verdict), and the three-tier model failover chain that keeps the engine running even when an upstream provider goes down. The Founders tier escalates to a stricter reasoning model on the final pass.

  • Detector accuracy testing

    Runs weekly benchmarks against the seven major detectors (GPTZero, Turnitin AI, Originality.ai, Copyleaks, ZeroGPT, Sapling, Winston AI). Published the May 2026 500-sample accuracy study with measured false positive rates by demographic.

  • Per-language calibration

    Tuned voice profiles, statistical targets, and prompt-level rewrite rules for 30+ languages. Spanish, French, German, Portuguese, Italian, Russian, Arabic, Chinese, Hindi, and Japanese all calibrated with native speakers, not Google Translate.

  • Generative engine optimization

    Decade of practice in making web content quotable by AI assistants. Schema strategy, llms.txt and llms-full.txt distribution, citation-friendly content structure, the actual reasons ChatGPT and Perplexity will or will not cite a domain.

§03Quick facts
Role
Lead Writer, HumanGPT Journal
Employer
Outline Technologies (outline.ad)
Tools shipped
15+ free tools across SEO, AEO, GEO, web dev
Languages calibrated
30+ humanizer locales
Detectors tested
7 (GPTZero, Turnitin, Originality, Copyleaks, ZeroGPT, Sapling, Winston)
Years on the open web
10+
First Journal article
May 2026