DEJAN's AI Outreach engine uses an adversarial Red/Blue workflow where a Red team agent writes grounded articles around a target link and a Blue team agent evaluates whether the placement appears paid or unnatural. Drafts that fail detection or cite competitor co-link domains are automatically folded, while passing drafts are cleaned with socket-level dead-link pruning that unwraps broken citations into plain text. When refining drafts, an Outreach Coach converts the detector's findings into exact, line-by-line rewrite suggestions grounded in an anti-fluff editorial guide.