The #1 reason GHL AI bots go off script (and how to fix it)
After building 15+ GHL chatbots for one client, I noticed a pattern in every bot that underperformed. They all had the same problem: the prompt was written like an essay. GHL's AI engine reads prompts in three distinct sections — Identity, Goal, and Additional Info. When you mix everything into one block of text, the bot loses context mid-conversation and starts making things up. Here's what each section should contain: Identity — Who the bot is, what business it represents, its name, tone, and hard limits on what it will and won't discuss. Be specific. Vague identity = vague bot. Goal — One primary objective only. Book an appointment. Qualify a lead. Answer FAQs. One goal. Bots that try to do everything do nothing well. Additional Info — Business-specific rules, escalation triggers, hours, services, objections, and your Knowledge Base connection. This is where most people underload. The bots that actually book appointments follow this structure exactly. Anyone else found this to be true? What's the biggest prompt mistake you've seen in GHL?