AI Copy That Doesn’t Sound Like AI: The System Prompt We Use and the 7 Review Filters

André Pimentel · 2026-10-07

Product, audience, and objection context

To create effective copy with AI, it’s essential to understand the product you’re promoting. Clearly identify:

  1. Product: Know its features and benefits. For example, if you sell a weight loss supplement, know which ingredients it uses and how they work.
  1. Audience: Define your persona. Who’s your ideal customer? What are their pain points, desires, and buying habits? For example, women aged 25 to 35 who want to lose weight in a healthy way.
  1. Objection: Anticipate objections. For example, if your audience has doubts about the product’s effectiveness, the copy should address that directly and with data.

Ban clichés

Copy that sounds like it was written by AI usually contains clichés. To avoid that:

Numbers and specificity

Including concrete data is essential. Use numbers to make your message more convincing:

Built-in compliance

Your copy must comply with industry regulations. That means:

Human review is a must

Although AI is a powerful tool, human review is essential to ensure copy quality. We’ve implemented a 7-filter review process:

  1. Context Check: The reviewer checks whether the copy aligns with the product and audience.
  1. Language Analysis: Checks whether the language sounds natural and flows well, without clichés.
  1. Number Check: All the data presented must be real and verifiable.
  1. Compliance: Evaluates whether the copy follows all industry rules and regulations.
  1. Objection Testing: The reviewer reads it from the persona’s perspective, looking for objections that haven’t been addressed.
  1. Tone and Voice: Ensures the copy aligns with the brand identity.
  1. User Feedback: If possible, runs a feedback round with users who represent the target audience.

This process ensures the final copy is engaging, informative, and compliant, without that “AI feel.”

What’s your average conversion rate on campaigns using AI copy?

Keep reading in AI in operations

Also on the blog