ShopifyBefore anything runs
- Brand guidelines and a written list of banned words, claims, competitor names and sensitive topics, both signed off by the client.
- Every AI tool in use is listed with its job: drafting, images, video, analysis or changes to accounts.
Content and people
- No claim the product cannot back up, including health, income and result claims.
- No competitor names unless the client approved the comparison and it is accurate.
- Written permission for any real person shown or heard, and no generated testimonial shown as a real customer.
Placements and account changes
- Placement and exclusion settings checked on the day a campaign starts and after any big change.
- The assistant has read access by default, and write access only for the accounts that need it.
- Budget and status changes go through approval, with a read-back afterwards.
When something goes wrong
- Pause the affected ad first, then investigate. Do not wait for the cause to be clear.
- Tell the client the same day, with what is known, what is not known and when the next check happens.
Made-up scenario: an AI drafted headline names a competitor in a comparison the client never approved. The ad is paused within the hour, the client hears the same day, and the headline goes on the banned list.
The AI assistant safety checklist covers the account side in more detail, and the approval workflow term explains the sign off.
