Command
Use my supplied host, approved positioning statements, and days of 7, 30, or 90. Confirm access with list_projects. Read get_brand_insights({host,days}), then list_prompts({host,status:"active",limit:100}), following nextCursor as cursor when needed. Select relevant prompt IDs and inspect get_prompt_runs({host,prompt_id,days}); use search_chats({host,query,days,limit:20}) for a short positioning phrase I provide. Build a matrix of intended claim, observed wording, supporting sample, and discrepancy. Distinguish absence in these samples from a universal omission. Propose three messaging experiments for review, with no invented perception score or claim that editing content will necessarily change model answers.What you need
- Provide approved positioning statements and a visible project host.
- Choose 7, 30, or 90 days and supply any phrases to search.
Make messaging discrepancies inspectable
Place your intended positioning beside the descriptions actually returned in sampled answers and brand insights. Each discrepancy retains its evidence and scope, allowing reviewers to judge whether the issue is incorrect wording, an omitted attribute, or insufficient coverage for a conclusion.
Design focused messaging experiments
Translate observed discrepancies into a small set of draft messaging changes and follow-up questions. The plan identifies what evidence would support improvement, while leaving content approval and publishing with your team and avoiding promises about future AI response behavior.