Command
Use the Peec MCP to pull our brand report broken down by chat_id and filter for conversations where our sentiment score is below 50. For those low-sentiment conversations, pull the URL report to identify which pages were retrieved in the same chats. Cross-reference to find the URLs that appear most frequently alongside negative sentiment scores. For each top offending URL, use get_url_content to fetch the page and identify what content is likely driving the negative association. Return a ranked list of the top offending URLs with: URL, page classification, retrieval count, average sentiment score in chats where this URL was retrieved, and a brief note on what the page says and why it is likely harming our score.
Most teams try to improve AI sentiment by creating more positive content, hoping it will dilute the negative signal over time. That approach is slow and often misses the actual problem. A single high-retrieval page, a critical Reddit thread, a comparison site where your product is ranked last, or a negative review from a high-authority domain, can account for a disproportionate share of your sentiment damage.
This finds the specific pages behind the score, ranked by how often they appear in low-sentiment conversations. You can then deal with them directly: update the content, request a correction, outrank the page, or deprioritize the source.
That is a strategy. Creating more blog posts and hoping for the best is not.
