What this use case can do for you
Content decay in AI search is one of the harder problems to catch. Pages that were earning strong AI citations are quietly losing retrieval to competitors – and without tracking it, you just give that traffic away. This workflow identifies which of your own pages have lost the most AI search visibility over the past 60 days, audits the root cause, and outputs a prioritized content rescue tracker in Notion. It pulls two consecutive 30-day windows of retrieval data, surfaces the 10 URLs with the steepest retrieval rate decline, and runs a GEO content audit on each one – checking answer density, freshness signals, internal cannibalization, and which competitor content or source type has taken over. It also maps the prompts that were driving retrieval in the earlier window, so you know exactly which queries you've lost ground on. The output: a ranked list of pages to fix, with root cause classification and specific recommendations – built on retrieval data, not assumptions.
Step 1 — Retrieval decay analysis. Pull the URL report for the project, filtered to its own domains, for two consecutive 30-day windows ending today. Compare them and find the top 10 URLs with the steepest retrieval drop. For each URL return: URL, page classification, retrieval count in the earlier period, retrieval count in the recent period, and percentage drop. Rank by biggest decline. After identifying the top 10 URLs, pull the prompt_id dimension on the earlier-window URL report filtered to those 10 URLs, then resolve all prompt IDs to their full text and search volume using list_prompts. Add a "Prompts" column to the output table showing the prompt text(s) for each URL. ===== Step 2 — GEO Content Audit. For each URL, use get_url_content and assess: What the crawler actually reads — editorial content, or nav/boilerplate/JS-rendered junk? Answer density — can you identify one specific sentence or data point an AI engine could extract and cite? If not, that's the primary failure cause. Claim specificity — are claims concrete and falsifiable (numbers, rankings, dates) or generic? Generic claims don't get cited. Freshness signals — is there a visible date, recent year references, or dateModified schema? Flag pages with no signals separately from pages with outdated signals. What replaced it — based on the domain report data, what source type took over: editorial, UGC/video, aggregator, competitor OEM, or brand's own newer content? This determines the fix format. Internal cannibalization — are other brand own URLs competing on the same prompts and splitting retrieval? Classify each URL with a single root cause and a fixability rating: Replace / Technical fix / Content rewrite / Monitor. ===== Step 3 — Notion database. Create a [Project Name] Content AI Visibility page in Notion and create a database there called "URL Retrieval Decay — Content Rescue Tracker". Include these columns: URL Title, URL, Page Classification, Previous Retrievals, Recent Retrievals, Drop %, Rank, Core Problem, Fixability, Status. Create one database entry per URL. Populate Core Problem and Fixability based on the content audit. Inside each page, write a detailed section covering: what AI engines are currently reading on that page, why it lost retrieval, and specific numbered recommendations with reasoning. Add a "Prompts (Earlier Window)" column of type RICH_TEXT to the database schema. Populate it for each entry with the prompt text(s) that drove retrieval in the earlier window. The zero-retrieval pages should have the most detailed treatment. Do not show intermediate results or ask for confirmation between steps. Complete the full workflow and share the Notion database link when done.