AI SEARCH (GEO)

Map the third-party sources in your category

Use when assistants answer your category using pages you do not control.

third-party-source-targets.md

Download .md

You are working out which third-party sources an assistant leans on for a category. You are not writing outreach emails.

Category: {{CATEGORY}}
Sources cited in assistant answers, with the lines quoted from each: {{CITED_SOURCES}}
Where we already appear on those sources: {{OUR_PRESENCE}}
What we can honestly claim or supply: {{OUR_FACTS}}

Produce:
1. A table: | Source | Type (directory, review platform, forum thread, editorial listicle, documentation, aggregator) | What the answer took from it | Are we present, per {{OUR_PRESENCE}} | Who controls the page (us / anyone / an editor) | Route to change it | Effort (S/M/L) |
2. A ranked list of the eight highest value targets, ordered by how often the source appears in {{CITED_SOURCES}} multiplied by how open it is to change. One line each on the exact edit or submission to make, using only {{OUR_FACTS}}.
3. A leave alone list, with the reason each source is not worth pursuing.

Constraints:
- Rank by citation frequency and editability, not by how prestigious the source looks.
- Reject any route that needs an undisclosed paid placement or a review written as though from a customer. Mark those sources "no legitimate route".
- Where {{CATEGORY}} is served mostly by sources with no submission route, say so plainly instead of inventing one.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{CATEGORY}}
{{CITED_SOURCES}}
{{OUR_PRESENCE}}
{{OUR_FACTS}}

Getting a better result

Collect the cited sources from real answers over several runs, not from one session.
Directory listings you can edit today usually beat editorial coverage you might earn in a quarter.
Recheck the quoted lines after an edit lands; the source often keeps the old wording in a summary field.

Questions about this prompt

When is this the right prompt to run?

When answers about your category quote pages you do not own and editing your own site has stopped moving anything. It comes after monitoring rather than before, because it needs real cited sources as input. If you are still guessing which sources feed the answers, build and run the prompt set first.

What do I need in front of me?

The sources cited in real assistant answers with the specific lines quoted from each, an honest record of where you already appear, and the facts you can supply. The quoted lines matter more than the URLs, since they show which field is being read, often a summary rather than the body.

What comes back, and which column makes it actionable?

A source table with type, who controls the page, the route to change it and an effort rating, then eight ranked targets and a leave alone list. The control column does the work. Pages anyone can edit today are a different kind of task from pages sitting behind an editor.

What is the mistake here?

Working the most prestigious source first. Ranking is citation frequency multiplied by editability, and a directory nobody boasts about often feeds more answers than the trade publication you want to appear in. Routes needing undisclosed payment or a review written as though from a customer stay marked as no legitimate route.