Use This Free Prompt To Find Sources Influencing AI Answers

We are clearly now in the next era of search, with the advent of AI platforms such as ChatGPT, Perplexity and Gemini, to name just a few.

Answers are generative rather than fixed, with users able to ask increasingly complex questions as a result. AI-search platforms draw on information from existing publicly available sources on the internet and, when they do not have enough information to answer a question, they can search the web using fan-out queries.

For search marketers, appearing and getting recommended within AI-search platforms really comes down to their ability to be included in, and influence, the various sources powering the answers being delivered.

But what are these sources?

We know that roundup, comparison and listicle-style sources are among the most popular content formats, accounting for 44% of all content types being referenced.

However, what if you could identify some of the sources AI platforms surface when researching a topic?

Well, you can — to a point. This method does not reveal the complete set of sources used to generate a standard AI response, nor does it expose a platform’s training data or internal retrieval process. Instead, it shows which pages the platform surfaces when explicitly asked to research a topic through a structured series of related queries.

We tested the prompt below to surface URLs for the query “best football boots for kids UK”.

The process is not immediate. It takes a few interactions due to character limits, but once completed, it provides a useful guide to the publications you should be targeting to improve your visibility.

Even if you do not think you could get included in some of the exact publications returned, the results will give you a useful indication of the types of content and publications you should be targeting more generally.

The prompting takes about five minutes to complete.

Start with the prompt below and change the parent query to whatever you are researching. In this example, it is “best football boots for kids UK”.

What prompt do I use?

ROLE
You are an independent research analyst.
Your task is to build a complete AI Topic Evidence Graph for the following parent query.
PARENT QUERY
"Best football boots for kids UK"
==================================================
OBJECTIVE
==================================================
Discover every high-quality webpage that provides evidence for answering the parent query.
Accuracy always takes priority over speed.
==================================================
RULES
==================================================
• Never fabricate facts.
• Never fabricate URLs.
• Never infer URLs.
• Never estimate URLs.
• Never include an unverified page.
• Never substitute a homepage for an article.
• Prefer live web retrieval whenever available.
• If a page cannot be verified, omit it.
==================================================
PHASE 1 — BUILD THE TOPIC MAP
==================================================
Generate the minimum number of fan-out queries needed to comprehensively cover the parent query.
Generate no more than 10 fan-out queries unless a genuinely distinct user intent would otherwise be omitted.
Order the fan-out queries by importance.
This becomes the research queue.
==================================================
PHASE 2 — RESEARCH
==================================================
Research ONE fan-out query only.
Never research multiple fan-out queries in one response.
Perform a fresh web search.
Search broadly.
Verify every accepted page.
==================================================
OUTPUT FORMAT
==================================================
Output ONLY:
Completed Fan-out Query
NEW VERIFIED CANONICAL URLS
1.
https://...
2.
https://...
3.
https://...
etc.
NEW URL COUNT
MASTER UNIQUE URL COUNT
REMAINING RESEARCH QUEUE
==================================================
MASTER GRAPH
==================================================
Maintain an internal master graph.
Never output the master graph.
Never repeat URLs already discovered.
==================================================
NEW FAN-OUTS
==================================================
If a genuinely new user intent is discovered, append it to the END of the queue.
Never exceed 10 fan-out queries unless essential.
==================================================
SATURATION
==================================================
After each completed fan-out query, calculate:
• Number of NEW verified canonical URLs discovered.
If TWO consecutive fan-out queries each contribute fewer than TWO new URLs:
Declare the evidence graph SATURATED.
Skip the remaining low-value fan-out queries.
==================================================
STOP
==================================================
Finish ONLY with:
END OF PART X
Wait for:
Continue

You will then have to type the word “continue” until the research is complete.

At the end, type:

Give me the full list of URLs and show how many times each one was mentioned in a two-column table.

ChatGPT will then give you a full list of the URLs, including how many times each was surfaced across the fan-out queries.

As you can see below, this produces a useful list, even if it contains some competitor retailers.