Track sources & citations
The sources view shows which websites the engines actually draw on when they answer your category, ranked by how often each is cited. The big lever it reveals is rarely your own site: if a review site or directory is cited again and again, earning a place there can beat publishing one more page of your own.
Before you start
Sources are extracted from saved answers, so you need a prompt set that has run. The view is richest on engines that cite the web heavily, especially Perplexity and Gemini. See mentions, citations and sources for the exact definitions.How to use the sources view
Read the ranked list
Open the sources view and you get a list of sites ordered by citation frequency across your prompts and engines, the most influential sites in your category sit at the top. Start here: these are the pages the engines reach for when deciding what to recommend.
Tell your citations apart from third-party sources
Each entry is tagged by ownership. A citation of one of your own pages means the engine linked you directly; a sourceon a site you don't control is a third party the engine trusts. Both matter, but they call for different moves, one you edit, the other you have to earn your way onto.
Find the high-leverage sites
Look for sites cited often where you are absent or thinly represented. A directory that feeds five of your prompts but never lists Acme is a single, concrete fix that can lift multiple prompts at once. That is far more leverage than a sixth page on your own blog.
Decide: get listed, or improve your listing
For a trusted site you're missing from, the job is to get listed, submit, pitch, or qualify for inclusion. For one where you already appear, the job is to improve the listing: fresher details, better reviews, a higher placement. Both change what the engine reads next time.
Track source influence over time
Source rankings shift as engines change what they trust. Re-read the view across runs: a site climbing the list is worth investing in early, and a once-dominant source fading tells you where not to spend more effort.
The core insight: borrow trust, don't just build it
AI answer engines don't read your site in isolation. They synthesise from the sources they trust, and those are mostly third-party: review sites, directories, forums, and roundups. So when a review site shows up cited again and again for your category, the highest-return move is usually to improve your presence on that site, not to publish another page on your own domain. You are borrowing the trust the engine already has in that source, instead of trying to build all of it yourself from scratch.
Worked example
Acme tracks "best project management tools for startups". In the sources view, the top cited sites across Perplexity and Gemini are:
- g2.com, cited on 9 of 12 prompts. Acme is listed but ranked low with 14 reviews; Beta has 200+.
- a startup-tools roundup blog, cited on 6 prompts. Acme is not mentioned at all.
- acme.com, cited on 2 prompts, only when the question already names Acme.
Acme's instinct is to write more blog posts. The sources view says otherwise: getting added to the roundup and growing reviews on G2 touches fifteen prompt-citations between them, while a new page on acme.com would only be reached when someone already searched for Acme. The third-party fix wins.
What good looks like
- You can name the top three sites the engines cite in your category from memory.
- You appear on the high-frequency trusted sources, not just your own domain.
- Your work plan is sorted by source leverage, how many prompts a fix touches, not by how easy the page is to write.
- You re-check the ranked list each month and notice when a source is rising or fading.
From sources to action
A source is only useful once it becomes a task: get listed on the directory, lift your G2 rating, pitch the roundup author. Translate the ranked list into a short worklist and revisit it as the rankings move.
Common mistakes
Optimising only your own pages.Polishing acme.com feels productive, but if the engines cite third-party sites far more often, you're improving the page they read least. Follow the citations, not your comfort zone.
Confusing a source with a citation of you.A site appearing in your sources doesn't mean it mentions Acme, it means the engine read it. Always check whether you're actually present there before assuming the source helps you.
Reacting to one run. Source lists are probabilistic too. Confirm a site is consistently influential across runs before you invest in earning a place on it.
Next steps
The exact difference between being named, being linked, and being read.
Turn high-leverage sources into a plan that moves your standing against competitors.
Why engines surface different sources, and what that means for you.