Being mentioned in an AI answer is the thing you are actually competing for, and it rarely comes down to one page. An assistant reads a handful of sources, weighs them, and writes an answer that may cite you, name you without a link, or leave you out entirely. To change that outcome you need three things in front of you at once: which sources the assistant pulls from, whether those sources cite you or a competitor, and which of your pages could close the gap.
The Citations and Sources view now puts all three in one place, with the evidence behind every call.
The words, quickly
A few terms do a lot of work here, so it helps to be precise about them:
- Source. A page or domain an AI answer drew on. Most answers lean on several.
- Citation. A source that is actually linked in the answer.
- Presence. How your brand showed up in a given answer: Cited (your page is in the answer’s sources), Named (your brand is in the text without a link), Considered (your page was in the retrieval pool but not the final answer), or Missing.
Keeping Cited and Named apart from Considered matters, because a page that keeps getting considered but never cited is a very different problem from one the assistant never sees at all.
The source landscape
The top of the view answers the big question first: across every answer in the window, which sources do the assistants actually pull from, and how often do they cite you? Brand Coverage is the share of answers that include your brand, and the platform breakdown shows where citations concentrate, because the cited mix on Perplexity is not the cited mix on Google AI Mode.
The lower chart plots brand coverage by source: how often the domains that shape your category cite you versus your competitors over time. When a reference site or a review domain carries a category and barely cites you, that line is the gap, drawn against the competitors who are already on it.
Every source, sorted by what it does for you
Below the landscape, the sources table lists every domain the assistants used, grouped by what kind of source it is: competitor, community, editorial, institutional, reference, or your own pages. Each row carries how often it is used, whether you are cited there, and which platform leans on it most.
The status filters are where this turns into a worklist. You missing surfaces high-usage sources that never cite you. Only co-citing finds sources that cite your competitors right alongside their other picks but skip you, which is usually the most winnable kind of gap. Sort by usage and you are looking at the sources worth the most, ranked by how much they move AI answers in your category.
From a gap to a page worth winning
A missing source is only useful if you know what to do about it. Open a row and the evidence drawer lays out the case: which prompts it affects, which competitors it cites instead, and a clear read on whether this is a page you can earn or a source that is out of reach. Page-level corroboration ties the answer back to specific URLs, so a fix has an address rather than a vague intent.
From there the gap becomes work. The same evidence drives an opportunity you can act on, and if you want the longer method behind reading these gaps, the source-gap audit guide walks through it step by step. It is also worth remembering that ranking well in classic search does not guarantee a citation, which is a gap in itself.
Read it honestly
Source data is easy to over-read, so the view is built to keep you honest. A brand mention in the text is not the same as a linked citation, and the presence ladder keeps them separate. A considered source sat in the retrieval pool; it was not necessarily in the final answer. And a brand-new brand will show a thin landscape because the assistants have not had much to read yet, which the view labels rather than dressing up as a measured zero. The point is to act on real gaps, not artifacts of how recently you started tracking.
You will find all of this under Citations and Sources for any tracked brand, and the same numbers feed the Overview so a figure never changes meaning between screens.
A note on timing
SurfacedBy changes often. We build from a mix of customer requests, our own research, and keeping pace as the AI platforms change what they cite and how they show it, so the screens here reflect how Citations and Sources looked at launch. The presence ladder, the per-platform breakdown, and the evidence drawer have all gained depth since, and they will keep gaining it.
