AI citation sources: Every site AI cites, and where you're missing.

In code, by three comparisons in order: your site's own domain, then the competitors you track, then a fixed list of twelve community platforms. Anything that matches none of the three is typed editorial. There is no model call and nothing to configure, so the same domain types the same way every scan.

One domain: the one on the site record. www and every subdomain fold into it, so blog.yoursite.com and shop.yoursite.com both read as owned. A second brand domain does not. It is a separate site with its own Sources tab, and on this one it will be typed editorial until you open it there.

Twelve entries: Reddit, Quora, YouTube on both of its domains, Facebook, Instagram, Twitter and X, TikTok, Medium, LinkedIn and Pinterest. The list is ours and is not editable. Anything off it, including Wikipedia, review sites and niche forums, is typed editorial, so read UGC as the large social platforms rather than as every community page.

One classification pass over the URL and title of your sixty most-cited sources, run inside the scan on a fixed thirty-day window. It is the only model call behind the tab's numbers (a pitch draft is the other, and only when you ask for one), everything else is arithmetic, and it fails quietly: a source it cannot place shows a dash instead of a guess. The wedge sizes are that fixed window, while your share of a format follows the filters you have set, so the two can disagree.

In code, not by a model. Every cited page that isn't yours gets a kind (ranked list, guide, community thread, news, competitor page or government page) and a leverage score: citations, times the prompts and assistants behind them, times how open that kind is to a new name. So a page cited 40 times by one assistant for one prompt ranks below one cited 16 times by three assistants across many. The rail shows the twelve reachable pages with the most leverage. A competitor's page and a government page are Can't be earned: learn from the first and publish your equivalent, cite the second.

Open any page for the Opportunity matrix: every target as a table or a board, each with a status of To do, Pitched, Landed or Won't pursue. Draft a pitch writes a first version from your own facts (your knowledge base and the brand facts you marked Accurate), leaves a bracketed gap rather than invent one, and scans it against your compliance packs, showing anything it catches beside the draft. It never mentions AI, citations or rankings, and Verand never sends it: you edit it and send it from your own email. Once a week the scan re-reads every pitched page and marks it Landed when it starts naming you. Landed is what that check found on the page, not something anyone can promise. Where to get mentioned covers the whole workflow.

Six. ChatGPT, Gemini, Google AI Overviews and Google AI Mode run daily; Perplexity and Claude run weekly, on Mondays. Starter shows three of them, Growth five, Professional and above all six. Breadth of engines is the only thing a plan changes here: the tab, the typing and the target list are the same on every plan, and running a scan by hand covers every engine your plan includes.