Six steps, one module
Until now, SEO software stopped at the gap in the AI answer. Closing it meant a spreadsheet, a PR agency or yet another tool. Here the whole route runs in the same system where your signals, briefings and reports already live.
- Find sourcesPages AI cites while your client is missing.
- JudgeDoes the audience fit? And how do you get in?
- CurateYou keep or drop each source.
- Find contactsThe right person, with evidence from the site.
- DraftPitch or form text, checked against sources.
- SendYou send it. The rest is pipeline.
Targets come from citation data, not from a media list
A list of thousands of media contacts tells you nothing about which page ChatGPT actually pulls up for your client's question. RankCockpit looks where answers are made: in the sources AI answers cite when your client's brand is missing and a competitor is often named in its place. Four more routes come on top, and you combine them per client.
- AI citationsSources from your AI visibility tracking, filtered to queries where your client does not appear.
- Search queriesGoogle top 10, AI Overview sources and optionally ChatGPT sources for your own queries.
- CompetitorsDomains linking to your competitors, filtered to your client's market.
- DirectoriesReview sites and industry directories as their own category, flagged when a listing looks free.
- VisibilityFor every source: how many queries cite it, organic traffic and rank in the target market.
- Your knowledgeImport your own lists on top. Everything lands in the same curation.
Two verdicts per source: does it fit, and how do you get in?
A judge reads your client's business brief and scores every domain for business fit from 0 to 3, with a one-sentence reason. A second verdict says which door is open: guest post, directory listing, stage or podcast, press, community, or none. You set the judge's rules per client, and they are versioned. Keeping or dropping a source is your call. The judge never changes a list on its own.
| Source | Cited in | Fit | Access | Status |
|---|---|---|---|---|
| Trade magazine, how-to sectionCompetitor named in 9 answers | 14 queriesChatGPT, AI Overviews | 3 · fits | Guest post | Review draft |
| Industry comparison siteCompetitor named in 6 answers | 11 queriesPerplexity, ChatGPT | 3 · fits | Listing | Published |
| Trade podcast with guestsInterview page with guest form | 5 queriesChatGPT | 2 · likely | Stage | Contact found |
| Regional daily paperLocal section, no topical fit | 3 queriesAI Overviews | 0 · no fit | no way in | dropped |
Shown with sample data
Contacts with evidence instead of guessed addresses
For every approved source, a worker reads up to eight pages of the site: legal notice, contact, team, editorial staff, guest authors, media kit. The model picks the contact that matches the way in. Then the code checks strictly: an address only enters the system if it appears verbatim on a loaded page, a name only if it appears on the source page. Role addresses such as privacy or billing are dropped. Every contact stays marked as unverified until you confirm it, with source and evidence right next to it. Cost: about 1 to 3 cents per domain.
The first draft, fact-checked by a second model
One click and the system writes a pitch aimed at that specific site: from the client brief, your voice profile, the way in, the site's current article titles and previously sent emails as a style reference. A second model checks the draft against exactly those sources and deletes anything that is not backed up. If the core claim is not backed up, the draft is blocked instead of polished. The code then checks length, salutation and every quoted article title. For directories and stages, you get the form text instead of an email. Cost: about 10 to 16 cents per draft.
One pipeline, all the way to the published URL
Every source moves through fixed states: planned, approved, sent, replied, in progress, published. Declined and no response are kept apart, because you never touch a real rejection again, but you can try someone who stayed silent in a second wave. At the end you record the URL where the mention went live, the starting point for checking the backlink and the AI citation. Your clients see their sources in their own login, help curate and maintain their own contacts. Status and history stay with you.
What this means for you
- GEO does not stop at measuring: the gap in the AI answer becomes a list of specific pages with a contact.
- Link building and digital PR aimed at sources AI demonstrably reads, not at volume.
- Every claim in a draft is backed up. No invented references to articles that do not exist.
- A new service package for your clients, without a PR database subscription.
What the module deliberately does not do
- It does not send emails. Only what you send goes out, from your own mailbox.
- It does not buy links or broker paid placements.
- It does not promise publication. It makes sure your request reaches the right page and the right person.
Frequently asked questions
Does RankCockpit send emails on my behalf?
No. The system writes drafts and opens them in the right mailbox. Only what you send yourself goes out, so your sender reputation stays in your hands.
Where do the mention targets come from?
From citation data: pages that AI answers cite as sources while your client's brand is missing. On top of that come the Google top 10 and AI Overview sources for your queries, the referring domains of your competitors and, if you want, industry directories. There is no purchased media list.
What does a run cost?
Costs accrue in your own API accounts. Contact research runs at about 1 to 3 cents per domain, a fact-checked draft at about 10 to 16 cents.
From the gap in the AI answer to the mention
Signals, keyword research, briefings, content and reporting are part of it. After the handover, all of it belongs to you.