Zicy for Publishers

What this guide is for, in one sentence: How an editorial organization uses Zicy to see where its journalism goes after the click stopped coming — which topics and pages still earn AI’s citation, where your work is used without credit, and what to publish next.

Who this is for: Editorial and media teams who’ve watched search referrals fall for two years straight while their rankings held — and who need the KPI stack for the era that replaced the click.

---

The problem you’re living

The traffic didn’t dip; the mechanism changed. Search referrals have collapsed hardest for the smallest publishers, click-through rates have cratered even on page-one positions, and AI referrals — however fast they’re growing — remain a small fraction of what search once sent. Meanwhile the citation slots inside AI answers are themselves shrinking: fewer links per answer, more competition for each one. Being the primary citation is the new page one — and the clicks it does send are better readers, pre-qualified by the answer itself.

So this guide begins with the honest sentence, because you’d distrust anything else: no tool restores the old traffic. What can be built is the intelligence layer for the new economics — your traffic dashboard shows what you lost; this shows what you’re owed: which pages AI relies on, which topics you still own, where your reporting is used without credit, and what evidence you’d bring to a licensing conversation.

How the loop maps to your newsroom

• Measure — cited-source tracking: are we the source AI routes readers to, per topic, per engine?
• Diagnose — entity accuracy, topic commissioning, and the citation map — including uncredited use.
• Act — archive modernisation and new coverage, structured to be extractable and attributable.
• Prove — which articles AI sends readers to, on which engines, tied to real sessions.

The worked example across the module articles is GreenGrid Media — Nadia’s energy publication — whose three-article arc (Key Topics → Citation Analysis → Site Audit) reads as one investigation, ending with a single robots.txt line that had silently blocked one engine’s crawler from the site’s most-cited section. Read it in order; it’s the diagnose loop working.

Your four core workflows

1. Track the beat, not the brand (Measure). Tag prompts by coverage area (Setting Up Prompt Tracking) so every question becomes “are we the cited source on this topic, on this engine?” The per-prompt view shows each engine’s full answer and its citations — the ground truth your editorial instincts get checked against.

2. Commission from the demand map (Diagnose). Key Topics Analysis is your commissioning meeting rendered as a screen: Leaders you own, Battlegrounds a rival leads, Blind Spots with proven demand and no coverage — shortlisted, demand-ranked, with each topic’s current owner labeled by type. When a manufacturer leads a topic, no publisher has written the definitive independent piece; those are often the cheapest wins on the board. Export the shortlist and it’s your pitch list.

3. Build the evidence file (Diagnose). Citation Analysis answers the publisher’s core question with receipts: your cited share, the rival that wins instead, and — the number your industry has never had — answers that likely use your reporting without credit, each with the prompt, the engine, the matched page, and the full answer attached. Reviewed and accumulated, that trail is negotiation ammunition for licensing and partnership conversations; the drill-down on your own domain (exact cited URLs, two clicks) is the proof artifact for partners and your own board. And the technical floor matters more than newsrooms expect: Site Audit catches the structural blockers — a robots rule, a missing llms.txt — that decide citation races before editorial quality gets a vote.

4. Modernise and prove (Act → Prove). The Content Optimiser makes archive coverage extractable without rewriting it; the Action Center ships schema and llms.txt so AI attributes your identity and sections correctly. Then Site Traffic closes the loop: which articles AI actually routes readers to, from which engines — smaller numbers than the old world, but attached to readers who arrive pre-qualified and stay longer.

Honest guidance — measure the owed, don’t chase the lost. Reporting this program against old referral baselines guarantees it looks like failure. The KPI stack that fits the era: cited share by topic, primary-citation wins on your beats, the uncredited-use count and its trend, and the quality (not just quantity) of AI-referred sessions. Those are also the numbers a licensing counterparty can’t wave away.

Honest guidance — the uncredited flag is a screen, not a verdict. It works on phrasing similarity. Read every flagged answer against the matched page before anything external; the count screens, the rows convict.

Your first 30 days

Week 1 — Instrument the beats. Profile live; prompts tagged by coverage area; first analyses complete. Run Site Audit immediately — a structural blocker found in week one saves a quarter of misdiagnosis. Week 2 — The commissioning map. Take the Key Topics Top-10 shortlist into your editorial meeting; baseline Citation Analysis, including your first uncredited-use count. Save both — they’re your before-state. Week 3 — Fix the floor, fill one gap. Ship the audit fixes (robots, llms.txt, schema); commission one Blind Spot piece and modernise two archive pieces on a Battleground topic. Week 4 — The evidence file. Connect analytics in Site Traffic; assemble the first monthly evidence file — cited pages with drill-downs, uncredited-use rows, topic positions — and decide who inside (and eventually outside) the building needs to see it.

What to read next

• Key Topics Analysis — the commissioning map; start here
• Citation Analysis — credit, competitors, and the uncredited-use count
• Site Audit — the technical floor under every citation race
• Site Traffic — what AI actually sends, and what those readers do