Prompt Tracking — Per-Prompt AI Visibility, Alert Badges, and Content Cluster Picking

Prompt Tracking — Your Operational AEO Page

The Prompt Tracking page (left sidebar → Monitor → Prompt Tracking) is the most-used surface in Citadex. It shows every prompt you're tracking, your coverage % per platform, alert badges for what's wrong, and the controls for generating new Strategy Articles from selected clusters.

Most users spend more time on this page than any other. This guide covers how to read it and act on it.

Adding Prompts to Tracking

Prompts arrive in your tracking list through three paths:

1. Onboarding — Your initial prompt library is generated when you create a project, based on your selected topics.
2. Prompt Library → Discover tab — Add prompts manually (auto-expands ~10 related ones) OR use the seed picker to expand from one of your existing tracked prompts (commercial-intent prompts are flagged for you — those make better seeds).
3. Bulk paste / CSV upload — Prompt Library → Bulk Upload tab for importing 50+ prompts at once.

Every prompt added auto-enters Citadex's scan queue. The queue runs sequentially with 3-attempt retry budget per prompt and exponential backoff — transient failures retry automatically.

Reading the Table

Each row shows one tracked prompt. Columns:

• Prompt — the query text + alert badge (if any) + template tag (listicle / comparison / how-to / faq / standard) + generation count ("已生成 N 次")
• Coverage % — the share of tracked AI platforms where your brand appears in the answer for this prompt
• Platforms — small icon row showing per-platform status (mentioned / cited / not mentioned)
• Last Checked — when the most recent scan ran

The Alert Badges — Triage Order

Two badges you'll see most often:

• ⚠️ Competitor Intercept (red) — competitors appeared but you didn't. Highest urgency. Click to see which platforms, which competitors, and a suggested fix playbook.
• ⚠️ Low Visibility (yellow) — neither you nor obvious competitors appear. Could be content gap or topic novelty.

The badge shows the most severe alert; +N indicates additional alerts of equal or lower severity. For full badge semantics, see Prompt Tracking Alert Badges.

The "Generate Strategy Article" Banner

Above the prompt table, when you have ≥1 uncovered prompt or selected rows, you'll see a blue banner with the Generate Strategy Article button.

This is the operational core of Citadex content creation:

1. Tick 3–7 related prompts in the table (shared topic, shared intent)
2. Click Generate Strategy Article
3. The modal opens with your selection pre-filled
4. Pick a generation mode (Classic / Multi-Agent Standard / Multi-Agent Deep Research)
5. Click Generate — output lands in Content Drafts

For the full workflow, see Strategy Article Workflow. For the agent pipeline details, see Multi-Agent Content Generation.

If you don't tick anything and click the banner, the modal opens in AI Auto-Pick mode — it chooses a cluster for you from your uncovered prompts.

Filtering and Sorting

The toolbar above the table includes:

• Search box — fuzzy match against prompt text
• Coverage filter — All / Covered (≥50%) / Uncovered (<50%) / Intercepted (uncovered with competitors)
• Category filter — by template type (listicle / comparison / how-to / faq / standard)
• Sort — by coverage, last checked, alphabetical, or custom drag-and-drop order

The "uncovered" count for the Generate Strategy Article banner counts your entire project (not just the filtered view) so the banner stays visible whenever you have gaps to act on.

The Evidence Panel

Click any row's evidence icon (or the badge) to open the evidence modal. You'll see:

• Platform tabs — switch between ChatGPT, Claude, Gemini, Perplexity, Copilot, Grok, DeepSeek, Meta AI
• The actual snippet of the AI's answer for this prompt on this platform
• "Competitors found in this answer" — filtered list of real brand names extracted from the answer (headings and topic phrases are excluded)
• Detected at — scan timestamp

This is the ground-truth view of how AI is talking about your brand right now. Read these snippets before making content decisions.

Bulk Operations

Once you tick multiple rows, an action bar appears at the bottom of the screen:

• Delete N prompts — remove them from tracking (data is preserved for 30 days for undo)
• Recover N deleted — if you just deleted by mistake, an undo banner appears at top for ~10 seconds
• Generate Strategy Article from N — same as the top banner but pre-filled with your bulk selection

Cross-Page Sync (Recent Improvement)

When you add prompts from the Prompt Library (a separate page), the Prompt Tracking page now silent-refetches automatically to pick them up. Previously, you had to manually refresh — that's been fixed. Newly added prompts appear in the table and enter the scan queue without intervention.

Frequently Asked Questions

Q: How is my "position" determined when the AI doesn't return a numbered list?

The parser analyses the order of mention in the response. The first brand mentioned in a relevant context is treated as position 1. For clearly ordered lists, we extract the position directly.

Q: Why does my coverage % change without me publishing anything new?

AI model updates, changes to training data, competitor content improvements, and new industry publications can all shift rankings. This is why continuous tracking matters.

Q: Some prompts say "checking…" forever — what's wrong?

The auto-scan loop has a 3-attempt retry budget. After 3 failed attempts, prompts are marked permanently failed for the session — refresh the page to retry. If failures persist across multiple sessions, check your API quota status in Settings → Billing (rate-limit errors from upstream AI providers can cascade).

Q: I added 50 prompts at once, but only some scanned. Why?

Each prompt scan is a real API call to each tracked AI platform. The queue runs sequentially to avoid rate limits. For 50 prompts you should see steady progression over 30–60 minutes. If progress stalls, check the scan loop in browser DevTools or refresh to restart.

Q: What's the fastest way to improve coverage on a low-coverage prompt?

1. Click into the evidence to see what the AI IS saying (and which competitors are winning, if any)
2. Generate a Strategy Article targeting this prompt cluster with Deep Research mode
3. Publish to your blog + syndicate via the extension
4. Re-scan in 7 days to measure delta

Q: Can I drag prompts to reorder?

Yes. Switch to "Custom" sort in the toolbar, then drag rows to reorder. Order persists per-project.