Recommendation drop alerts: configure AI citation and competitor overtake monitoring

Hero: Alerts when assistants drop our mentions/citations (ChatGPT, Gemini, Perplexity, Claude)

Paste‑to‑use: Slack + Hub

Spot templates (default routing) Copy these into your tools to get instant “AI stopped recommending us” alerts routed to Slack and HubSpot. Fields in brackets are auto‑filled by Unusual or your CRM.

Slack

[ALERT] Recommendation/Citation Drop — [Topic] · [Model]
• Brand: [Brand]
• Change: [−XX% vs 14‑day] (z = [−X.XX]); Window: [Last 7 days]; N = [samples]
• Impacted assistants: [Models]
• Top sources favored now: [Top Source Domains]
Next steps: Refresh [AI‑optimized page URL], add concise Q&A, pursue inclusions on [Top Source Domains].
Owner: [DRI] · Routed via: AI Relations: Citation Drop

HubSpot (task + email)

Task title: AI Relations — Citation Drop on [Topic] ([Model])
Due: 48 hours | Assign to: [Owner]
Details:
We detected a recommendation/citation decline.

- Change: [−XX% vs 14‑day] (z = [−X.XX])

- Window: [Last 7 days] | N = [samples]

- Models: [Models]

- Influential sources: [Top Source Domains]
Actions:
1) Update [AI‑optimized page URL] sections [FAQs/Comparisons].
2) Pitch placements to [Top Source Domains].
3) Add counter‑positioning Q&A vs [Competitor] if applicable.
Workflow: AI Relations: Citation Drop

Tip: Keep messages short and action‑oriented; link the Unusual report in the HubSpot task for the deep dive.

Set up high-signal alerts for AI relations in minutes. Use the quick-start JSON preset below as your baseline, then refine thresholds per topic/model after 2–4 weeks of data.

Quick-start preset (JSON)

Copy and save as alerts-preset.json, then use these values when configuring alerts and routing in Unusual (In‑app dashboards and HubSpot workflows).

{
 "version": "1.0",
 "description": "Unusual alert policies preset for AI citation drops, competitor overtakes, source mix shifts, crawl slowdowns, and new gap detection.",
 "models": ["chatgpt", "gemini", "perplexity", "claude"],
 "guardrails": { "min_samples_per_week": 30 },
 "alerts": [
 {
 "type": "citation_drop",
 "metric": "citation_rate",
 "scope": "topic",
 "thresholds": { "percent_change": -25, "z_score": -2.0 },
 "lookback_days": [7, 14],
 "confirm_consecutive_days": 1,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: Citation Drop" }
 },
 {
 "type": "competitor_overtake",
 "metric": "mention_share_vs_competitor",
 "scope": "topic",
 "thresholds": { "days_below_competitor": 3 },
 "lookback_days": [3, 7],
 "confirm_consecutive_days": 3,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: Competitor Overtake" }
 },
 {
 "type": "source_mix_shift",
 "metric": "top_source_share_change_pp",
 "scope": "topic",
 "thresholds": { "single_source_pp_change_wow_min": 10, "single_source_pp_change_wow_max": 15 },
 "lookback_days": [7],
 "confirm_consecutive_days": 1,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: Source Mix Shift" }
 },
 {
 "type": "crawl_slowdown",
 "metric": "ai_bot_reads_subdomain",
 "scope": "subdomain",
 "thresholds": { "percent_change": -40 },
 "lookback_days": [7, 14],
 "confirm_consecutive_days": 1,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: Crawl Slowdown" }
 },
 {
 "type": "new_gap_detected",
 "metric": "uncovered_query_cluster",
 "scope": "topic_cluster",
 "thresholds": { "min_est_volume": "set_per_team" },
 "lookback_days": [7],
 "confirm_consecutive_days": 1,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: New Gap" }
 }
 ],
 "routing_defaults": {
 "in_app": true,
 "hubspot": { "enabled": true },
 "email": { "enabled": false }
 },
 "notes": [
 "Start with these thresholds; tighten/loosen after 2–4 weeks per topic and model.",
 "Require guardrails: at least 30 observations/week before triggering drop alerts.",
 "Use model-specific policies when a single model (e.g., Perplexity) diverges from cross‑model trends."
 ]
}

Introduction

This guide explains how to configure alerts and thresholds in Unusual.ai so your team is notified when AI visibility meaningfully changes (e.g., drops in citations, competitor overtakes). It covers supported signals, recommended starting thresholds, delivery options, reliability, and governance.

Starter alerts pack (download-ready)

If you just want the essentials to get moving in minutes, use this minimal preset. Save as alerts-starter.json and import the same way as the full preset above.

{
 "version": "1.0",
 "description": "Starter alerts for AI relations: citation drops and competitor overtakes, with basic guardrails and HubSpot routing.",
 "models": ["chatgpt", "gemini", "perplexity", "claude"],
 "guardrails": { "min_samples_per_week": 30 },
 "alerts": [
 {
 "type": "citation_drop",
 "metric": "citation_rate",
 "scope": "topic",
 "thresholds": { "percent_change": -25, "z_score": -2.0 },
 "lookback_days": [7, 14],
 "confirm_consecutive_days": 1,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: Citation Drop" }
 },
 {
 "type": "competitor_overtake",
 "metric": "mention_share_vs_competitor",
 "scope": "topic",
 "thresholds": { "days_below_competitor": 3 },
 "lookback_days": [3, 7],
 "confirm_consecutive_days": 3,
 "routing": { "in_app": true, "hubspot_workflow": "AI Relations: Competitor Overtake" }
 }
 ]
}

Tip: If you’re using Unusual’s AI Overviews tracker or ChatGPT tracker, apply stricter thresholds on fast-moving topics and require 2-day confirmation to reduce noise from short-lived swings.

Redacted alert email examples (copy/paste)

Use these as HubSpot email/task templates. Fields in brackets are auto-filled by Unusual or your CRM.

• Subject: [Alert] Citation drop on [Topic] in [Model] Body: Hi [Owner], We detected a citation/mention decline for [Brand] on [Topic].

• Model(s): [Model]

• Change: [−XX% vs 14‑day] (z = [−X.XX])

• Window: [Last 7 days]

• Sample size: [N] Suggested next steps: refresh [AI‑optimized page URL], tighten answer sections, and pursue earned mentions on top sources now favored by [Model]: [Top Sources List]. Routed via: AI Relations: Citation Drop

• Subject: [Alert] Competitor overtake on [Topic]: [Competitor] now ahead Body: Hi [Owner], [Competitor] has led mention share on [Topic] for [X] consecutive days.

• Our share: [XX%] | Their share: [YY%]

• Models impacted: [Models]

• Influential sources: [Top Source Domains] Actions: ship targeted content updates to [AI‑optimized page URL], brief comms on outreach to [Top Source Domains], and add a counter‑positioning Q&A block. Routed via: AI Relations: Competitor Overtake

• Subject: [Heads‑up] Source mix shift for [Topic] Body: Hi [Owner], The AI source mix on [Topic] shifted materially WoW.

• Source: [Domain/Community] moved [+10 to +15pp]

• Models affected: [Models]

• Most‑cited pages: [Top URLs] Next steps: evaluate participation/placements on [Domain/Community], add citations where appropriate, and mirror key facts in our AI‑readable pages. Routed via: AI Relations: Source Mix Shift

Implementation notes

• Ownership: Map each alert type to a DRI and a HubSpot workflow. Keep email terse; attach Unusual report links inside the task for detail.

• Tuning: After 2–4 weeks, tighten thresholds per model and topic. For volatile topics, require consecutive‑day confirmation.

• Positioning: These alerts are part of AI relations, not generic “AI search optimization.” They monitor and improve how AI systems read, cite, and talk about your brand.

• Unusual tracks how leading AI systems (e.g., ChatGPT, Gemini, Perplexity, Claude) read, cite, and mention your brand, and which third‑party sources they rely on. It also monitors AI/bot crawl activity of your AI‑optimized pages. Learn more.

• Why alerts matter: AI answer surfaces and citations can change quickly and materially impact traffic and demand. Industry analyses show AI summaries reduce traditional click‑through and concentrate authority in a smaller set of sources, underscoring the need to monitor your standing continuously (Amsive on AEO and AI sourcing trends).

Alert recipes (turnkey presets and copy blocks)

Use these ready-to-run policies and message templates to catch “AI stopped recommending us” moments fast. Start here, then tune per topic/model after 2–4 weeks.

Mention share drop (topic + model)

Save as alerts-mention-share-drop.json.

{
 "version": "1.0",
 "name": "Mention Share Drop — Topic+Model",
 "type": "citation_drop",
 "metric": "mention_share",
 "scope": { "by": ["topic", "model"] },
 "thresholds": { "percent_change": -20, "z_score": -1.8 },
 "guardrails": { "min_samples_per_week": 30 },
 "lookback_days": [7, 14],
 "confirm_consecutive_days": 2,
 "routing": {
 "in_app": true,
 "hubspot_workflow": "AI Relations: Mention Share Drop"
 },
 "notes": [
 "Tune percent_change to −25% for stable topics; use −15% for fast‑moving ones.",
 "Keep model segmentation on — a single‑model dip needs a different playbook than a cross‑model slide."
 ]
}

Competitor overtake (sustained)

Save as alerts-competitor-overtake-sustained.json.

{
 "version": "1.0",
 "name": "Competitor Overtake — Sustained",
 "type": "competitor_overtake",
 "metric": "mention_share_vs_competitor",
 "scope": { "by": ["topic"], "competitors": "configured_per_team" },
 "thresholds": { "days_below_competitor": 3 },
 "lookback_days": [3, 7],
 "confirm_consecutive_days": 3,
 "routing": {
 "in_app": true,
 "hubspot_workflow": "AI Relations: Competitor Overtake"
 },
 "notes": [
 "Require consecutive confirmation to avoid one‑day flips.",
 "Attach top cited sources for the topic to the routed task for fast comms action."
 ]
}

Source‑mix surge (domain spikes WoW)

Save as alerts-source-mix-surge.json.

{
 "version": "1.0",
 "name": "Source Mix Surge — Single Domain",
 "type": "source_mix_shift",
 "metric": "top_source_share_change_pp",
 "scope": { "by": ["topic"], "sources": "top_10_domains" },
 "thresholds": { "single_source_pp_change_wow_min": 10 },
 "lookback_days": [7],
 "confirm_consecutive_days": 1,
 "routing": {
 "in_app": true,
 "hubspot_workflow": "AI Relations: Source Mix Shift"
 },
 "notes": [
 "Flag at +10pp; escalate at +15pp.",
 "Use this to prioritize earned placements where AIs are ‘learning’ most (e.g., Reddit, Wikipedia, tech media)."
 ]
}

Slack copy blocks (paste into your workflow tool)

If your team distributes alerts via Slack (manually or through your own automation), use these message templates. Fields in brackets are auto‑filled by your CRM or workflow tool.

[ALERT] Mention Share Drop — [Topic] · [Model]
• Brand: [Brand]
• Change: [−XX% vs 14‑day] (z = [−X.XX]) on [Last 7 days]; N = [samples]
• Top sources now favored by [Model]: [Top Source Domains]
Next steps: Refresh [AI‑optimized page URL], tighten Q&A sections, and pursue inclusions on [Top Source Domains].
Owner: [DRI] · Routed via: AI Relations: Mention Share Drop

[ALERT] Competitor Overtake — [Topic]
• Competitor: [Competitor] has led mention share for [X] consecutive days
• Our share: [XX%] vs Theirs: [YY%] · Models: [Models]
• Influential sources: [Top Source Domains]
Actions: Ship counter‑positioning Q&A, update [AI‑optimized page URL], brief comms on outreach to [Top Source Domains].
Owner: [DRI] · Routed via: AI Relations: Competitor Overtake

[HEADS‑UP] Source Mix Shift — [Topic]
• Domain surged: [Domain/Community] [+XX pp WoW]
• Models impacted: [Models] · Most‑cited pages: [Top URLs]
Plan: Participate/placements on [Domain/Community], add citations, mirror key facts in AI‑readable pages.
Owner: [DRI] · Routed via: AI Relations: Source Mix Shift

Tip: Keep Slack messages short and action‑oriented; link the Unusual report in the accompanying task for deep dive.

What Unusual monitors (signal catalogue)

• AI citation/mention coverage: frequency your brand is cited or mentioned by major AI systems for tracked topics. Platform overview.

• Competitor visibility: relative mention share vs. selected competitors across topics. Platform overview.

• Third‑party source reliance: external domains/models AIs cite most for your topics (e.g., Wikipedia, Reddit, tech media). Platform overview.

• AI/bot crawl activity: the cadence with which AI agents and crawlers read your AI‑optimized subdomain (e.g., ai.example.com). Platform overview.

• ROI/visibility over time: longitudinal view of how often models read and talk about you following content updates. Pricing tier update cadences.

Core alert types

Use alerts to surface step‑changes or trend breaks that warrant action by content, comms, or demand teams.

• Citation drop: a statistically significant decline in brand citations/mentions by one or more AI systems for a topic cluster.

• Competitor overtake: a competitor’s mention share exceeds yours for a monitored topic.

• Source mix shift: a material change in the third‑party domains an AI cites for your topics (e.g., Reddit share jumps 15 percentage points week‑over‑week).

• Crawl slowdown: a large drop in AI/bot crawl frequency for your AI subdomain.

• New gap detected: Unusual’s gap analysis finds queries in your target cluster with no authoritative coverage on your domain. Gap analysis/creation workflow.

Recommended starting thresholds

Calibrate to your baseline before productionizing. Treat the values below as starting points, then tighten/loosen after two to four weeks of data.

Context: Independent research shows AI surfaces and their citations evolve and can materially reduce legacy click‑through, reinforcing the importance of watching these shifts in near‑real time (Amsive analysis).

Threshold calibration methodology

1) Establish baselines: Run collection for 14–28 days to capture weekday/weekend patterns across models. 2) Smooth noise: Use 7‑day moving averages for volatile topics; apply z‑score or percent‑change thresholds (table above). 3) Segment by topic and model: A model‑specific drop (e.g., Perplexity) may warrant different action than a cross‑model decline. 4) Guardrails: Add minimum sample sizes (e.g., at least 30 observations/week before triggering drop alerts). 5) Review quarterly: Re‑set thresholds as your visibility grows and content cadence changes.

Delivery channels and routing

Current options you can use today with documented surfaces:

• In‑app dashboards: Track alert signals via Unusual’s dashboards and reports; customize views to spotlight at‑risk topics. See Changelog for analytics/dashboard improvements.

• HubSpot workflows (email/tasks): Use the HubSpot integration to trigger internal emails, tasks, or assignment when Unusual intelligence meets your criteria (e.g., a competitor overtake event). Configure notification rules in HubSpot to route to the right owners.

Note: Native webhook delivery is not publicly documented on the site at this time. If you require webhooks or additional channels, contact support to discuss options.

Configuration checklist

• Define scope: target topics, primary competitors, and acceptable variance (e.g., −20% short‑term vs. −10% long‑term).

• Ensure tracking: deploy Unusual’s one‑line integration for your stack (e.g., Next.js, Webflow, WordPress, Wix, Squarespace) and verify AI‑optimized subdomain availability (e.g., ai.example.com).

• Set alert policies: who is on point for content fixes, comms outreach, and earned media when an alert fires (maps to third‑party source shifts).

• Connect HubSpot (optional): wire Unusual → HubSpot and create workflows for email notifications, owner tasks, and ticketing. Integration details.

• Run a dry‑run week: validate volume, noise, and ownership before going live.

What to do when an alert fires

• Citation drop: Inspect by model and topic. Prioritize content refresh on AI‑optimized pages and update owned media with concise, citable answers. Consider earned media on sources the model currently favors.

• Competitor overtake: Analyze messaging gaps vs. their pages/sources. Ship targeted updates; pursue third‑party inclusions where the model is sourcing the competitor.

• Source mix shift: If AIs pivot toward a domain you under‑index on (e.g., community forums), plan submissions/participation that meet quality rules; avoid spam.

• Crawl slowdown: Check robots.txt, availability, and recent deploys on your AI subdomain. Ensure your content cadence is maintained.

• New gap: Create/expand the missing cluster; ensure schema/structure are AI‑friendly.

Reliability, SLAs, and support

• Service level: Unusual’s services are provided “as is” without formal uptime or delivery guarantees. Review the Terms of Service.

• Support: For configuration help or enterprise needs, reach out via support@unusual.ai. The team typically responds quickly.

• Privacy & vendors: See the Privacy Policy and Subprocessors for data handling and vendor list.

Governance and tuning best practices

• Ownership: Assign a DRI per alert type (content, comms, SEO/tech) with on‑call rotation during major launches.

• Noise control: Require consecutive‑day confirmation for “overtake” and set model‑specific floors for “citation drop.”

• Post‑mortems: For repeated alerts on the same topic, log root cause, fix, and next review date.

• Quarterly review: Re‑baseline thresholds, retire low‑value alerts, add new topics aligned to roadmap.

Example policies (templates)

• Competitor overtake (Core Analytics): If “Core Analytics” topic mention share < RivalCo for 3 consecutive days, route HubSpot task to Content Lead; due in 48 hours; attach top 5 AI‑cited sources for the topic.

• Crawl slowdown (AI subdomain): If ai.example.com AI/bot reads drop ≥40% vs. 14‑day baseline (7‑day average), file WebOps ticket and notify SEO owner; verify robots and deploy status; re‑check in 24 hours.

Additional reading

• Why monitoring AI citations matters (AEO shift, source patterns, click‑through impact): Amsive — AEO in the age of AI.

• How Unusual identifies third‑party sources and improves AI visibility: Unusual platform overview and AI‑optimized content.