ai-search-monitoring
This skill should be used when the user asks to "track AI search visibility", "monitor brand mentions in ChatGPT", "measure AEO performance", "track citations in Perplexity", "monitor AI search rankings", "set up AI search tracking", "measure generative search visibility", "track brand in AI engines", or any variation of monitoring, tracking, or measuring brand visibility, citations, and performance across AI search engines.
AI Search Monitoring
You can't optimize what you don't measure. AI search monitoring tracks how often — and how accurately — AI engines cite your brand when users ask queries you should own. This is fundamentally different from SEO monitoring. Google Search Console doesn't track AI citations. Rank trackers don't show Perplexity results. You need a separate monitoring stack.
Most SaaS companies discover they're invisible in AI search because they never checked. By the time they notice, competitors have been cited for months and are entrenched. Start monitoring before you start optimizing.
What to Monitor
The four metrics that matter
| Metric | Definition | Why it matters |
|---|---|---|
| Citation rate | % of target queries where your brand is cited as a source | The primary AEO KPI. Are you being cited at all? |
| Citation accuracy | % of citations where the AI's answer about you is factually correct | Being cited with wrong info is worse than not being cited |
| Share of voice | Your citation count vs. competitors across the same query set | Relative position matters. If a competitor is cited 3x more, they're winning |
| Citation retention | % of citations maintained month-over-month | Measures whether you're holding position or losing ground |
Secondary metrics
| Metric | Definition | When to track |
|---|---|---|
| Answer sentiment | Is the AI's answer about you positive, neutral, or negative? | When your brand is established enough to have sentiment variation |
| Source position | Where in the citation list do you appear (1st, 2nd, 3rd)? | When you're consistently cited but want to improve prominence |
| Query coverage gap | Target queries where you're NOT cited | Always — this is your optimization roadmap |
| Engine-specific performance | Citation rate per engine (ChatGPT vs Perplexity vs Gemini) | When you see inconsistent citation across engines |
Monitoring Tools
Dedicated AI search monitoring platforms
| Tool | What it does | Best for | Pricing |
|---|---|---|---|
| Profound | Tracks brand visibility across ChatGPT, Perplexity, Gemini, Claude. Automated query monitoring | Comprehensive automated monitoring at scale | Paid — starts ~$200/month |
| Otterly | AI search rank tracking. Monitors citation position across engines | Citation position tracking | Paid — starts ~$100/month |
| PeecAI | AI search monitoring focused on brand mentions and competitive analysis | Competitive citation analysis | Paid — varies |
| Scrunch AI | Monitors AI-generated answers for your target queries | Broad AI visibility tracking | Paid — varies |
Manual monitoring (free, always do this)
No tool replaces manual testing. Automated tools sample queries periodically. Manual testing catches nuance they miss.
Manual testing process:
- Build a list of 20-50 target queries
- Test each query in ChatGPT, Perplexity, and Gemini
- Record results in a spreadsheet
- Re-test the full list monthly
- Spot-check 5-10 priority queries weekly
Setting Up Your Monitoring System
Step 1: Build the target query list
Your query list is the foundation of all monitoring. Build it carefully.
Query categories to include:
| Category | Examples | # to include |
|---|---|---|
| Brand queries | "What is [Brand]?", "[Brand] pricing", "[Brand] reviews" | 5-10 |
| Competitor comparisons | "[Brand] vs [Competitor]", "[Competitor] alternatives" | 5-10 per major competitor |
| Category queries | "What is [category]?", "Best [category] tools" | 5-10 |
| Problem queries | "How to [solve problem you solve]" | 5-10 |
| Feature queries | "Best tool for [feature]", "How to do [task]" | 5-10 |
Target: 30-50 queries for initial monitoring. Scale to 100+ as your AEO program matures.
Rules:
- Include both Google-style queries ("best CRM tools 2026") and conversational AI queries ("What's the best CRM for a 50-person SaaS startup?"). These may return different citations
- Include queries you currently win AND queries you currently lose. Monitoring only wins gives you a false sense of security
- Update the query list quarterly as buyer language and market evolve
Step 2: Establish the baseline
Before making any AEO changes, record your current state.
Baseline tracking template:
| Query | ChatGPT cited? | ChatGPT accurate? | Perplexity cited? | Perplexity accurate? | Gemini cited? | Gemini accurate? | Competitors cited |
|---|---|---|---|---|---|---|---|
| "What is [Brand]?" | Yes | Yes | Yes | Partial — wrong pricing | No | N/A | Competitor A, Competitor B |
| "[Brand] vs [Competitor]" | No | N/A | Yes | Yes | No | N/A | Competitor A (their own page) |
Record the baseline over 2-3 test sessions (different days) to account for response variability. AI engines don't always return identical answers.
Step 3: Configure automated monitoring
If using a tool like Profound or Otterly:
| Configuration | Setting |
|---|---|
| Query list | Import your 30-50 target queries |
| Engines to monitor | ChatGPT, Perplexity, Gemini (at minimum) |
| Monitoring frequency | Weekly for priority queries, monthly for full list |
| Alerts | Notify on: new citation gained, citation lost, accuracy change |
| Competitor tracking | Add top 3-5 competitors to track their citations alongside yours |
Step 4: Build the dashboard
Whether using a tool or manual tracking, you need a single-view dashboard.
Dashboard elements:
| Element | Visualization | Update frequency |
|---|---|---|
| Overall citation rate | Single number — % of queries where you're cited | Weekly |
| Citation rate trend | Line chart — citation rate over time | Weekly |
| Citation rate by engine | Bar chart — ChatGPT vs Perplexity vs Gemini | Weekly |
| Share of voice vs competitors | Stacked bar — your citations vs competitor citations | Monthly |
| Query coverage gap | Table — queries where you're NOT cited, sorted by priority | Monthly |
| Citation accuracy | % accurate — flag any inaccuracies | Monthly |
Monitoring Workflows
Weekly workflow (15 minutes)
- Check automated alerts for new citations gained or lost
- Spot-check 5 priority queries manually in each engine
- Flag any new inaccuracies for content team to fix
- Update the citation tracking spreadsheet
Monthly workflow (1-2 hours)
- Run the full 30-50 query test across all engines
- Update baseline tracking template with current results
- Calculate citation rate, share of voice, and retention metrics
- Identify top 5 "fixable" gaps (queries where you're close to being cited but aren't)
- Create fix tickets for the content team
- Compare month-over-month trends
Quarterly workflow (half day)
- Re-evaluate the target query list. Remove irrelevant queries, add emerging ones
- Full competitive citation analysis — who's gaining, who's losing?
- Audit citation accuracy across all cited queries
- Assess entity recognition (re-run entity audit questions)
- Adjust AEO strategy based on trends
- Report to stakeholders
Interpreting Results
Common patterns and what they mean
| Pattern | Likely cause | Action |
|---|---|---|
| Cited in Perplexity but not ChatGPT | Perplexity uses real-time retrieval; ChatGPT relies more on training data. Your content is retrievable but not yet in training data | Keep content updated. Entity building will improve ChatGPT citations over time |
| Cited but inaccurately | AI is extracting from an outdated page or a third-party source with wrong info | Update your page with correct info. Add dateModified. Check which source the AI is citing — if it's a third party, try to get them to update |
| Citation rate dropping | A competitor published better content, your content went stale, or the AI engine reweighted sources | Audit the pages that lost citations. Check competitor pages. Refresh content |
| High citation rate, low traffic | Users are getting answers from the AI without visiting your site | This is expected. Focus on citation accuracy and brand visibility. Add CTAs that survive extraction (branded terms, product names) |
| Competitor cited for your brand query | Competitor has a comparison page about you that outranks your own content | Publish or improve your own comparison pages. Ensure your product pages are AEO-optimized |
Pre-Setup Checklist
Before starting AI search monitoring:
- [ ] Target query list built with 30-50 queries across all categories
- [ ] Accounts created for ChatGPT, Perplexity, and Gemini
- [ ] Monitoring tool evaluated and selected (or manual process defined)
- [ ] Baseline measurement completed over 2-3 test sessions
- [ ] Tracking spreadsheet or dashboard created
- [ ] Competitor list defined (top 3-5 competitors to track)
- [ ] Alert thresholds set (what triggers action?)
- [ ] Responsible person assigned for weekly/monthly monitoring
- [ ] Reporting cadence defined (who sees results, how often?)
- [ ] Query list review schedule set (quarterly updates)
Anti-Pattern Check
- Not monitoring at all → You're flying blind. Competitors may be cited for your brand queries right now and you wouldn't know. Set up basic manual monitoring this week — it takes 1 hour
- Only monitoring ChatGPT → Each engine has different citation behavior. A brand cited in Perplexity but not Gemini has different issues than one cited nowhere. Monitor all three major engines
- Checking once and never again → AI citations shift as engines retrain, competitors publish, and content ages. Weekly spot-checks + monthly full audits catch changes before they compound
- Monitoring queries you'll never win → If you're a 10-person startup, don't monitor "best enterprise CRM." Focus on queries where your content can realistically be the best source
- Not acting on monitoring data → Monitoring without action is just watching yourself lose. Every monthly review should produce 3-5 specific content fixes. If it doesn't, your query list needs updating
- Expecting stable results → AI engine responses vary between sessions. A query that cites you today might not tomorrow. Track trends over weeks, not individual test results
Want agents that use skill files like this?
We customize skill files for your brand voice and methodology, then run content agents against them.
Book a call