Skip to main content

Citation Tracking

Understanding how AEO/GEO Analytics extracts, tracks, and analyzes citations from LLM responses.

What is a Citation?​

Citation: A URL or domain reference that appears in an LLM's response to a query.

Citation Examples​

Direct URL Citation:

Response: "According to this guide (https://familyhandyman.com/
plumbing/faucet-repair/), you should first turn off the water supply..."

Citation Extracted: familyhandyman.com

Inline Reference:

Response: "Family Handyman recommends turning off the water supply
before starting any faucet repair..."

Citation Extracted: May or may not include URL depending on provider

Multiple Citations:

Response: "Several sources including This Old House, Home Depot,
and YouTube tutorials suggest the following steps..."

Citations Extracted: thisoldhouse.com, homedepot.com, youtube.com

How Citation Extraction Works​

The Extraction Process​

Step 1: LLM Response

  • Query sent to provider (Claude, GPT-4, etc.)
  • Provider generates response with citations
  • Full text returned to AEO/GEO Analytics

Step 2: URL Detection

  • Regular expression scans response text
  • Identifies patterns matching URLs
  • Extracts complete URL strings

Step 3: Domain Parsing

  • Full URL broken down
  • Root domain extracted
  • Subdomains removed (www. stripped)

Step 4: Storage

  • Citation saved to database
  • Linked to original query and response
  • Domain tracked separately for aggregation

Example Flow​

Query: "How to fix a leaking faucet?"

Provider Response:
"To fix a leaking faucet, follow these steps from
https://familyhandyman.com/plumbing/faucet-repair/:
1. Turn off water supply
2. Remove faucet handle
..."

Extraction:
URL Found: https://familyhandyman.com/plumbing/faucet-repair/
Domain Parsed: familyhandyman.com
Stored: Citation record created
        Domain counter incremented

Citation Patterns​

Provider Differences​

Claude (Anthropic):

  • Highest citation rate (6-10 per response)
  • Explicit URL format
  • Clear source attribution
  • Educational and how-to sites favored

GPT-4 (OpenAI):

  • Medium-high citation rate (4-8 per response)
  • Mix of URLs and source names
  • Balanced source types
  • Review and comparison sites favored

Gemini (Google AI):

  • Lower citation rate (3-6 per response)
  • Search-oriented citations
  • Google properties sometimes favored
  • Educational sites common

Perplexity:

  • Medium citation rate (4-7 per response)
  • Real-time web search based
  • Fresh content cited
  • News and current sources

Query Type Impact​

How-To Queries:

  • Citation Rate: High (7-12 citations)
  • Source Types: Tutorial sites, YouTube, how-to guides
  • Why: Multiple approaches exist, LLM cites several

Comparison Queries:

  • Citation Rate: Medium-High (5-9 citations)
  • Source Types: Review sites, comparison tools
  • Why: Need multiple perspectives

Problem-Solving:

  • Citation Rate: High (6-10 citations)
  • Source Types: Troubleshooting guides, expert sites
  • Why: Multiple solutions cited

Educational:

  • Citation Rate: Medium (4-6 citations)
  • Source Types: Wikipedia, .edu sites, encyclopedias
  • Why: Fewer authoritative sources

Decision-Making:

  • Citation Rate: Medium (4-6 citations)
  • Source Types: Expert opinions, buying guides
  • Why: Specific recommendations

Citation Quality Indicators​

Position in Response​

Early Citation (First paragraph):

  • Highest visibility
  • User sees immediately
  • Primary source signal

Mid-Response Citation:

  • Supporting evidence
  • Additional context
  • Medium visibility

Late Citation (Last paragraph, footnotes):

  • Secondary sources
  • Further reading
  • Lower visibility

Attribution Style​

Explicit Attribution:

"According to Family Handyman (familyhandyman.com)..."
  • High visibility
  • Clear source
  • Trust building

Implicit Attribution:

"Most plumbers recommend... (sources: familyhandyman.com, thisoldhouse.com)"
  • Medium visibility
  • Aggregated sources
  • Less direct

URL Only:

"...(https://familyhandyman.com/guide)"
  • Lower visibility
  • Technical reference
  • Click-through possible

Citation Context​

Recommendation Citation:

  • "Expert source X recommends..."
  • High trust signal
  • Authority building

Data Citation:

  • "According to research..."
  • Credibility signal
  • Statistical authority

Example Citation:

  • "As shown in this guide..."
  • Practical application
  • How-to authority

Domain Tracking​

What We Track​

Per Domain:

  • Total citation count
  • First seen date
  • Last seen date
  • Queries triggering citations
  • Providers citing it
  • Specific URLs cited

Aggregate:

  • Citation share percentage
  • Rank among all domains
  • Trending up or down

Domain Authority Signals​

High Citation Count:

  • Frequent citations = trusted source
  • Consistent across queries = breadth
  • Multiple providers = universal trust

Citation Diversity:

  • Many different URLs = comprehensive
  • Various query types = versatile
  • All providers = robust authority

Citation Consistency:

  • Always cited for topic = go-to source
  • First position often = primary authority
  • Explicit attribution = recognized expert

Citation Metrics​

Primary Metrics​

Total Citations:

  • Absolute count across all responses
  • Growth metric over time
  • Volume indicator

Citation Rate:

  • Citations per response
  • Quality indicator
  • Provider comparison metric

Citation Share:

  • Your citations ÷ Total citations
  • Competitive metric
  • Market position indicator

Secondary Metrics​

Unique Domains:

  • Number of different domains cited
  • Diversity indicator
  • Competition level

Average Citations per Query:

  • Total citations ÷ Total queries
  • Query quality indicator
  • Expected citation baseline

Citation Distribution:

  • How evenly citations spread
  • Monopoly vs competitive landscape
  • Opportunity identification

Citation Analysis Techniques​

Competitive Benchmarking​

Compare Your Domain:

  1. Execute 50 queries in your industry
  2. Track citation counts by domain
  3. Rank domains by citation share
  4. Identify your position

Example Results:

1. competitor-a.com - 15% share (150 citations)
2. competitor-b.com - 12% share (120 citations)
3. industry-leader.com - 10% share (100 citations)
...
10. your-domain.com - 3% share (30 citations)

Insights:

  • Gap from leaders: 12% points
  • Opportunity: Increase share by 4x
  • Focus: Match competitor content coverage

Content Gap Analysis​

Find Uncited Queries:

  1. Execute query set
  2. Identify queries with 0-2 citations
  3. Low competition = opportunity

Example:

Query: "How to replace toilet flapper valve?"
Citations: 2 (very low)

Opportunity: Create definitive guide
Competition: Minimal
Potential: High citation share capture

Track Over Time:

  • Weekly: Execute same query set
  • Monthly: Compare citation shares
  • Quarterly: Trend analysis

Example Trend:

Jan: your-domain.com - 3% share
Feb: your-domain.com - 4% share (+33%)
Mar: your-domain.com - 5.5% share (+38%)

Trend: Growing share, positive trajectory
Action: Continue current strategy

Why Citations Matter​

Direct Benefits​

Brand Awareness:

  • Domain appears in AI responses
  • Users see your brand name
  • Builds familiarity over time

Traffic Potential:

  • Users may click citations
  • 5-15% CTR typical
  • Qualified traffic

Trust Building:

  • AI trusts your content
  • Users trust AI recommendations
  • Transitive trust effect

Indirect Benefits​

Search Rankings:

  • AI citations may signal quality to Google
  • Authority building across platforms
  • Backlink-like effect

Brand Recall:

  • Even without clicks, visibility matters
  • "I've heard of that before"
  • Future search behavior influenced

Content Validation:

  • Citations confirm content quality
  • Topic coverage verification
  • Strategic direction validation

Citation Limitations​

What Citations Don't Tell You​

User Behavior:

  • Can't track if user clicked
  • Can't measure time on site
  • Can't attribute conversions

Citation Intent:

  • Positive or negative mention?
  • Agreement or counterexample?
  • Primary or supplementary?

Full Context:

  • How citation was framed
  • What else was cited
  • User's ultimate choice

Measurement Challenges​

Provider Variability:

  • Different citation rates
  • Different source preferences
  • Inconsistent formatting

Query Dependence:

  • Some queries cite more than others
  • Can't compare across industries easily
  • Context matters

Temporal Effects:

  • Provider updates change behavior
  • Content freshness varies
  • Seasonal patterns exist

Best Practices​

Testing Methodology​

Consistent Query Sets:

  • Use same queries over time
  • Compare apples to apples
  • Track trends accurately

Multiple Providers:

  • Don't rely on single LLM
  • Provider diversity important
  • Comprehensive view

Regular Cadence:

  • Weekly for active monitoring
  • Monthly for trend analysis
  • Quarterly for strategy

Data Interpretation​

Statistical Significance:

  • Minimum 20-50 queries for patterns
  • Don't over-interpret single citations
  • Look for consistent trends

Context Awareness:

  • Consider query type
  • Account for provider differences
  • Note market changes

Competitive Context:

  • Always compare to benchmarks
  • Track relative not just absolute
  • Industry varies widely

Action-Oriented Analysis​

Identify Opportunities:

  • Low-citation queries = create content
  • Competitor gaps = differentiation
  • Trending topics = prioritize

Measure Impact:

  • Before/after content updates
  • A/B test content approaches
  • Validate hypotheses

Iterate Quickly:

  • Test → Measure → Refine
  • Don't wait for perfect
  • Continuous improvement

Next Steps​