Prompt
How do I measure which source pages drive AI citations?
Latest observation
To measure which source pages drive AI citations, treat it like a referral attribution problem with a few AI-specific twists.
1) Start by defining what counts as an “AI citation”
Pick a consistent definition, for example:
- A visible link in an AI answer
- A cited source panel / footnote / reference
- A mention of your domain even without a link
- A page surfaced in “More sources” or expandable citations
Be clear whether you care about:
- Domain-level citations or page-level citations
- Direct citations vs. mentions that influenced the answer
- One-time exposure vs. repeat appearance across queries
2) Capture citation events
You usually need two layers of data:
A. AI response monitoring
Track when your site is cited in:
- ChatGPT / GPT search experiences
- Perplexity
- Google AI Overviews
- Copilot / other answer engines
For each citation event, store:
- Date/time
- AI platform
- Query/topic
- Cited URL
- Citation position/order
- Whether it was a link, snippet, or mention
- Geography/device if available
B. Your own web analytics
In GA4, server logs, or a warehouse, track:
- Landing page
- Referrer
- UTM tags if present
- Session quality metrics
This helps connect cited source pages to actual downstream traffic.
3) Normalize URLs
AI citations can be messy:
- Canonicalize URLs
- Remove tracking params
- Resolve redirects
- Group variants of the same page
Without normalization, you’ll overcount or split credit across duplicates.
4) Build a citation-to-page mapping
Create a table like:
| AI Platform | Query | Cited URL | Canonical Page | Date | Citation Type |
|---|
Then aggregate by canonical page:
- Number of citations
- Number of unique queries
- Number of unique AI platforms
- Estimated share of citations
This tells you which source pages are most frequently cited.
5) Measure “driving” impact, not just appearance
A cited page may not be the one that generates traffic or conversions. Add these metrics:
- Citation frequency: how often the page is cited
- Citation diversity: how many distinct queries it supports
- Traffic lift: visits after citations
- Engagement quality: bounce rate, time on page, conversion rate
- Assisted conversions: did AI-driven visits later convert?
A page that’s cited often but never brings useful traffic may be less valuable than a page with fewer but high-intent citations.
6) Use controlled tests to infer causality
If you want to know which pages cause citations, do experiments:
- Update a page with clearer facts, structured headings, and citations
- Create a new page targeting a topic
- Compare citation frequency before/after
- Use time-based comparisons or A/B-style content changes where possible
If a page starts appearing in AI answers more often after improvements, that’s useful evidence.
7) Use server logs when possible
If AI bots access your content directly, logs can show:
- Which pages are being crawled
- How often
- By which bots/user agents
This won’t prove citation, but it can help identify pages likely contributing to AI training/retrieval.
8) Track content attributes of cited pages
Often the pages that get cited share patterns:
- Short, direct answers
- Clear headings
- Lists and tables
- High topical authority
- Strong internal linking
- Up-to-date statistics
- Schema markup
Capturing these attributes helps explain why certain pages drive citations.
9) Recommended reporting dashboard
Build a dashboard with:
- Top cited pages
- Top cited queries
- Citation share by page
- Traffic from AI platforms by page
- Conversion rate from AI traffic
- New citations over time
- Pages gaining/losing citation share
10) Practical stack
A common setup:
- AI citation collection: manual monitoring, scraping, or third-party AI visibility tools
- Analytics: GA4 + Search Console + server logs
- Storage: BigQuery / Snowflake / SQL database
- Dashboard: Looker Studio / Tableau / Power BI
Simple formula
For each page:
AI Citation Score =
- citation count
- × query diversity
- × platform diversity
- × downstream traffic/conversion value
You can weight these based on your goals.
If you want, I can give you:
- a spreadsheet template for tracking AI citations,
- a SQL schema for storing citation data, or
- a step-by-step GA4 + dashboard setup.