Browse guides
Measure citations and change
Compare consistent conditions and distinguish bot visits, answer citations, and referrals.
Before you start
Record prompts, markets, engines, publishing URLs, and change dates. Without a baseline, start with the first observation instead of estimating historical improvement.
Step by step
- Save actual answers and cited URLs before publishing. Record the period and collected-answer count.
- Publish reviewed content or correct a source, then verify the public URL.
- For bot-request records, check the URL and time. Page access, model training, and answer citations are different events.
- Inspect links in subsequent answers under the same prompt, engine, and market conditions. A product mention also differs from a linked citation.
- Record mention and citation changes separately from GA4 referrals and ad performance. Explain changes in prompt sets or product scope.
- Report changes after publishing as observations. Without a control or another suitable design, do not claim that publishing caused every change.
Check the result
Your report should retain comparison conditions, original answers, source URLs, check times, and limits. Do not count unverified citations as results.
If something goes wrong
- A bot may visit without a citation. Review prompt-page relevance and observe subsequent collections.
- Referrals can change without a citation change. Check each metric’s collection path.