Citation experiments

Test what makes your pages more likely to be cited

Kelsey turns page improvements into controlled citation experiments. Establish a baseline, test one evidence-backed change, and verify whether the same buyer questions cite your page more often after it is published.

Kelsey Citation Experiments editor showing side-by-side draft review with evidence attached

How it works

From a missed citation to a verified improvement

AI answers naturally change between runs, so one promising response is not proof. Kelsey keeps the page, buyer questions, model settings, and testing conditions fixed before and after your change.

Baseline

Page and buyer questions locked

Diagnose

Competing sources reviewed

Change

One improvement approved

Verify

The same questions run again

Citation experiment

Improve pricing page citation eligibility

Ready to publish

Target page

/pricing

Target questions

3 buyer questions

Control questions

2 unchanged questions

Baseline citation rate

22%

2 of 9 runs

Test conditions

Locked for verification

Controlled page improvements

Test one meaningful change at a time

Kelsey compares your page with the sources currently winning your buyer questions and identifies the clearest difference worth testing.

Instead of rewriting an entire page and leaving you unsure what helped, each experiment isolates one change. Other recommendations stay available as possible future tests.

Recommended change

Add a direct plan comparison

Competing pricing pages clearly explain which plan is best for different customer types. Your current page lists features, but does not answer that decision directly.

Predicted improvement

This change may make the page easier to retrieve and use for questions about pricing, team size, and plan selection.

Review recommended change Start experiment
Original
Proposed

Plans for every team

Choose the plan that fits your needs and upgrade as your team grows.

Which plan is right for you?

The Starter plan is best for small marketing teams tracking a single brand. Pro is designed for teams that need more buyer questions, additional experiments, and deeper competitive analysis.

Evidence attached

  • 3 target buyer questions
  • 4 competing pricing pages reviewed
  • Named plan recommendations found across winning sources
Approve change Edit draft

Review before publishing

See exactly what Kelsey wants to change and why

Review the original section beside the proposed version, with the relevant buyer questions, cited sources, and competing page patterns attached.

Nothing is published automatically. Your team decides whether to accept the change, edit it, and publish it through your existing workflow.

Kelsey Citation Experiments editor with publish options and CMS handoff

Indexing and retrieval

Kelsey waits until the new version can actually be found

Running the experiment immediately after publishing could test the old version of your page. Kelsey first confirms that the updated section is live, extractable, and available for retrieval.

Only then does Kelsey repeat the locked buyer questions and begin measuring the result.

Published

Updated section found on the live page

Extraction confirmed

New passage matches the approved version

Retrieval ready

Page is available for verification

Verification running

Repeating the original buyer questions

Verified results

Compare the same questions before and after

Once the updated page is ready, Kelsey repeats the same questions under the same conditions and measures three separate outcomes:

Cited

Did the page appear in the answer’s citations?

Retrieved

Did the model find the page while researching the answer?

Used

Did the final answer use information from the changed section?

Control questions help Kelsey distinguish an improvement to your page from normal movement across AI answers.

Experiment result

Citation rate improved

Verified improvement

Baseline citation rate

22%

2 of 9 runs

Current citation rate

67%

6 of 9 runs

Control result

Stable

Verdict: Verified improvement

The updated plan comparison increased citations for the target buyer questions while the control group remained stable.

Next experiment

Add supporting evidence for the plan recommendations and test whether the page is used more consistently in comparison answers.

View complete results Start next experiment

Predicted vs. verified

Know what Kelsey expects and what actually happened

Before publishing, Kelsey can predict whether a change may improve page extraction, retrieval, or citation eligibility. That prediction is always labeled clearly.

A result is only labeled Verified after the updated page is available for retrieval and the original questions have been repeated.

Kelsey never treats a single answer as proof and never mixes predicted improvements into your verified citation rate.

Run your first citation experiment

Start with a page that is already losing valuable buyer questions. Kelsey will identify the clearest improvement to test, guide you through the change, and verify whether it made a measurable difference.

Start an experiment