Products Knowledge BaseBuyer QuestionsAI CrawlabilityAI VisibilityCitation SourcesMonitoring & AlertsIn-app ReportsContent CreationOpportunity DiscoveryMedia & DistributionShopify AppExplore the Web Platform
Solutions E-commerce & DTCBrand SitesContent & MediaAgenciesAI EnginesPlatformsGEO for ShopifyView all solutions
Resources Resource CenterBlogLearning CenterGEO GlossaryResearch Reports
PricingTools
About About InsightWonderContact usSecurity
Start free trial
Citation Sources

Trace real AI citations to exact pages.

Inspect cited domains and URLs, then trace each citation back to the question and model that used it.

Primary unit URL and domain
Trace Question and model
Owned evidence Cited brand pages
Integrity Answer markers only
Core capability

What Citation Sources helps the team understand.

Inspect cited domains and URLs, then trace each citation back to the question and model that used it.

01

URL-level evidence

Open the exact page cited inside an AI answer.

02

Domain-level view

See which publishers and communities appear most often.

03

Question trace

Connect every citation to the buyer question and model response.

Operating model

Keep every result connected to evidence and action.

The product explains the result, preserves its source context and makes the next review explicit.

01
Use actual answer citationsDo not count a retrieval candidate as if the model cited it.
02
Keep breadth and count distinctOne question citing many pages should not distort source breadth.
03
Separate citation from recommendationA brand page can be cited without the brand being recommended.
Measurement rules

The rules citation sources follows

A feature list is easy to copy; the rules behind the numbers are not. Each of these is enforced in the product, and most of them were added after getting it wrong once.

Rule Why it exists
Only sources marked inside the answer text count as citations.An assistant typically retrieves a dozen or more pages and cites a few. Counting the retrieval pool inflates citation numbers several times over — the most common way AI visibility reports overstate reach.
A citation marker that does not resolve to a real URL is discarded.Models occasionally emit a marker that points at nothing. We would rather record one citation fewer than record one that is not real.
Breadth is reported alongside volume.One answer citing twelve of your pages is not twelve times the reach. The number of distinct questions that cited a domain is much harder to inflate than the raw count.
Being cited and being recommended are kept apart.Your page can be cited as a source while the answer recommends a competitor. Merging the two flatters the report and hides the actual gap.
Illustrative product evidence

Move from summary metrics into the underlying question.

Sample data demonstrates the evidence depth of a formal project.

InsightWonder / Citation Sources Sample data

Visibility evidence

Last 30 days
AI visibility 34% +4%
Mentioned questions 10/29 +2
Share of voice 31% +4%
Cited pages 27 +9
Buyer question Brand status Evidence
Best modular sofa for a small apartment Gap Competitor cited
Most durable washable sofa under $2,000 Mentioned Product page
Low-VOC sofa for families Recommended Guide + review
FAQ

Questions about Citation Sources

Does this feature run from the public website?

No. Marketing pages explain the capability. Formal data collection begins only after a user enters the App and confirms the project scope.

Can snapshot results become part of this workflow?

Yes. A valid snapshot can be claimed as an initial sample, while onboarding still confirms markets, competitors and the formal monitored question set.

Are all metrics comparable across AI engines?

Core answer states can be normalized, but citations, search assistance and answer formats still need engine-specific evidence.

How often is the data refreshed?

Measurement runs on the schedule your plan allows, and monitoring quota is shared at the account level. Alerts require two consecutive runs to confirm, so model sampling noise does not become a false alarm.

Can the original AI answer be inspected?

Yes. Every result drills down to the answer that produced it, question by question and model by model, together with the citations found in that answer.

What happens when a model is unavailable?

That run is marked partial rather than being silently filled in. The remaining engines still complete, and the gap stays visible in the result.

From signal to action

Put Citation Sources into a complete GEO workflow.

Create a project, confirm your market and start measuring the questions that matter.