Sources, Citations & Diagnosis

QueryBurst Answer Engine retains the full evidence trail for each answer. A source can be retrieved, selected, grounded, or cited — and can disappear at any stage. This makes it possible to diagnose where a page entered or exited the process, rather than only observing the final citations.

Can I see sources that were retrieved but never cited?

Yes.

Answer Engine retains the candidate evidence trail rather than only showing the final citations.

A source can move through stages such as:

retrieved → selected → grounded → cited

and can disappear at any point.

Is retrieval the same thing as citation?

No.

A page can be fetched, chunked, and scored without being selected.

It can be selected without becoming a visible citation.

And a cited page does not necessarily mean the cited company was recommended. QueryBurst’s breakdown of sources, citations, mentions and recommendations explains why these distinctions matter.

These are separate events.

Is being cited by AI automatically a success?

No.

A source can support a factual claim while the answer recommends a competitor.

Citation is useful diagnostic evidence.

It is not automatically the commercial outcome.

The more useful question is whether your brand or product influenced the answer in the direction you care about.

What is the knowledge graph showing?

Answer Engine can extract entities, relationships, n-grams, and semantic triples from the retrieved evidence and organise them into a graph.

This is an analytical representation of the evidence environment surrounding the answer.

It is not a claim that you are viewing ChatGPT's literal internal knowledge graph.

It helps answer questions such as:

  • which entities dominate the retrieved evidence
  • what relationships repeatedly appear
  • which concepts cluster together
  • how products, features, criteria, and sources are connected

Can I test my own page even if it was not naturally retrieved?

Yes.

Answer Engine can inject a crawled page from your own site into the candidate evidence set.

This is useful as a counterfactual test.

Instead of asking only:

"Why wasn't this page discovered?"

you can temporarily bypass discovery and ask:

"If the model had retrieved this page, would the evidence have been competitive?"

Discovery and evidence fitness are different problems.

What is skeptical mode?

Skeptical mode applies stronger scrutiny to self-serving evidence.

For example, a company ranking itself number one in its own "best products" article should not necessarily carry the same evidential weight as independent corroboration.

The aim is to investigate how the evidence set changes when obvious self-interest is treated more cautiously.