Citations & Sources

Every Queria answer includes precise citations linking each piece of information to its original document. This guarantees transparency, traceability and reliability.

Inline citations

In the answer text you find numbered references in square brackets: [1], [2], [3], and so on. Each number corresponds to a specific document or source consulted by the AI.

Example:

The ordinary statute of limitations is ten years [1], unless the law provides otherwise [2]. For periodic obligations the five-year term applies [3].

Sources footer

Below every answer an expandable panel shows the complete details of each cited source:

• Title of the document or source
• Type of source (identified by colored badge)
• Excerpt of the relevant passage
• Relevance score as a percentage
• Date of the document (when available)

Badge system

Color badges identify the origin of each citation:

Badge	Color	Description
Document	Blue	Company documents uploaded to the platform
Knowledge Base	Green	Content from the curated knowledge base
Legal Source	Amber	Legislation and regulations from certified sources
Food Source	Orange	Food data from certified sources
Chem Source	Teal	Chemical data and SDS from certified sources
Pharma Source	Purple	Pharmaceutical data from certified sources
Web Page	Gray	Web pages indexed by the system

Interacting with citations

Click a citation [N] in the text to:

• Highlight the corresponding source in the footer
• View the excerpt from the original document
• Open the full document (when available)

The link between citation and source is bidirectional: you can also click a source in the footer to highlight where it's cited in the text.

Citation map

Every answer internally includes a citationMap that associates each citation number with the specific document. This mechanism ensures:

• Every claim is linked to a verifiable source
• No orphan citations or wrong references
• Numbering is always consistent and sequential

Relevance score

Next to every source a percentage score indicates how pertinent the source content is to the question:

• 90-100%: very high match, the source directly answers the question
• 70-89%: good match, strongly correlated information
• 50-69%: partial match, useful but not direct context
• Below 50%: the source is generally excluded by re-ranking

How to verify information

To make sure of an answer's accuracy:

1. Check the badges: identify the source types used
2. Verify the score: high-score sources are more reliable
3. Read the excerpts: compare the cited text with the generated answer
4. Open the document: access the full original content
5. Cross-check sources: answers with citations from multiple different sources are generally more robust

Why citations matter

Queria's citation system is designed to counter AI hallucinations:

• Anti-hallucination: the AI is instructed to base answers only on real documents
• Traceability: every piece of information can be traced back to its source
• Trust: users can independently verify claims
• Compliance: critical in regulated industries where the source matters as much as the content
• Transparency: the AI's decision-making process is visible and verifiable

Notice

If an answer contains no citations, the AI is responding from its general knowledge rather than from specific documents. In those cases, always verify the information with independent sources.

---

Queria v3.5.0 -- Cog-RAG Architecture