Overview
How might we design source quality scores so people can trust and act on AI output?
When this pattern fits
- Ideal for research tools, information platforms, and applications where displaying source quality scores helps users assess information credibility.
When to skip or lighten it
- Closed corpora of equally trusted internal docs where scores add noise.
- When the scoring model is opaque and frequently contradicts user judgment.
- Creative generation with no real sources.
States
State model coming soon
Key UX elements
Key UX elements coming soon
Anti-patterns to avoid
Green “trusted” badges with no criteria explained.
Scoring the domain only while citing a low-quality page on that domain.
Hiding low-quality sources instead of showing them with a clear weak score.
Using scores as a substitute for excerpts and links.
How products use it
| Product | Implementation |
|---|---|
| Academic / research AI (Elicit, Consensus) | Signals paper quality, citation counts, or study type beside claims. |
| Enterprise search | Ranks internal sources by freshness, ACL, and owner authority. |
| Perplexity | Source lists emphasize reputable publishers alongside raw URLs. |
| News AI products | Publisher credibility labels next to grounded summaries. |
Implementation
Copy this prompt to generate a production-ready implementation in Cursor, Claude Code, Lovable, or any AI coding agent.
Generate a production-ready implementation of the "Source Quality Scores" AI interface design pattern.
Pattern Definition:Frequently asked questions
What are source quality scores in AI UX?
Source quality scores are visible reliability ratings for the documents or sites behind an AI answer, helping users decide how much to trust each citation.
Should every citation show a quality score?
Score when the corpus is mixed and users must prioritize. Skip decorative scores when all sources are pre-vetted or when you cannot explain the rating.
How should low-quality sources be handled?
Show them with an honest weak score or “unverified” label rather than deleting them silently, and prefer higher-quality alternatives when available.