# Conflicting Sources in Context

## The Problem

Retrieved chunks contain contradictory information from different documents, causing AI to provide inconsistent or confused answers.

### Symptoms

* ❌ AI hedges: "Some sources say X, others say Y"
* ❌ Contradictory statements in response
* ❌ Unclear which source is authoritative
* ❌ Old vs new docs both retrieved
* ❌ Different product tiers mixed up

### Real-World Example

```
Retrieved chunks:
→ Chunk A (2022 docs): "API rate limit is 100 req/hour"
→ Chunk B (2024 docs): "API rate limit is 1000 req/hour"

Both retrieved with similar scores

AI response: "The API rate limit is 100 requests per hour, though
some documentation indicates it may be 1000 requests per hour."

Confused answer - which is correct?
```

***

## Deep Technical Analysis

### Version Conflicts

**Document Versioning:**

```
Same document, multiple versions:
→ v1.0: Feature X works this way
→ v2.0: Feature X redesigned

Both embedded in knowledge base:
→ No version filtering
→ Both retrieved
→ AI must reconcile incompatible info
```

**Timestamp Metadata:**

```
Solution: Tag with version/date
{
  metadata: {
    version: "2.0",
    published_date: "2024-01-01"
  }
}

Retrieval filter:
→ Prefer latest version
→ Deprioritize old docs
```

### Multi-Source Inconsistencies

**Different Systems of Record:**

```
Knowledge base contains:
→ Public docs: Rate limit 100/hour
→ Internal wiki: Rate limit 200/hour (recent increase)
→ API spec: Rate limit 1000/hour (not yet documented)

Which is authoritative?
→ No source prioritization
→ Equal weight
→ Conflicting retrieval
```

**Source Trust Levels:**

```
Assign authority scores:
→ Official API spec: Priority 1
→ Published docs: Priority 2
→ Internal wiki: Priority 3
→ Community forums: Priority 4

Retrieval boosting:
→ Boost high-priority sources
→ Use lower-priority only if high unavailable
```

### Reconciliation Strategies

**LLM Arbitration:**

```
Prompt: "If sources conflict, prefer the most recent"

AI must:
→ Detect conflict
→ Check dates
→ Choose most recent
→ But: LLM not 100% reliable at this
```

**Pre-Retrieval Filtering:**

```
Better: Filter before retrieval
→ Only retrieve docs from last 6 months
→ Exclude deprecated content
→ Tag as "archived" and filter out

Prevents conflict at source
```

### Citation Clarity

**Source Attribution:**

```
When conflict exists:
"According to API Specification v2.0 (Jan 2024), rate limit
is 1000/hour. Note: Older documentation (2022) states 100/hour
- this is outdated."

Explicit:
→ Which source
→ When published
→ Why one supersedes other
```

***

## How to Solve

**Tag chunks with version/timestamp + filter to most recent docs + assign source authority levels (official > community) + implement conflict detection + prefer highest-priority source + add "last updated" in responses + archive old docs with metadata flag + prompt LLM to prefer recent over old when conflict detected.** See [Source Conflicts](/rag-scenarios-and-solutions/accuracy/conflicting-info.md).


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://help.twig.so/rag-scenarios-and-solutions/accuracy/conflicting-info.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.