/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Anthropic debuts Citations, a new API feature that lets developers “ground” answers in source documents, available for Claude 3.5 Sonnet and Claude 3.5 Haiku

At least we're in good company: … Forums: Hacker News : Citations on the Anthropic API

TechCrunch Kyle Wiggers

Context & Ripple Effects

Anthropic had been expanding Claude’s developer workflow with prompt-generation and testing tools and an analysis capability for code, calculations, and file data. Citations adds source traceability to that stack for applications built on Claude 3.5 Sonnet and Haiku.

The feature matters because it makes document-grounded output a product capability rather than an implementation developers must assemble entirely themselves. It also precedes Anthropic’s later addition of web search to the API, extending the kinds of sources Claude-based applications can work from.

First-order effects

  • Developers using Claude 3.5 Sonnet or Haiku can attach source support to answers generated from documents, making it easier for end users to inspect where a response came from.
  • Anthropic strengthens its API proposition for document-heavy applications where unsupported model output is harder to use operationally.

Second-order effects

  • Application builders can shift effort from constructing basic source-attribution layers toward integrating citations into their own review, retrieval, and user-interface workflows.
  • Competing model APIs face added pressure to make grounding and source visibility standard features, not merely developer-built patterns.

Third-order effects

  • If document-level attribution becomes a baseline API capability, enterprise AI differentiation will increasingly depend on the quality, governance, and usability of the underlying information corpus rather than fluent generation alone.
  • The pattern points toward AI systems designed to expose evidence alongside answers, creating a clearer foundation for human review in high-accountability workflows.

The trend: Generative-AI platforms are evolving from standalone answer engines into evidence-aware application infrastructure built around governed source material.

Discussion

  • @crisgiardina Cristiano Giardina on x
    Anthropic launching a built-in RAG system essentially. Pretty nice for Enterprise customers who integrate Claude with a knowledge base, now you don't need a vector store etc. Pricing is the same as regular tokens but you don't pay for the quoted text output tokens, which is nice.
  • @alexalbert__ Alex Albert on x
    We've rolled out Citations in the Anthropic API. Citations allows Claude to ground its answers in user-provided information and provide precise references to the sentences and passages used in its responses. Here's how it works: [image]
  • @alexalbert__ Alex Albert on x
    With Citations enabled, Claude can cite chunks of content from pdfs, plain text docs, and text chunks. The returned citations are easy to parse, and Claude makes sure to only highlight the part of the response that is applicable to the citation. [image]
  • @anthropicai @anthropicai on x
    Introducing Citations. Our new API feature lets Claude ground its answers in sources you provide. Claude can then cite the specific sentences and passages that inform each response. [image]