[Call for Feedback] Proposal to sunset experimental Writer and Rewriter APIs in favor of the Prompt API
Kenji Baheux
Hello,
As part of our commitment to incubation and listening to real-world developer experience, we are opening a community consultation window regarding our proposal to sunset the experimental Writer and Rewriter APIs (window.Writer and window.Rewriter) in favor of the foundational Prompt API (window.LanguageModel).
The Journey So Far
The Writer and Rewriter APIs were originally explored as dedicated task-specific abstractions. However, through feedback from the dev & origin trials, web developers overwhelmingly told us they prefer using the Prompt API directly to implement their own writing, tone, and rewriting workflows.
Because our initial implementation of Writer and Rewriter operated as templatized system prompts over our general-purpose on-device model, developers achieved identical or superior results with custom prompting.
Furthermore, rapid advancements in our base on-device models (including Gemini Nano and Gemma 4) have made general-purpose prompting exceptionally capable at tone, length, and format transformations without requiring specialized fine-tuned weights, complex system templates, and dedicated IDL surfaces.
Streamlining Built-in AI
Retiring these thin abstractions allows us to reduce specification churn and browser maintenance toil, refocusing our platform engineering resources on areas with distinct architectural value:
Specialized task APIs where dedicated fine-tuning yields unique client performance or quality gains (such as the Summarizer, Translator, and Language Detector APIs).
Foundational platform primitives that unlock new client-side capabilities such as the Semantic Embedder API, native Tool Calling, Thinking Mode, and more recently our exploration of a Decisions API!
Transition Resources & Polyfills
To ensure developers currently experimenting with Writer or Rewriter experience zero disruption:
Zero-dependency Polyfill: We have published an open-source, zero-dependency drop-in polyfill that maps window.Writer and window.Rewriter calls directly onto window.LanguageModel: https://github.com/GoogleChromeLabs/web-ai-demos/tree/main/writer-rewriter-polyfills
Canonical System Prompts: We have documented the exact system prompt templates used by the underlying implementation so you can replicate or customize them directly with the Prompt API.
Call for Feedback (Open through November 17, 2026)
We want to ensure no critical developer workflows or unmet needs are overlooked before proceeding with Chromium code removal (targeted for Chrome M160 in January 2027).
If you are currently evaluating Writer or Rewriter and feel that direct prompting via window.LanguageModel falls short of your performance, latency, or UX requirements, please let us know:
Complete our short feedback questionnaire.
Or file an issue on the Writing Assistance APIs repository.
Best regards,