Writer API

Stay organized with collections Save and categorize content based on your preferences.

Alexandra Klepper

Thomas Steiner

Published: May 20, 2025

Explainer	Web	Extensions	Chrome Status	Intent
GitHub	Origin trial	Origin trial	View	Intent to Experiment

The Writer API helps you create new content that conforms to a specified writing task. The Writer API and the Rewriter API are part of the Writing Assistance APIs proposal.

These partner APIs can help you improve content created by users.

Use cases

Write new content, based on your initial idea and optional context. This could be used to:

• Support users write any kind of content, like reviews, blog posts, or emails.
• Help users write better support requests.
• Draft an introduction for a series of work samples, to better capture certain skills.

Is your use case missing? Join the early preview program to share your feedback.

Get started

Join the Writer API origin trial, running in Chrome 137 to 142.

Sign up for the origin trial

The Writer API is available in a joint origin trial with the Rewriter API. To start using these APIs:

1. Acknowledge Google's Generative AI Prohibited Uses Policy.
2. Go to the Writer API origin trial.
3. Click Register and fill out the form. In the Web origin field, provide your origin or extension ID, chrome-extension://YOUR_EXTENSION_ID.
4. To submit, click Register.
5. Copy the token provided, and add it to every participating web page on your origin or include it in your Extension manifest.
6. Start using the Writer and Rewriter APIs.

Learn more about how to get started with origin trials.

Add support to localhost

To access the Writer and Rewriter APIs on localhost during the origin trial, you must update Chrome to the latest version. Then, follow these steps:

1. Go to chrome://flags/#writer-api-for-gemini-nano.
2. Select Enabled.
3. Click Relaunch or restart Chrome.

Use the Writer API

First, run feature detection to see if the browser supports these APIs.

if ('Writer' in self) {
  // The Writer API is supported.
}

The Writer API, and all other built-in AI APIs, are integrated in the browser. Gemini Nano is downloaded separately the first time any website uses a built-in AI API. In practice, if a user has already interacted with a built-in API, they have downloaded the model to their browser.

To determine if the model is ready to use, call the asynchronous Writer.availability() function. It returns a string that can take four possible values:

• unavailable: The browser supports the Writer API, but it can't be used at the moment. This could be for a number of reasons, such as insufficient available disk space to download the model.
• available: The browser supports the Writer API, and it can be used right away.
• downloadable: The browser supports the Writer API, but it needs to download the model first.
• downloading: The browser supports the Writer API, and is currently downloading the model.

To trigger model download and start the writer, call the Writer.create() function. If the response to availability() was downloadable, listen for download progress and inform the user, as the download may take time.

const writer = await Writer.create({
  monitor(m) {
    m.addEventListener("downloadprogress", e => {
      console.log(`Downloaded ${e.loaded * 100}%`);
    });
  }
});

API functions

The create() function lets you configure a new writer object. It takes an optional options object with the following parameters:

• tone: Writing tone can refer to the style, character, or attitude of the content. The value can be set to formal, neutral (default), or casual.
• format: The output formatting, with the allowed values markdown (default) and plain-text.
• length: The length of the output, with the allowed values short, medium (default), and long.
• sharedContext: When writing multiple outputs, a shared context can help the model create content better aligned with your expectations.

The following example demonstrates how to initiate a writer object:

const options = {
  sharedContext: 'This is an email to acquaintances about an upcoming event.',
  tone: 'casual',
  format: 'plain-text',
  length: 'medium',
};

const available = await Writer.availability();
let writer;
if (available === 'unavailable') {
  // The Writer API isn't usable.
  return;
}
if (available === 'available') {
  // The Writer API can be used immediately .
  writer = await Writer.create(options);
} else {
  // The Writer can be used after the model is downloaded.
  writer = await Writer.create(options);
  writer.addEventListener('downloadprogress', (e) => {
    console.log(e.loaded, e.total);
  });
}

Start writing

There are two ways to output writing from the model: non-streaming and streaming.

Non-streaming output

With non-streaming writing, the model processes the input as a whole and then produces the output.

To get a non-streaming output, call the asynchronous write() function. You must include a prompt for the content you want written. You can add an optional context to provide the model background information, which may help the model better meet your expectations for the output.

// Non-streaming
const writer = await Writer.create();
const result = await writer.write(
  "An inquiry to my bank about how to enable wire transfers on my account.", {
    context: "I'm a longstanding customer",
  },
);

Stream writing output

Streaming offers results in real-time. The output updates continuously as the input is added and adjusted.

To get a streaming writer, call the writeStreaming() function and iterate over the available segments of text in the stream. You can add an optional context to provide the model background information, which may help the model better meet your expectations for the output.

// Streaming
const writer = await Writer.create();
const stream = writer.writeStreaming(
  "An inquiry to my bank about how to enable wire transfers on my account.", {
    context: "I'm a longstanding customer",
  },
);
for await (const chunk of stream) {
  composeTextbox.append(chunk);
}

Share context for multiple tasks

You may want to use a writer to generate multiple pieces of content. In this case, it's useful to add sharedContext. For example, you may want to help reviewers give better feedback in comments.

// Shared context and per writing task context
const writer = await Writer.create({
sharedContext: "This is for publishing on [popular website name], a business and employment-focused social media platform."
});

const stream = writer.writeStreaming(
  "Write a blog post about how I love all this work on gen AI at Google!" +
  "Mention that there's so much to learn and so many new things I can do!",
  { context: " The request comes from someone working at a startup providing an e-commerce CMS solution."}
);

for await (const chunk of stream) {
  composeTextbox.append(chunk);
}

Reuse a writer

You can use the same writer to create multiple pieces of content.

// Reuse a writer
const writer = await Writer.create({ tone: "formal" });

const reviews = await Promise.all(
  Array.from(
    document.querySelectorAll("#reviews > .review"),
    (reviewEl) => writer.write(reviewEl.textContent)
  ),
);

Stop the writer

To end the writing process, abort the controller and destroy the writer.

// Aborting a writer
const controller = new AbortController();
stopButton.onclick = () => controller.abort();

const writer = await Writer.create({ signal: controller.signal });
await writer.write(reviewEl.textContent, { signal: controller.signal });

// Destroying a writer
writer.destroy();

Demo

Engage and share feedback

The Writer and Rewriter API are under active discussion and subject to change in the future. If you try this API and have feedback, we'd love to hear it.

• Read the explainer, raise questions and participate in discussion.
• Review the implementation for Chrome on Chrome Status.
• Join the early preview program for an early look at new APIs and access to our mailing list.
• If you have feedback on Chrome's implementation, file a Chromium bug.

Discover all of the built-in AI APIs which use models, including Gemini Nano and other expert models, in the browser.