diff --git a/docs/product/ai-in-sentry/seer/autofix/index.mdx b/docs/product/ai-in-sentry/seer/autofix/index.mdx
index 1b961c856b7955..cd980930fed231 100644
--- a/docs/product/ai-in-sentry/seer/autofix/index.mdx
+++ b/docs/product/ai-in-sentry/seer/autofix/index.mdx
@@ -40,10 +40,39 @@ Seer's Issue Autofix flow follows a three step process:
 2. [Solution Identification](#solution-identification): Identify a solution based on the analysis
 3. [Code Generation](#code-generation): Generate a code fix and (optionally) open a PR, or hand off to a coding agent
 
-You can always manually trigger the Autofix flow from the Issue Details page. Issues will automatically trigger the flow when an agent is configured for background handoff and the Issue is highly actionable with 10+ events captured and high medium or above fixability score.
+You can always manually trigger the Autofix flow from the Issue Details page.
+
+### When Autofix Runs Automatically
+
+When automation is enabled in your project's Seer settings, Sentry can also trigger Autofix automatically on issues that meet the following criteria:
+
+1. **The issue has 10 or more events** — Sentry waits until the issue has captured at least 10 events to ensure it's worth investigating.
+2. **The issue occurred within the last 14 days** — Only recent issues are eligible for automatic runs.
+3. **The issue has a sufficient fixability score** — Sentry's ML model must estimate the issue is reasonably fixable (see below).
+
+#### Fixability Score
+
+When Seer evaluates an issue, it assigns a **fixability score** — a value between 0 and 1 that estimates how straightforward the issue is to resolve. This score is generated by a machine-learning model that analyzes the issue summary and classifies it on a spectrum from "complex and very difficult to resolve" to "simple and easily resolved."
+
+The fixability score determines **whether** Autofix runs automatically, and **how far** it goes:
+
+- **Below 0.40 (Low)** — The issue is not considered fixable enough for automation. Autofix will not run automatically, but you can still trigger it manually.
+- **0.40 – 0.66 (Medium)** — Autofix runs automatically but stops at **root cause analysis**. Sentry identifies the likely cause and presents it for your review.
+- **0.66 – 0.78 (High)** — Autofix runs automatically through **code changes**. Sentry identifies the root cause and suggests a fix.
+- **0.78 and above (Very High)** — Autofix runs automatically and can **open a pull request** with the suggested fix.
+
+#### Capping Automation Depth
+
+Even when the fixability score suggests a deeper level of automation, you can set a maximum stopping point in your project's Seer settings. The available options are:
+
+- **Root cause analysis only** — Autofix identifies the cause but doesn't suggest code changes.
+- **Code changes** — Autofix suggests a fix but doesn't open a PR.
+- **Open a pull request** — Autofix can go all the way to opening a PR (default).
+
+Your chosen stopping point acts as an upper bound — Autofix will never go further than what you allow, regardless of the fixability score.
 
 <Alert>
-If you've recently set up Seer, existing issues will need Autofix to be run manually. 
+If you've recently set up Seer, existing issues will need Autofix to be run manually.
 </Alert>
 
 ### Root Cause Analysis