diff --git a/docs/modules/sanitize-html.md b/docs/modules/sanitize-html.md
new file mode 100644
index 00000000..0ed342cd
--- /dev/null
+++ b/docs/modules/sanitize-html.md
@@ -0,0 +1,89 @@
+---
+description: Replacing sanitize-html with neosanitize, a zero-dependency, isomorphic HTML sanitizer with a drop-in legacy engine and a faster WHATWG-based engine
+---
+
+# Replacements for `sanitize-html`
+
+[`sanitize-html`](https://github.com/apostrophecms/sanitize-html) is a widely used HTML sanitizer, but it pulls in a large dependency tree (including `htmlparser2`, `parse-srcset`, `postcss`, and others) and only runs in Node-like environments.
+
+[`neosanitize`](https://github.com/PuruVJ/neosanitize) is a zero-dependency, isomorphic (works in Node, browsers, and edge runtimes) HTML sanitizer written in TypeScript. It ships two engines:
+
+- A **legacy** engine (`neosanitize/legacy`) that is a byte-identical, drop-in port of `sanitize-html` 2.x — same API, same options, same output.
+- A **root** engine (`neosanitize`) built on a fast, browser-faithful WHATWG parser with a deny-by-default policy, recommended for new code and for the best performance and security.
+
+## `neosanitize/legacy` (drop-in replacement)
+
+The legacy engine mirrors the `sanitize-html` API and output exactly, so migrating is usually just a matter of changing the import:
+
+```ts
+import sanitizeHtml from 'sanitize-html' // [!code --]
+import sanitizeHtml from 'neosanitize/legacy' // [!code ++]
+
+const clean = sanitizeHtml(dirty, {
+  allowedTags: ['b', 'i', 'em', 'strong', 'a'],
+  allowedAttributes: {
+    a: ['href']
+  }
+})
+```
+
+All `sanitize-html` options (`allowedTags`, `allowedAttributes`, `allowedSchemes`, `transformTags`, `exclusiveFilter`, etc.) are supported with identical behaviour.
+
+## `neosanitize` (recommended)
+
+For new code, the root module is recommended. It is built on a fast, browser-faithful WHATWG parser (100% html5lib tokenizer conformance) and follows a deny-by-default policy, so only what you explicitly allow gets through.
+
+Unlike `sanitize-html`, the root engine uses a builder pattern: you compile a policy once with `Sanitizer.builder(...)`, then reuse the resulting sanitizer everywhere.
+
+```ts
+import { Sanitizer } from 'neosanitize'
+import * as presets from 'neosanitize/presets'
+
+// Build once (compiles the policy), reuse everywhere.
+const sanitizer = Sanitizer.builder(presets.ugc)
+  .allow('img', ['src', 'alt'])
+  .build()
+
+sanitizer.sanitize(
+  '<p>hi <img src=x onerror=alert(1)> <script>bad()</script></p>'
+)
+// → '<p>hi <img src="x"> </p>'
+```
+
+You can also build a policy from scratch instead of starting from a preset, and chain `.allow()` / `.deny()` to refine it:
+
+```ts
+const sanitizer = Sanitizer.builder({
+  tags: ['a', 'b', 'p'],
+  attrs: { a: ['href'] }
+})
+  .allow('img', ['src', 'alt'])
+  .deny('span')
+  .build()
+
+sanitizer.sanitize(
+  '<p>see <a href="/docs" onclick="x()">docs</a><iframe></iframe></p>'
+)
+// → '<p>see <a href="/docs">docs</a></p>'
+```
+
+Because it has no dependencies, the same code runs in browsers and edge runtimes unchanged (bundlers resolve `neosanitize` to the browser build automatically). The underlying parser can also be used directly via `neosanitize/whatwg-parser`.
+
+### Swapping the parser
+
+The default parser is the fastest option and is browser-faithful with 100% html5lib tokenizer conformance. If you need different trade-offs, the parser is pluggable via `.parser(...)` on the builder — the policy engine and sanitization logic stay identical, only the HTML-to-tree step changes. Adapters for [`parse5`](https://github.com/inikulin/parse5) and [`htmlparser2`](https://github.com/fb55/htmlparser2) ship as separate exports (their libraries are peer dependencies, loaded only when imported):
+
+```ts
+import { Sanitizer } from 'neosanitize'
+import { parse5Adapter } from 'neosanitize/parse5'
+import { htmlparser2Adapter } from 'neosanitize/htmlparser2'
+import * as presets from 'neosanitize/presets'
+
+// Full WHATWG spec conformance on adversarial markup (~50% throughput)
+const strict = Sanitizer.builder(presets.ugc).parser(parse5Adapter).build()
+
+// Fast, lenient parsing (no full WHATWG tree builder — e.g. no foster-parenting)
+const lenient = Sanitizer.builder(presets.ugc)
+  .parser(htmlparser2Adapter)
+  .build()
+```
diff --git a/manifests/preferred.json b/manifests/preferred.json
index bd638b5b..01248bf3 100644
--- a/manifests/preferred.json
+++ b/manifests/preferred.json
@@ -2540,6 +2540,12 @@
       "replacements": ["crypto.timingSafeEqual"],
       "url": {"type": "e18e", "id": "buffer-equal-constant-time"}
     },
+    "sanitize-html": {
+      "type": "module",
+      "moduleName": "sanitize-html",
+      "replacements": ["neosanitize"],
+      "url": {"type": "e18e", "id": "sanitize-html"}
+    },
     "scmp": {
       "type": "module",
       "moduleName": "scmp",
@@ -3364,6 +3370,11 @@
       "type": "documented",
       "replacementModule": "neoqs"
     },
+    "neosanitize": {
+      "id": "neosanitize",
+      "type": "documented",
+      "replacementModule": "neosanitize"
+    },
     "neotraverse": {
       "id": "neotraverse",
       "type": "documented",