diff --git a/doc/api/buffer.md b/doc/api/buffer.md
index 7e380f7425fb..840e130bfb13 100644
--- a/doc/api/buffer.md
+++ b/doc/api/buffer.md
@@ -2092,6 +2092,10 @@ console.log(buf.fill('zz', 'hex'));
 <!-- YAML
 added: v5.3.0
 changes:
+  - version: REPLACEME
+    pr-url: https://github.com/nodejs/node/pull/64917
+    description: UTF-16LE and UCS2 searches now inspect every byte offset.
+                 Buffer and Uint8Array values are compared in their entirety.
   - version: v26.1.0
     pr-url: https://github.com/nodejs/node/pull/62390
     description: Added the `end` parameter.
@@ -2160,6 +2164,10 @@ console.log(buf.includes('this', 4));
 <!-- YAML
 added: v1.5.0
 changes:
+  - version: REPLACEME
+    pr-url: https://github.com/nodejs/node/pull/64917
+    description: UTF-16LE and UCS2 searches now inspect every byte offset.
+                 Buffer and Uint8Array values are compared in their entirety.
   - version: v26.1.0
     pr-url: https://github.com/nodejs/node/pull/62390
     description: Added the `end` parameter.
@@ -2343,6 +2351,10 @@ for (const key of buf.keys()) {
 <!-- YAML
 added: v6.0.0
 changes:
+  - version: REPLACEME
+    pr-url: https://github.com/nodejs/node/pull/64917
+    description: UTF-16LE and UCS2 searches now inspect every byte offset.
+                 Buffer and Uint8Array values are compared in their entirety.
   - version: v26.1.0
     pr-url: https://github.com/nodejs/node/pull/62390
     description: Added the `end` parameter.
diff --git a/src/node_buffer.cc b/src/node_buffer.cc
index 19c28609660d..f1ef480a8ce4 100644
--- a/src/node_buffer.cc
+++ b/src/node_buffer.cc
@@ -1045,9 +1045,7 @@ void IndexOfString(const FunctionCallbackInfo<Value>& args) {
   int64_t end_i64 = args[5].As<Integer>()->Value();
 
   const char* haystack = buffer.data();
-  // Round down to the nearest multiple of 2 in case of UCS2.
-  const size_t haystack_length = (enc == UCS2) ?
-      buffer.length() &~ 1 : buffer.length();  // NOLINT(whitespace/operators)
+  const size_t haystack_length = buffer.length();
 
   size_t needle_length;
   if (!StringBytes::Size(isolate, needle, enc).To(&needle_length)) return;
@@ -1055,7 +1053,6 @@ void IndexOfString(const FunctionCallbackInfo<Value>& args) {
   // search_end is the exclusive upper bound of the search range.
   size_t search_end = static_cast<size_t>(std::min(
       std::max(end_i64, int64_t{0}), static_cast<int64_t>(haystack_length)));
-  if (enc == UCS2) search_end &= ~static_cast<size_t>(1);
 
   int64_t opt_offset = IndexOfOffset(haystack_length,
                                      offset_i64,
@@ -1102,27 +1099,27 @@ void IndexOfString(const FunctionCallbackInfo<Value>& args) {
     if constexpr (IsBigEndian()) {
       StringBytes::InlineDecoder decoder;
       if (decoder.Decode(env, needle, enc).IsNothing()) return;
-      const uint16_t* decoded_string =
-          reinterpret_cast<const uint16_t*>(decoder.out());
+      const uint8_t* decoded_string =
+          reinterpret_cast<const uint8_t*>(decoder.out());
 
       if (decoded_string == nullptr)
         return args.GetReturnValue().Set(-1);
 
-      result = nbytes::SearchString(reinterpret_cast<const uint16_t*>(haystack),
-                                    search_end / 2,
+      result = nbytes::SearchString(reinterpret_cast<const uint8_t*>(haystack),
+                                    search_end,
                                     decoded_string,
-                                    decoder.size() / 2,
-                                    offset / 2,
+                                    decoder.size(),
+                                    offset,
                                     is_forward);
     } else {
-      result = nbytes::SearchString(reinterpret_cast<const uint16_t*>(haystack),
-                                    search_end / 2,
-                                    needle_value.out(),
-                                    needle_value.length(),
-                                    offset / 2,
-                                    is_forward);
+      result = nbytes::SearchString(
+          reinterpret_cast<const uint8_t*>(haystack),
+          search_end,
+          reinterpret_cast<const uint8_t*>(needle_value.out()),
+          needle_length,
+          offset,
+          is_forward);
     }
-    result *= 2;
   } else if (enc == UTF8) {
     Utf8Value needle_value(isolate, needle);
     if (*needle_value == nullptr)
@@ -1163,8 +1160,6 @@ void IndexOfBuffer(const FunctionCallbackInfo<Value>& args) {
   CHECK(args[4]->IsBoolean());
   CHECK(args[5]->IsNumber());
 
-  enum encoding enc = static_cast<enum encoding>(args[3].As<Int32>()->Value());
-
   Environment* env = Environment::GetCurrent(args);
   THROW_AND_RETURN_UNLESS_BUFFER(env, args[0]);
   THROW_AND_RETURN_UNLESS_BUFFER(env, args[1]);
@@ -1182,7 +1177,6 @@ void IndexOfBuffer(const FunctionCallbackInfo<Value>& args) {
   // search_end is the exclusive upper bound of the search range.
   size_t search_end = static_cast<size_t>(std::min(
       std::max(end_i64, int64_t{0}), static_cast<int64_t>(haystack_length)));
-  if (enc == UCS2) search_end &= ~static_cast<size_t>(1);
 
   int64_t opt_offset = IndexOfOffset(haystack_length,
                                      offset_i64,
@@ -1218,27 +1212,13 @@ void IndexOfBuffer(const FunctionCallbackInfo<Value>& args) {
     return args.GetReturnValue().Set(-1);
   }
 
-  size_t result = search_end;
-
-  if (enc == UCS2) {
-    if (search_end < 2 || needle_length < 2) {
-      return args.GetReturnValue().Set(-1);
-    }
-    result = nbytes::SearchString(reinterpret_cast<const uint16_t*>(haystack),
-                                  search_end / 2,
-                                  reinterpret_cast<const uint16_t*>(needle),
-                                  needle_length / 2,
-                                  offset / 2,
-                                  is_forward);
-    result *= 2;
-  } else {
-    result = nbytes::SearchString(reinterpret_cast<const uint8_t*>(haystack),
-                                  search_end,
-                                  reinterpret_cast<const uint8_t*>(needle),
-                                  needle_length,
-                                  offset,
-                                  is_forward);
-  }
+  size_t result =
+      nbytes::SearchString(reinterpret_cast<const uint8_t*>(haystack),
+                           search_end,
+                           reinterpret_cast<const uint8_t*>(needle),
+                           needle_length,
+                           offset,
+                           is_forward);
 
   args.GetReturnValue().Set(result >= search_end ? -1
                                                  : static_cast<int>(result));
diff --git a/test/parallel/test-buffer-includes.js b/test/parallel/test-buffer-includes.js
index 3d6138faca81..2059e028ee50 100644
--- a/test/parallel/test-buffer-includes.js
+++ b/test/parallel/test-buffer-includes.js
@@ -74,7 +74,18 @@ assert(b.includes(Buffer.from('f'), 5));
 assert(b.includes(Buffer.from('f'), -1));
 assert(!b.includes(Buffer.from('f'), 6));
 
-assert(!Buffer.from('ff').includes(Buffer.from('f'), 1, 'ucs2'));
+assert(Buffer.from('ff').includes(Buffer.from('f'), 1, 'ucs2'));
+
+{
+  const oddIndexBuffer = Buffer.from('00aaaa', 'hex');
+  const oddIndexNeedle = Buffer.from('\uaaaa', 'utf16le');
+  assert(oddIndexBuffer.includes('\uaaaa', 0, 'utf16le'));
+  assert(!oddIndexBuffer.includes('\uaaaa', 0, 2, 'utf16le'));
+  assert(oddIndexBuffer.includes('\uaaaa', 0, 3, 'utf16le'));
+  assert(oddIndexBuffer.includes(oddIndexNeedle, 0, 'utf16le'));
+  assert(oddIndexBuffer.includes(
+    new Uint8Array(oddIndexNeedle), 0, 'utf16le'));
+}
 
 // test hex encoding
 assert.strictEqual(
diff --git a/test/parallel/test-buffer-indexof.js b/test/parallel/test-buffer-indexof.js
index 2dd62c91e25c..99d297ace847 100644
--- a/test/parallel/test-buffer-indexof.js
+++ b/test/parallel/test-buffer-indexof.js
@@ -81,7 +81,7 @@ assert.strictEqual(b.indexOf(Buffer.from('f'), 5), 5);
 assert.strictEqual(b.indexOf(Buffer.from('f'), -1), 5);
 assert.strictEqual(b.indexOf(Buffer.from('f'), 6), -1);
 
-assert.strictEqual(Buffer.from('ff').indexOf(Buffer.from('f'), 1, 'ucs2'), -1);
+assert.strictEqual(Buffer.from('ff').indexOf(Buffer.from('f'), 1, 'ucs2'), 1);
 
 // Test invalid and uppercase encoding
 assert.strictEqual(b.indexOf('b', 'utf8'), 1);
@@ -192,7 +192,7 @@ assert.strictEqual(Buffer.from('aaaa0').indexOf('30', 'hex'), 4);
 assert.strictEqual(Buffer.from('aaaa00a').indexOf('3030', 'hex'), 4);
 
 {
-  // Test usc2 and utf16le encoding
+  // Test ucs2 and utf16le encodings.
   ['ucs2', 'utf16le'].forEach((encoding) => {
     const twoByteString = Buffer.from(
       '\u039a\u0391\u03a3\u03a3\u0395', encoding);
@@ -308,6 +308,51 @@ assert.strictEqual(Buffer.from('aaaa').indexOf('你好', 'ucs2'), -1);
 // Haystack has odd length, but the needle is UCS2.
 assert.strictEqual(Buffer.from('aaaaa').indexOf('b', 'ucs2'), -1);
 
+{
+  // Search UTF-16LE values at byte offsets that are not aligned to the start
+  // of the Buffer.
+  const value = '\u6881\u6882\u6881';
+  const valueBuffer = Buffer.from(value, 'utf16le');
+  assert.strictEqual(valueBuffer.indexOf('\u6881', 1, 'utf16le'), 4);
+  assert.strictEqual(valueBuffer.slice(1).indexOf('\u6881', 'utf16le'), 3);
+  assert.strictEqual(valueBuffer.indexOf('\u6881', -5, 'utf16le'), 4);
+  assert.strictEqual(valueBuffer.indexOf('\u6881', 1, 4, 'utf16le'), -1);
+  assert.strictEqual(valueBuffer.indexOf('\u6881', 1, 6, 'utf16le'), 4);
+
+  const prefixed = Buffer.alloc(7);
+  prefixed.write(value, 1, 'utf16le');
+  assert.strictEqual(prefixed.indexOf(value, 1, 'utf16le'), 1);
+
+  // UTF-16LE searches compare bytes at every offset, even when a match spans
+  // two code units. The same needle also occurs on a code unit boundary at 4.
+  const crossUnitBuffer = Buffer.from([
+    0x41, 0x00, 0x01, 0x00, 0x00, 0x01,
+  ]);
+  assert.strictEqual(crossUnitBuffer.indexOf('\u0100', 0, 'utf16le'), 1);
+
+  const oddIndexBuffer = Buffer.from('00aaaa', 'hex');
+  const oddIndexNeedle = Buffer.from('\uaaaa', 'utf16le');
+  assert.strictEqual(oddIndexBuffer.indexOf('\uaaaa', 0, 'utf16le'), 1);
+  assert.strictEqual(oddIndexBuffer.indexOf('\uaaaa', 0, 2, 'utf16le'), -1);
+  assert.strictEqual(oddIndexBuffer.indexOf('\uaaaa', 0, 3, 'utf16le'), 1);
+  assert.strictEqual(oddIndexBuffer.indexOf(
+    oddIndexNeedle, 0, 'utf16le'), 1);
+  assert.strictEqual(oddIndexBuffer.indexOf(
+    new Uint8Array(oddIndexNeedle), 0, 'utf16le'), 1);
+  assert.strictEqual(oddIndexBuffer.lastIndexOf(
+    '\uaaaa', 2, 3, 'utf16le'), 1);
+  assert.strictEqual(oddIndexBuffer.lastIndexOf(
+    oddIndexNeedle, 2, 3, 'utf16le'), 1);
+
+  // The encoding argument only applies to string needles. Buffer needles are
+  // compared byte-for-byte, so an odd-length needle includes its final byte.
+  const oddLengthNeedle = Buffer.from([0x61, 0x00, 0xff]);
+  assert.strictEqual(Buffer.from([0x61, 0x00, 0x62, 0x00])
+    .indexOf(oddLengthNeedle, 0, 'utf16le'), -1);
+  assert.strictEqual(Buffer.from([0x61, 0x00, 0xff, 0x00])
+    .indexOf(oddLengthNeedle, 0, 'utf16le'), 0);
+}
+
 {
   // Find substrings in Utf8.
   const lengths = [1, 3, 15];  // Single char, simple and complex.
@@ -349,12 +394,18 @@ assert.strictEqual(Buffer.from('aaaaa').indexOf('b', 'ucs2'), -1);
 
       const patternBufferUcs2 =
           allCharsBufferUcs2.slice(index, index + length);
+      const expectedBufferIndex =
+          allCharsBufferUcs2.indexOf(patternBufferUcs2);
       assert.strictEqual(
-        index, allCharsBufferUcs2.indexOf(patternBufferUcs2, 0, 'ucs2'));
+        expectedBufferIndex,
+        allCharsBufferUcs2.indexOf(patternBufferUcs2, 0, 'ucs2'));
 
       const patternStringUcs2 = patternBufferUcs2.toString('ucs2');
+      const expectedStringIndex = allCharsBufferUcs2.indexOf(
+        Buffer.from(patternStringUcs2, 'ucs2'));
       assert.strictEqual(
-        index, allCharsBufferUcs2.indexOf(patternStringUcs2, 0, 'ucs2'));
+        expectedStringIndex,
+        allCharsBufferUcs2.indexOf(patternStringUcs2, 0, 'ucs2'));
     }
   }
 }
@@ -722,7 +773,11 @@ assert.strictEqual(reallyLong.lastIndexOf(pattern), 0);
   assert.strictEqual(buf.indexOf('', 0, 3), 0);
   assert.strictEqual(buf.indexOf('', 5, 3), 3);
   assert.strictEqual(buf.indexOf(Buffer.from(''), 5, 3), 3);
+  assert.strictEqual(buf.indexOf('', 5, 3, 'utf16le'), 3);
+  assert.strictEqual(buf.indexOf(Buffer.from(''), 5, 3, 'utf16le'), 3);
   assert.strictEqual(buf.indexOf('', 0, 0), 0);
   assert.strictEqual(buf.lastIndexOf('', 5, 3), 3);
   assert.strictEqual(buf.lastIndexOf(Buffer.from(''), 5, 3), 3);
+  assert.strictEqual(buf.lastIndexOf('', 5, 3, 'utf16le'), 3);
+  assert.strictEqual(buf.lastIndexOf(Buffer.from(''), 5, 3, 'utf16le'), 3);
 }