diff --git a/modules/caddyhttp/matchers.go b/modules/caddyhttp/matchers.go
index 2bdd874b7..e10c675fb 100644
--- a/modules/caddyhttp/matchers.go
+++ b/modules/caddyhttp/matchers.go
@@ -640,6 +640,20 @@ func (MatchPath) matchPatternWithEscapeSequence(escapedPath, matchPath string) b
 		iPattern++
 	}
 
+	// if the pattern was fully consumed before the path, the remaining path
+	// bytes still have to be part of the string we match against; otherwise a
+	// pattern with no trailing wildcard (e.g. "/foo%2fbar") would behave like a
+	// prefix match and incorrectly match longer paths (e.g. "/foo%2fbarbaz").
+	// the pattern has no more escape hints here, so compare in normalised space.
+	if iPath < len(escapedPath) {
+		remaining := escapedPath[iPath:]
+		if unescaped, err := url.PathUnescape(remaining); err == nil {
+			sb.WriteString(unescaped)
+		} else {
+			sb.WriteString(remaining)
+		}
+	}
+
 	// we can now treat rawpath globs (%*) as regular globs (*)
 	matchPath = strings.ReplaceAll(matchPath, "%*", "*")
 
diff --git a/modules/caddyhttp/matchers_test.go b/modules/caddyhttp/matchers_test.go
index b129aa4b7..b98743baf 100644
--- a/modules/caddyhttp/matchers_test.go
+++ b/modules/caddyhttp/matchers_test.go
@@ -422,6 +422,21 @@ func TestPathMatcher(t *testing.T) {
 			input:  "/ADMIN%2fPaAzZLm123NEL",
 			expect: true,
 		},
+		{
+			match:  MatchPath{"/foo%2fbar"},
+			input:  "/foo%2fbarbaz",
+			expect: false,
+		},
+		{
+			match:  MatchPath{"/foo%2f"},
+			input:  "/foo%2fx",
+			expect: false,
+		},
+		{
+			match:  MatchPath{"/admin%2fpanel"},
+			input:  "/admin%2fpanelX",
+			expect: false,
+		},
 	} {
 		err := tc.match.Provision(caddy.Context{})
 		if err == nil && tc.provisionErr {