diff --git a/.github/workflows/R-CMD-check.yaml b/.github/workflows/R-CMD-check.yaml
index 0d4c965db..330a4e9d9 100644
--- a/.github/workflows/R-CMD-check.yaml
+++ b/.github/workflows/R-CMD-check.yaml
@@ -66,6 +66,13 @@ jobs:
       actions: read
 
     steps:
+      - name: Start HTML validation server
+        run: |
+          docker run --rm -p 8888:8888 -d ghcr.io/validator/validator:latest &&
+          echo "W3C_MARKUP_VALIDATOR_BASEURL=http://0.0.0.0:8888" >> "$GITHUB_ENV"
+        if: runner.os == 'Linux'
+        shell: bash
+
       - uses: actions/checkout@v4
 
       - uses: r-lib/actions/setup-r@v2
diff --git a/DESCRIPTION b/DESCRIPTION
index fa3537624..325621513 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -73,6 +73,7 @@ Suggests:
     downlit (>= 0.4.0),
     htmlwidgets,
     jsonlite,
+    curl,
     rstudioapi,
     miniUI,
     rsconnect (>= 0.4.3),
diff --git a/R/html.R b/R/html.R
index dad883f82..6959364cd 100644
--- a/R/html.R
+++ b/R/html.R
@@ -279,7 +279,7 @@ split_chapters = function(
   # Need to take care of the div tags here before restore_part_html and
   # restore_appendix_html erase the section ids of the hidden PART or APPENDIX
   # sections.
-  if (split_level > 1) {
+  if (FALSE) {
     body = x[(i5 + 1):(i6 - 1)]
     h1 = grep('^<div (id="[^"]+" )?class="section level1("| )', body) + i5
     h2 = grep('^<div (id="[^"]+" )?class="section level2("| )', body) + i5
diff --git a/tests/rmd/split-section.Rmd b/tests/rmd/split-section.Rmd
index 4b6b4f167..34e342524 100644
--- a/tests/rmd/split-section.Rmd
+++ b/tests/rmd/split-section.Rmd
@@ -18,10 +18,16 @@ See chapter 2 now at \@ref(section-2)
 
 # Section 2
 
-## subsection 2 {#sub2}
+## subsection 21 {#sub2}
 
 ```{r iris-plot, fig.cap = "A plot"}
 plot(iris)
 ```
 
 See figure \@ref(fig:iris-plot)
+
+# subsection 22
+
+# Section 3
+
+## subsection 3
diff --git a/tests/test-rmd.R b/tests/test-rmd.R
index c77537cc2..70f93fe07 100644
--- a/tests/test-rmd.R
+++ b/tests/test-rmd.R
@@ -1,3 +1,4 @@
+source('./testthat/helper-validate_html.R')
 # only run this when NOT_CRAN is true (e.g., on Travis CI)
 if (Sys.getenv('NOT_CRAN') == 'true') local({
   all_files = function() {
@@ -9,10 +10,23 @@ if (Sys.getenv('NOT_CRAN') == 'true') local({
   for (f in list.files('rmd', '[.]Rmd$', full.names = TRUE)) {
     rmarkdown::render(f, envir = globalenv(), quiet = TRUE)
   }
+ 
+  html_issues = simplify_html_validation(
+    validate_html(list.files("rmd", ".html$", full.names = TRUE))
+  )
+  if(nrow(html_issues) > 0)
+    stop(nrow(html_issues),
+         " HTML issues detected:\n  ",
+         paste(html_issues$file, html_issues$messages, sep = ': ', collapse = '\n  '))
 
   # split by section works correctly
   ## id is used for html file name
-  sections_files = c("section-1.html", "subsection-1.html", "section-2.html", "sub2.html")
+  sections_files = c(
+    "section-1.html", "subsection-1.html",
+    "section-2.html", "sub2.html", "subsection-22.html",
+    "section-3.html", "subsection-3.html"
+  )
+
   if (any(!file.exists(file.path("rmd", sections_files))))
     stop("Failed to generate sections files")
   ## reference is working correctly (see #787)
diff --git a/tests/testthat/helper-validate_html.R b/tests/testthat/helper-validate_html.R
new file mode 100644
index 000000000..9e5448169
--- /dev/null
+++ b/tests/testthat/helper-validate_html.R
@@ -0,0 +1,39 @@
+validate_html <- function(
+  files,
+  # using same env variable old package to keep some compatibility
+  base_url = Sys.getenv('W3C_MARKUP_VALIDATOR_BASEURL')
+) {
+  if(base_url == '' || length(files) == 0) return(NULL)
+  lapply(files, function(file) {
+    h <- curl::new_handle()
+    curl::handle_setform(handle = h, out = 'json', file = curl::form_file(file, "text/html"))
+    jsonlite::fromJSON(rawToChar(
+      curl::curl_fetch_memory(base_url, h)$content
+    ))
+  })
+}
+
+simplify_html_validation <- function(results) {
+  if(is.null(results) || length(results) == 0) return(
+    data.frame(file=character(0), messages=character(0))
+  )
+  expected_errors <- c(
+    "Attribute “number” not allowed on element “div” at this point.",
+    "CSS: “border-top”: “solid\\9” is not a “color” value.",
+    "CSS: “border-bottom”: “solid\\9” is not a “color” value."
+
+  )
+  do.call(rbind, lapply(results,
+    function(result) {
+      if(is.null(result)) return(
+        data.frame(file=character(0), messages=character(0))
+      )
+      messages <- result$messages$message[result$messages$type == 'error']
+      messages <- messages[!messages %in% expected_errors]
+      data.frame(
+        file = if (length(messages) > 0) result$url else character(0),
+        messages = messages
+      )
+    }
+  ))
+}
\ No newline at end of file