diff --git a/R/gtsave.R b/R/gtsave.R
index 5cba3c709..320631e70 100644
--- a/R/gtsave.R
+++ b/R/gtsave.R
@@ -485,14 +485,18 @@ gt_save_docx <- function(
     con = word_md_file
   )
 
+  temp_filename <- tempfile(fileext = paste0(".",tools::file_ext(filename)))
   rmarkdown::pandoc_convert(
     input = word_md_file,
-    output = filename
+    output = temp_filename,
   )
 
   if (needs_gt_as_word_post_processing(word_md_text)) {
-    gt_as_word_post_processing(path = filename)
+    gt_as_word_post_processing(path = temp_filename)
   }
+
+  file.rename(temp_filename, filename)
+
 }
 
 #' Get the lowercase extension from a filename
diff --git a/R/utils.R b/R/utils.R
index 02f681d7d..0dae313f9 100644
--- a/R/utils.R
+++ b/R/utils.R
@@ -1177,6 +1177,7 @@ markdown_to_xml <- function(text) {
     USE.NAMES = FALSE,
     FUN = function(x, ...) commonmark::markdown_xml(linebreak_br(x), ...)
   )
+
   vapply(
     res,
     FUN.VALUE = character(1L),
@@ -1225,9 +1226,21 @@ markdown_to_xml <- function(text) {
         }
       }
 
-      res <- lapply(children, apply_rules)
-      res <- vapply(res, FUN = as.character, FUN.VALUE = character(1L))
-      res <-  paste0(res, collapse = "")
+      if(length(children)){
+        res <- lapply(children, apply_rules)
+        res <- vapply(res, FUN = as.character, FUN.VALUE = character(1L))
+        res <-  paste0(res, collapse = "")
+      }else{
+        res <- xml_p(
+          xml_pPr(
+            xml_spacing(before = 0, after = 60)
+          ),
+          xml_r(
+            xml_rPr(),
+            xml_t()
+          )
+        )
+      }
       paste0("<md_container>", res, "</md_container>")
     }
   )
diff --git a/R/utils_render_xml.R b/R/utils_render_xml.R
index ca89da601..9b8a90e2c 100644
--- a/R/utils_render_xml.R
+++ b/R/utils_render_xml.R
@@ -2958,7 +2958,7 @@ parse_to_xml <- function(x, ...) {
   ##check if wrapped in ooxml
   ## get what it starts with and assign
 
-  if (is.null(x)) {
+  if (is.null(x) | identical(x,"<md_container></md_container>")) {
     x <-
       xml_p(
         xml_pPr(
diff --git a/tests/testthat/test-as_word.R b/tests/testthat/test-as_word.R
index 0df5b6dcd..0980964f3 100644
--- a/tests/testthat/test-as_word.R
+++ b/tests/testthat/test-as_word.R
@@ -2835,3 +2835,33 @@ test_that("multicolumn stub are supported", {
   expect_equal(xml_attr(xml_find_all(tcPr[[1]], ".//w:gridSpan"), "val"), "3")
 
 })
+
+test_that("word ooxml handles empty markdown cells gt object", {
+
+  # Create a one-row table for these tests
+  gt_empty_md <- exibble[1, 1:2] |>
+    dplyr::mutate(num = " ") |>
+    gt() |>
+    cols_label(
+      num = md(" "),
+      char = md("char")
+    ) |>
+    fmt_markdown()
+
+  ## convert to word xml nodeset
+  xml <- gt_empty_md %>% as_word() %>% xml2::read_xml()
+
+  table_cells <- xml_find_all(xml, ".//w:tc//w:t")
+
+  expect_equal(
+    as.character(table_cells),
+    c(
+      "<w:t xml:space=\"default\"/>",
+      "<w:t xml:space=\"preserve\">char</w:t>",
+      "<w:t xml:space=\"default\"/>",
+      "<w:t xml:space=\"preserve\">apricot</w:t>"
+    )
+  )
+
+})
+