diff --git a/NEWS.md b/NEWS.md
index 1a3f466..9b1db4a 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,7 @@
 # reuseme (development version)
 
+* `complete_todo()` no longer deletes the full line. It only deletes what it says it deletes (#27).
+
 * `file_outline()` works better with news files and headings at the end of files.
 
 * `file_outline()` gives a better error for empty paths.
diff --git a/R/outline.R b/R/outline.R
index 77ac4b0..e23ebb8 100644
--- a/R/outline.R
+++ b/R/outline.R
@@ -592,7 +592,8 @@ construct_outline_link <- function(.data, is_saved_doc, dir_common, pattern) {
     "- {.run [Done{cli::symbol$tick}?](reuseme::complete_todo(",
     # Removed ending dot. (possibly will fail with older versions)
     .data$line_id[cn], ", '", .data$file[cn], "', '",
-    stringr::str_sub(stringr::str_replace_all(.data$content[cn], "'|\\{|\\}|\\)|\\(|\\[\\]|\\+", "."), start = -15L), "'))}",
+    # modify regex twice if needed (see below)
+    stringr::str_sub(stringr::str_replace_all(.data$content[cn], "\\^|\\$|'|\\{|\\}|\\)|\\(|\\[\\]|\\+", "."), start = -15L), "'))}",
     .data$rs_version[cn]
   )
   # truncate other elements
@@ -610,7 +611,9 @@ construct_outline_link <- function(.data, is_saved_doc, dir_common, pattern) {
         outline_el,
         "- {.run [Done{cli::symbol$tick}?](reuseme::complete_todo(",
         # Removed ending dot. (possibly will fail with older versions)
-        line_id, ", '", file, "', '", stringr::str_sub(stringr::str_replace_all(content, "'|\\{|\\}|\\)|\\(|\\[\\]|\\+", "."), start = -15L), "'))}",
+
+        # modify regex twice if needed (see above)
+        line_id, ", '", file, "', '", stringr::str_sub(stringr::str_replace_all(content, "\\^|\\$|'|\\{|\\}|\\)|\\(|\\[\\]|\\+", "."), start = -15L), "'))}",
         rs_version
       ),
       outline_el2
diff --git a/R/use-todo.R b/R/use-todo.R
index d87a676..017dbe7 100644
--- a/R/use-todo.R
+++ b/R/use-todo.R
@@ -133,37 +133,44 @@ complete_todo <- function(line_id, file, regexp, rm_line = NULL) {
     ))
   }
 
+  line_content_before_comment <- stringr::str_extract(line_content, "([^#]+)#", group = 1)
   if (is.null(rm_line)) {
-    rm_line <- tag_type %in% c("TODO", "FIXME")
+    rm_line <- tag_type %in% c("TODO", "FIXME") # && is.na(line_content_before_comment)
   }
+  line_content_todo <- stringr::str_extract(line_content, "#[^#]+")
   line_content_show <- stringr::str_squish(line_content)
 
   if (rm_line) {
-    line_content_show <- cli::style_strikethrough(line_content_show)
+    if (is.na(line_content_before_comment)) {
+      line_content_show <- cli::style_strikethrough(line_content_show)
+    } else {
+      line_content_show <- paste(line_content_before_comment, cli::style_strikethrough(line_content_todo))
+    }
   } else {
     # Only strikethrough the tag
-    regex <- paste0("\\s", tag_type)
+    regex <- paste0(" ", tag_type)
     regex_new <- cli::style_strikethrough(regex)
     line_content_show <- stringr::str_replace(line_content_show, regex, regex_new)
   }
-
+  file_line <- paste0(file, ":", line_id)
   cli::cli_alert_success(
-    "Removed {.code {line_content_show}} from {.file {file}}!"
+    "Removed {.code {line_content_show}} from {.file {file_line}}!"
   )
+  # Rem
+  line_content_new <- strip_todo_line(line_content, only_rm_tag = !rm_line)
 
-  if (rm_line) {
-    file_content_new <- file_content[-line_id]
-    line_content_new <- ""
-  } else {
-    line_content_new <- sub(
-      pattern = paste0(tag_type, "\\s+"),
-      replacement = "",
-      line_content
-    )
-
+  if (nzchar(line_content_new)) {
     file_content[line_id] <- line_content_new
     file_content_new <- file_content
+  } else {
+    file_content_new <- file_content[-line_id]
+    if (!rm_line) {
+      # WIll remove this line eventually
+      # remove line if it ends up empty. not supposed to happen
+      cli::cli_abort("This should not happen. We were supposed not to remove lines", .internal = TRUE)
+    }
   }
+
   if (length(file_content_new) > 0) {
     usethis::write_over(path = file, lines = file_content_new, overwrite = TRUE, quiet = TRUE)
   } else {
@@ -236,3 +243,25 @@ compute_path_todo <- function(todo, proj) {
   list(path_todo = path_todo, todo = todo)
 }
 # this doesn't work
+
+# accepts a single line
+strip_todo_line <- function(x, only_rm_tag = FALSE) {
+  check_string(x)
+  if (!stringr::str_detect(x, "TODO|WORK|FIXME")) {
+    cli::cli_abort("Could not detect a todo tag in x")
+  }
+  if (only_rm_tag) {
+    x_new <- stringr::str_remove(x, "\\s(TODO|WORK|FIXME)")
+  } else {
+    x_new <- stringr::str_extract(x, "([^#]+)\\#+", group = 1)
+    if (is.na(x_new)) {
+      x_new <- ""
+      # cli::cli_abort("Could not extract content before tag")
+    }
+  }
+  if (x_new == x) {
+    cli::cli_abort("Could not make any change to x = {.val {x}}")
+  }
+  x_new <- stringr::str_trim(x_new, "right")
+  x_new
+}
diff --git a/tests/testthat/test-use-todo.R b/tests/testthat/test-use-todo.R
index e2af578..7024b2f 100644
--- a/tests/testthat/test-use-todo.R
+++ b/tests/testthat/test-use-todo.R
@@ -17,12 +17,23 @@ test_that("Marking TODO as done detects tags", {
   )
 })
 
+test_that("todo items are correctly stripped", {
+  expect_equal(
+    strip_todo_line("2^2 # TODO this"),
+    "2^2"
+  )
+  expect_equal(
+    strip_todo_line("2^2 # TODO this", only_rm_tag = TRUE),
+    "2^2 # this"
+  )
+})
+
 test_that("Marking a TODO item as done works", {
   tmp <- tempfile(fileext = "R")
   content <- c(
     "# I Want this done",
-    "# TODO item to delete",
-    "# WORK Explain what the next code does.",
+    "2^2 # TODO item to delete",
+    "2^2 # WORK Explain what the next code does.",
     "# TODO with {.href [cli hyperlinks](https://cli.r-lib.org/reference/links.html)}",
     "# FIXME Repair this function",
     "   # TODO Check r-lib/usethis#1890",
@@ -48,7 +59,7 @@ test_that("Marking a TODO item as done works", {
 
   expect_complete_todo(
     out <- complete_todo(
-      line_id = 3,
+      line_id = 4,
       file = tmp,
       regexp = "Explain what the next code does"
     ),
@@ -56,7 +67,7 @@ test_that("Marking a TODO item as done works", {
   )
   expect_equal(
     out,
-    "# Explain what the next code does."
+    "2^2 # Explain what the next code does."
   )
   expect_complete_todo(
     out <- complete_todo(
@@ -70,7 +81,8 @@ test_that("Marking a TODO item as done works", {
     read_utf8(tmp),
     c(
       "# I Want this done",
-      "# Explain what the next code does.",
+      "2^2",
+      "2^2 # Explain what the next code does.",
       "# TODO with {.href [cli hyperlinks](https://cli.r-lib.org/reference/links.html)}",
       "# FIXME Repair this function",
       "# TODO Check https://github.com/r-lib/usethis/issues/1890",