diff --git a/NAMESPACE b/NAMESPACE
index 36d9b92..b6fb459 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -25,6 +25,7 @@ export(jsonpointer)
 export(version)
 importFrom(cli,cli_progress_bar)
 importFrom(cli,cli_progress_done)
+importFrom(tools,file_ext)
 importFrom(utils,head)
 importFrom(utils,tail)
 useDynLib(rjsoncons, .registration = TRUE)
diff --git a/R/j_data_type.R b/R/j_data_type.R
index e0d5832..f5f54df 100644
--- a/R/j_data_type.R
+++ b/R/j_data_type.R
@@ -40,6 +40,8 @@
 #'
 #' @inheritParams j_query
 #'
+#' @importFrom tools file_ext
+#'
 #' @details
 #'
 #' `j_data_type()` without any arguments reports possible return
@@ -49,10 +51,13 @@
 #'
 #' - For a scalar (length 1) character `data`, either `"url"`
 #'   (matching regular expression `"^https?://"`, `"file"`
-#'   (`file.exists(data)` returns `TRUE`), or `"json"`. When `"file"`
-#'   or `"url"` is inferred, the return value is a length 2 vector,
-#'   with the first element the inferred type of data (`"json"` or
-#'   `"ndjson"`) obtained from the first 2 lines of the file.
+#'   (`file.exists(data)` returns `TRUE`), or `"json"`. Compression
+#'   type is accounted for and supported compression extensions are
+#'   `".gz"`, `".bz2"`, and `".xz"`. `".zip"` is not supported and
+#'   will return an error. When `"file"` or `"url"` is inferred,
+#'   the return value is a length 2 vector, with the first element
+#'   the inferred type of data (`"json"` or `"ndjson"`) obtained
+#'   from the first 2 lines of the file.
 #' - For character data with `length(data) > 1`, `"ndjson"` if all
 #'   elements start a square bracket or curly brace, consistently
 #'   (i.e., agreeing with the start of the first record), otherwise
@@ -91,10 +96,24 @@ j_data_type <-
         )
 
         if (length(data) == 1L) {
-            ## url or file path or json
-            if (length(grep("^https?://", data))) {
+            is_url <- grepl("^https?://", data, TRUE)
+            is_file <- file.exists(data)
+            ## compressed and uncompressed url or file path
+            if (file_ext(data) %in% c("gz", "bz2", "xz")) {
+                format <-
+                    if (grepl("\\.ndjson\\.(gz|bz2|xz)$", data, TRUE))
+                        "ndjson"
+                    else if (grepl("\\.json\\.(gz|bz2|xz)$", data, TRUE))
+                        "json"
+                c(
+                    format,
+                    if (is_url) "url" else if (is_file) "file"
+                )
+            } else if (identical(file_ext(data), "zip")) {
+                stop("`j_data_type()` of type `.zip` is not supported")
+            } else if (is_url) {
                 c(j_data_type(readLines(data, 2L, warn = FALSE)), "url")
-            } else if (file.exists(data)) {
+            } else if (is_file) {
                 c(j_data_type(readLines(data, 2L, warn = FALSE)), "file")
             } else if (.is_scalar_character(data) && !inherits(data, "AsIs")) {
                 "json"
diff --git a/man/j_data_type.Rd b/man/j_data_type.Rd
index 6dd3b15..087ddaf 100644
--- a/man/j_data_type.Rd
+++ b/man/j_data_type.Rd
@@ -33,10 +33,13 @@ validate) the type of \code{data} based on the following rules:
 \itemize{
 \item For a scalar (length 1) character \code{data}, either \code{"url"}
 (matching regular expression \code{"^https?://"}, \code{"file"}
-(\code{file.exists(data)} returns \code{TRUE}), or \code{"json"}. When \code{"file"}
-or \code{"url"} is inferred, the return value is a length 2 vector,
-with the first element the inferred type of data (\code{"json"} or
-\code{"ndjson"}) obtained from the first 2 lines of the file.
+(\code{file.exists(data)} returns \code{TRUE}), or \code{"json"}. Compression
+type is accounted for and supported compression extensions are
+\code{".gz"}, \code{".bz2"}, and \code{".xz"}. \code{".zip"} is not supported and
+will return an error. When \code{"file"} or \code{"url"} is inferred,
+the return value is a length 2 vector, with the first element
+the inferred type of data (\code{"json"} or \code{"ndjson"}) obtained
+from the first 2 lines of the file.
 \item For character data with \code{length(data) > 1}, \code{"ndjson"} if all
 elements start a square bracket or curly brace, consistently
 (i.e., agreeing with the start of the first record), otherwise