diff --git a/R-package/R/lgb.cv.R b/R-package/R/lgb.cv.R
index f81026fe673f..96c55573e205 100644
--- a/R-package/R/lgb.cv.R
+++ b/R-package/R/lgb.cv.R
@@ -46,6 +46,9 @@ CVBooster <- R6::R6Class(
 #' @param eval_train_metric \code{boolean}, whether to add the cross validation results on the
 #'               training data. This parameter defaults to \code{FALSE}. Setting it to \code{TRUE}
 #'               will increase run time.
+#' @param print_metrics \code{boolean} when calculating evaluation metrics after boosting rounds, whether
+#'                      to print informational messages with these metrics (see parameters \code{eval_freq},
+#'                      \code{eval_train_metric}, and \code{eval}).
 #' @inheritSection lgb_shared_params Early Stopping
 #' @return a trained model \code{lgb.CVBooster}.
 #'
@@ -92,6 +95,7 @@ lgb.cv <- function(params = list()
                    , reset_data = FALSE
                    , serializable = TRUE
                    , eval_train_metric = FALSE
+                   , print_metrics = TRUE
                    ) {
 
   if (nrounds <= 0L) {
@@ -244,7 +248,8 @@ lgb.cv <- function(params = list()
   }
 
   # Add printing log callback
-  if (params[["verbosity"]] > 0L && eval_freq > 0L) {
+  print_metrics <- as.logical(print_metrics)
+  if (print_metrics && eval_freq > 0L) {
     callbacks <- add.cb(cb_list = callbacks, cb = cb_print_evaluation(period = eval_freq))
   }
 
@@ -287,7 +292,7 @@ lgb.cv <- function(params = list()
       , cb = cb_early_stop(
         stopping_rounds = early_stopping_rounds
         , first_metric_only = isTRUE(params[["first_metric_only"]])
-        , verbose = params[["verbosity"]] > 0L
+        , verbose = print_metrics
       )
     )
   }
diff --git a/R-package/man/lgb.cv.Rd b/R-package/man/lgb.cv.Rd
index 555cb11c7bb3..01647394cd28 100644
--- a/R-package/man/lgb.cv.Rd
+++ b/R-package/man/lgb.cv.Rd
@@ -26,7 +26,8 @@ lgb.cv(
   callbacks = list(),
   reset_data = FALSE,
   serializable = TRUE,
-  eval_train_metric = FALSE
+  eval_train_metric = FALSE,
+  print_metrics = TRUE
 )
 }
 \arguments{
@@ -126,6 +127,10 @@ into a predictor model which frees up memory and the original datasets}
 \item{eval_train_metric}{\code{boolean}, whether to add the cross validation results on the
 training data. This parameter defaults to \code{FALSE}. Setting it to \code{TRUE}
 will increase run time.}
+
+\item{print_metrics}{\code{boolean} when calculating evaluation metrics after boosting rounds, whether
+to print informational messages with these metrics (see parameters \code{eval_freq},
+\code{eval_train_metric}, and \code{eval}).}
 }
 \value{
 a trained model \code{lgb.CVBooster}.
diff --git a/R-package/tests/testthat/test_basic.R b/R-package/tests/testthat/test_basic.R
index 9b84017476a7..c58ea9b53e51 100644
--- a/R-package/tests/testthat/test_basic.R
+++ b/R-package/tests/testthat/test_basic.R
@@ -3256,7 +3256,7 @@ test_that("lightgbm() accepts 'weight' and 'weights'", {
   expect_equal(record_evals[["valid"]][["auc"]][["eval_err"]], list())
 }
 
-.train_for_verbosity_test <- function(train_function, verbose_kwarg, verbose_param) {
+.train_for_verbosity_test <- function(train_function, verbose_kwarg, verbose_param, print_metrics = FALSE) {
   set.seed(708L)
   nrounds <- 5L
   params <- list(
@@ -3301,6 +3301,7 @@ test_that("lightgbm() accepts 'weight' and 'weights'", {
     )
     train_kwargs[["nfold"]] <- 3L
     train_kwargs[["showsd"]] <- FALSE
+    train_kwargs[["print_metrics"]] <- print_metrics
   }
   log_txt <- capture.output({
     bst <- do.call(
@@ -3549,6 +3550,7 @@ test_that("lgb.cv() only prints eval metrics when expected to", {
     out <- .train_for_verbosity_test(
       verbose_kwarg = verbose_keyword_arg
       , verbose_param = -1L
+      , print_metrics = FALSE
       , train_function = lgb.cv
     )
     .assert_has_expected_logs(
@@ -3583,6 +3585,7 @@ test_that("lgb.cv() only prints eval metrics when expected to", {
     out <- .train_for_verbosity_test(
       verbose_kwarg = verbose_keyword_arg
       , verbose_param = 1L
+      , print_metrics = TRUE
       , train_function = lgb.cv
     )
     .assert_has_expected_logs(
@@ -3638,6 +3641,7 @@ test_that("lgb.cv() only prints eval metrics when expected to", {
   out <- .train_for_verbosity_test(
     verbose_kwarg = verbose_keyword_arg
     , verbose_param = 1L
+    , print_metrics = TRUE
     , train_function = lgb.cv
   )
   .assert_has_expected_logs(
@@ -3828,3 +3832,29 @@ test_that("Evaluation metrics aren't printed as a single-element vector", {
   })
   expect_false(grepl("[1] \"[1]", log_txt, fixed = TRUE))
 })
+
+test_that("lgb.cv() doesn't mix booster messages with evaluation metrics messages", {
+  log_txt <- capture_output({
+    data(mtcars)
+    y <- mtcars$mpg
+    x <- as.matrix(mtcars[, -1L])
+    cv_result <- lgb.cv(
+        data = lgb.Dataset(x, label = y)
+        , params = list(
+            objective = "regression"
+            , metric = "l2"
+            , min_data_in_leaf = 5L
+            , max_depth = 3L
+            , num_threads = .LGB_MAX_THREADS
+        )
+        , nrounds = 2L
+        , nfold = 3L
+        , verbose = -1L
+        , eval_train_metric = TRUE
+        , print_metrics = TRUE
+    )
+  })
+  expect_false(grepl("Warning", log_txt, fixed = TRUE))
+  expect_false(grepl("gain", log_txt, fixed = TRUE))
+  expect_false(grepl("inf", log_txt, fixed = TRUE))
+})