diff --git a/NEWS.md b/NEWS.md
index 6e8c0844a..b7b365b72 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -15,6 +15,7 @@
 ## Bug fixes
 
 * `Lint()`, and thus all linters, ensures that the returned object's `message` attribute is consistently a simple character string (and not, for example, an object of class `"glue"`; #2740, @MichaelChirico).
+* Files with encoding inferred from settings read more robustly under `lint(parse_settings = TRUE)` (#2803, @MichaelChirico).
 
 ## New and improved features
 
diff --git a/R/lint.R b/R/lint.R
index 00459c6ff..0f09ee524 100644
--- a/R/lint.R
+++ b/R/lint.R
@@ -45,6 +45,13 @@ lint <- function(filename, linters = NULL, ..., cache = FALSE, parse_settings =
 
   needs_tempfile <- missing(filename) || re_matches(filename, rex(newline))
   inline_data <- !is.null(text) || needs_tempfile
+  parse_settings <- !inline_data && isTRUE(parse_settings)
+
+  if (parse_settings) {
+    read_settings(filename)
+    on.exit(reset_settings(), add = TRUE)
+  }
+
   lines <- get_lines(filename, text)
 
   if (needs_tempfile) {
@@ -58,11 +65,6 @@ lint <- function(filename, linters = NULL, ..., cache = FALSE, parse_settings =
   filename <- normalize_path(filename, mustWork = !inline_data) # to ensure a unique file in cache
   source_expressions <- get_source_expressions(filename, lines)
 
-  if (isTRUE(parse_settings)) {
-    read_settings(filename)
-    on.exit(reset_settings(), add = TRUE)
-  }
-
   linters <- define_linters(linters)
   linters <- Map(validate_linter_object, linters, names(linters))
 
@@ -81,12 +83,7 @@ lint <- function(filename, linters = NULL, ..., cache = FALSE, parse_settings =
   lints <- list()
   if (!is_tainted(source_expressions$lines)) {
     for (expr in source_expressions$expressions) {
-      if (is_lint_level(expr, "expression")) {
-        necessary_linters <- expression_linter_names
-      } else {
-        necessary_linters <- file_linter_names
-      }
-      for (linter in necessary_linters) {
+      for (linter in necessary_linters(expr, expression_linter_names, file_linter_names)) {
         # use withCallingHandlers for friendlier failures on unexpected linter errors
         lints[[length(lints) + 1L]] <- withCallingHandlers(
           get_lints(expr, linter, linters[[linter]], lint_cache, source_expressions$lines),
@@ -709,3 +706,11 @@ zap_temp_filename <- function(res, needs_tempfile) {
   }
   res
 }
+
+necessary_linters <- function(expr, expression_linter_names, file_linter_names) {
+  if (is_lint_level(expr, "expression")) {
+    expression_linter_names
+  } else {
+    file_linter_names
+  }
+}
diff --git a/tests/testthat/test-get_source_expressions.R b/tests/testthat/test-get_source_expressions.R
index bb092d3dd..eeaff905b 100644
--- a/tests/testthat/test-get_source_expressions.R
+++ b/tests/testthat/test-get_source_expressions.R
@@ -105,17 +105,14 @@ test_that("Multi-byte character truncated by parser is ignored", {
 })
 
 test_that("Can read non UTF-8 file", {
-  withr::local_options(lintr.linter_file = tempfile())
-  file <- test_path("dummy_projects", "project", "cp1252.R")
-  lintr:::read_settings(file)
-  expect_null(get_source_expressions(file)$error)
+  proj_dir <- test_path("dummy_projects", "project")
+  withr::local_dir(proj_dir)
+  expect_no_lint(file = "cp1252.R", linters = list())
 })
 
-test_that("Warns if encoding is misspecified", {
-  file <- test_path("dummy_projects", "project", "cp1252.R")
-  lintr:::read_settings(NULL)
-  the_lint <- lint(filename = file, parse_settings = FALSE)[[1L]]
-  expect_s3_class(the_lint, "lint")
+test_that("Warns if encoding is misspecified, Pt. 1", {
+  proj_dir <- test_path("dummy_projects", "project")
+  withr::local_dir(proj_dir)
 
   lint_msg <- "Invalid multibyte character in parser. Is the encoding correct?"
   if (!isTRUE(l10n_info()[["UTF-8"]])) {
@@ -125,17 +122,21 @@ test_that("Warns if encoding is misspecified", {
     lint_msg <- "unexpected '<'"
   }
 
-  expect_identical(the_lint$linter, "error")
-  expect_identical(the_lint$message, lint_msg)
-  expect_identical(the_lint$line_number, 4L)
-
-  file <- test_path("dummy_projects", "project", "cp1252_parseable.R")
-  lintr:::read_settings(NULL)
-  the_lint <- lint(filename = file, parse_settings = FALSE)[[1L]]
-  expect_s3_class(the_lint, "lint")
-  expect_identical(the_lint$linter, "error")
-  expect_identical(the_lint$message, "Invalid multibyte string. Is the encoding correct?")
-  expect_identical(the_lint$line_number, 1L)
+  expect_lint(
+    file = "cp1252.R",
+    parse_settings = FALSE,
+    checks = list(rex::rex(lint_msg), linter = "error", line_number = 4L)
+  )
+
+  expect_lint(
+    file = "cp1252_parseable.R",
+    parse_settings = FALSE,
+    checks = list(
+      rex::rex("Invalid multibyte string. Is the encoding correct?"),
+      linter = "error",
+      line_number = 1L
+    )
+  )
 })
 
 test_that("Can extract line number from parser errors", {