diff --git a/arch/x86/core/fatal.c b/arch/x86/core/fatal.c
index 07bd2315b1579..9e38c6a335b82 100644
--- a/arch/x86/core/fatal.c
+++ b/arch/x86/core/fatal.c
@@ -13,6 +13,10 @@
 #include <mmu.h>
 LOG_MODULE_DECLARE(os, CONFIG_KERNEL_LOG_LEVEL);
 
+#ifdef CONFIG_DEMAND_PAGING
+#include <zephyr/kernel/mm/demand_paging.h>
+#endif
+
 #if defined(CONFIG_BOARD_QEMU_X86) || defined(CONFIG_BOARD_QEMU_X86_64)
 FUNC_NORETURN void arch_system_halt(unsigned int reason)
 {
@@ -476,6 +480,20 @@ void z_x86_page_fault_handler(struct arch_esf *esf)
 #endif /* CONFIG_X86_KPTI */
 		if (was_valid_access) {
 			/* Page fault handled, re-try */
+
+#ifdef CONFIG_EVICTION_LRU
+			/* Currently only LRU eviction algorithm needs to be marked.
+			 * So for now, skip it for others to avoid unnecessary
+			 * processing.
+			 */
+			uintptr_t phys, ret;
+
+			ret = arch_page_info_get(virt, &phys, false);
+			if ((ret & ARCH_DATA_PAGE_NOT_MAPPED) != ARCH_DATA_PAGE_NOT_MAPPED) {
+				k_mem_paging_eviction_accessed(phys);
+			}
+#endif
+
 			return;
 		}
 	}
diff --git a/subsys/demand_paging/eviction/Kconfig b/subsys/demand_paging/eviction/Kconfig
index 2c582be201a36..7cd2983cba7a0 100644
--- a/subsys/demand_paging/eviction/Kconfig
+++ b/subsys/demand_paging/eviction/Kconfig
@@ -48,3 +48,16 @@ config EVICTION_NRU_PERIOD
 	  pages that are capable of being paged out. At eviction time, if a page
 	  still has the accessed property, it will be considered as recently used.
 endif # EVICTION_NRU
+
+if EVICTION_LRU
+
+config EVICTION_LRU_SW_SIMULATED
+	bool
+	default y if X86
+	help
+	  This is for architectures where hardware does not provide a native
+	  mean to figure out least recently used pages. Enabling this changes
+	  the eviction selection to clear the access flag if it is set and
+	  requeues the page frame for later consideration.
+
+endif # EVICTION_LRU
diff --git a/subsys/demand_paging/eviction/lru.c b/subsys/demand_paging/eviction/lru.c
index 1156fa1a549ad..da4bab9eee520 100644
--- a/subsys/demand_paging/eviction/lru.c
+++ b/subsys/demand_paging/eviction/lru.c
@@ -160,14 +160,48 @@ void k_mem_paging_eviction_accessed(uintptr_t phys)
 
 struct k_mem_page_frame *k_mem_paging_eviction_select(bool *dirty_ptr)
 {
-	uint32_t head_pf_idx = LRU_PF_HEAD;
+	uint32_t pf_idx = LRU_PF_HEAD;
+	struct k_mem_page_frame *pf;
+	uintptr_t flags;
 
-	if (head_pf_idx == 0) {
+	if (pf_idx == 0) {
 		return NULL;
 	}
 
-	struct k_mem_page_frame *pf = idx_to_pf(head_pf_idx);
-	uintptr_t flags = arch_page_info_get(k_mem_page_frame_to_virt(pf), NULL, false);
+#ifndef CONFIG_EVICTION_LRU_SW_SIMULATED
+	pf = idx_to_pf(pf_idx);
+	flags = arch_page_info_get(k_mem_page_frame_to_virt(pf), NULL, false);
+#else
+	do {
+		/*
+		 * During page selection for eviction, if a page has access bit set,
+		 * clear the access bit and put it back at the end of queue.
+		 *
+		 * A subtle fail safe to avoid infinite loop is that the head of queue
+		 * when this function is entered will have its access bit cleared.
+		 * So when we have exhausted every page and looped back to that page,
+		 * it will be selected for eviction.
+		 */
+		pf = idx_to_pf(pf_idx);
+		flags = arch_page_info_get(k_mem_page_frame_to_virt(pf), NULL, true);
+
+		if ((flags & ARCH_DATA_PAGE_ACCESSED) != ARCH_DATA_PAGE_ACCESSED) {
+			/* Page has not been accessed recently, so select this. */
+			break;
+		}
+
+		/*
+		 * Page has been accessed recently.
+		 * Put the page frame back to the end of queue.
+		 * Use unlink() here since we do not need the extra logic inside remove().
+		 */
+		lru_pf_unlink(pf_idx);
+		lru_pf_append(pf_idx);
+
+		/* Grab the new head of queue. */
+		pf_idx = LRU_PF_HEAD;
+	} while (pf_idx != 0);
+#endif /* !CONFIG_EVICTION_LRU_SW_SIMULATED */
 
 	__ASSERT(k_mem_page_frame_is_evictable(pf), "");
 	*dirty_ptr = ((flags & ARCH_DATA_PAGE_DIRTY) != 0);