diff --git a/ydb/core/base/fulltext.cpp b/ydb/core/base/fulltext.cpp
index 9a877f3726ca..4c7fb3264c33 100644
--- a/ydb/core/base/fulltext.cpp
+++ b/ydb/core/base/fulltext.cpp
@@ -1,5 +1,6 @@
 #include "fulltext.h"
-#include <regex>
+#include <util/charset/utf8.h>
+#include <util/generic/xrange.h>
 
 namespace NKikimr::NFulltext {
 
@@ -45,35 +46,65 @@ namespace {
         return result;
     }
 
-    // Note: written by llm, can be optimized a lot later
-    TVector<TString> Tokenize(const TString& text, const Ydb::Table::FulltextIndexSettings::Tokenizer& tokenizer) {
-        TVector<TString> tokens;
-        switch (tokenizer) {
-            case Ydb::Table::FulltextIndexSettings::WHITESPACE: {
-                std::istringstream stream(text);
-                TString token;
-                while (stream >> token) {
-                    tokens.push_back(token);
+    inline bool IsNonStandard(wchar32 c) {
+        return !IsAlphabetic(c) && !IsDecdigit(c);
+    }
+
+    void Tokenize(const TString& text, TVector<TString>& tokens, auto isDelimiter) {
+        const unsigned char* ptr = (const unsigned char*)text.data();
+        const unsigned char* end = ptr + text.size();
+
+        while (ptr < end) {
+            wchar32 symbol;
+            size_t symbolBytes = 0;
+
+            while (ptr < end) { // skip delimiters
+                if (SafeReadUTF8Char(symbol, symbolBytes, ptr, end) != RECODE_OK) {
+                    tokens.clear();
+                    return;
+                }
+                if (!isDelimiter(symbol)) {
+                    break;
                 }
+                ptr += symbolBytes;
+            }
+            if (ptr >= end) {
                 break;
             }
-            case Ydb::Table::FulltextIndexSettings::STANDARD: {
-                std::regex word_regex(R"(\b\w+\b)"); // match alphanumeric words
-                std::sregex_iterator it(text.begin(), text.end(), word_regex);
-                std::sregex_iterator end;
-                while (it != end) {
-                    tokens.push_back(it->str());
-                    ++it;
+
+            const unsigned char* tokenPtr = ptr;
+            while (ptr < end) { // read token
+                if (SafeReadUTF8Char(symbol, symbolBytes, ptr, end) != RECODE_OK) {
+                    tokens.clear();
+                    return;
                 }
-                break;
+                if (isDelimiter(symbol)) {
+                    break;
+                }
+                ptr += symbolBytes;
             }
+            tokens.emplace_back((const char*)tokenPtr, ptr - tokenPtr);
+        }
+    }
+
+
+    TVector<TString> Tokenize(const TString& text, const Ydb::Table::FulltextIndexSettings::Tokenizer& tokenizer) {
+        TVector<TString> tokens;
+        switch (tokenizer) {
+            case Ydb::Table::FulltextIndexSettings::WHITESPACE:
+                Tokenize(text, tokens, IsWhitespace);
+                break;
+            case Ydb::Table::FulltextIndexSettings::STANDARD:
+                Tokenize(text, tokens, IsNonStandard);
+                break;
             case Ydb::Table::FulltextIndexSettings::KEYWORD:
-                tokens.push_back(text);
+                if (UTF8Detect(text) != NotUTF8) {
+                    tokens.push_back(text);
+                }
                 break;
             default:
                 Y_ENSURE(TStringBuilder() << "Invalid tokenizer: " << static_cast<int>(tokenizer));
         }
-
         return tokens;
     }
 
@@ -131,8 +162,8 @@ TVector<TString> Analyze(const TString& text, const Ydb::Table::FulltextIndexSet
     TVector<TString> tokens = Tokenize(text, settings.tokenizer());
 
     if (settings.use_filter_lowercase()) {
-        for (auto& token : tokens) {
-            token.to_lower();
+        for (auto i : xrange(tokens.size())) {
+            tokens[i] = ToLowerUTF8(tokens[i]);
         }
     }
 
diff --git a/ydb/core/base/ut/fulltext_ut.cpp b/ydb/core/base/ut/fulltext_ut.cpp
index e7b2fff76499..2450d74e33a9 100644
--- a/ydb/core/base/ut/fulltext_ut.cpp
+++ b/ydb/core/base/ut/fulltext_ut.cpp
@@ -1,6 +1,7 @@
 #include "fulltext.h"
 
 #include <library/cpp/testing/unittest/registar.h>
+#include <util/generic/xrange.h>
 
 namespace NKikimr::NFulltext {
 
@@ -100,20 +101,68 @@ Y_UNIT_TEST_SUITE(NFulltext) {
 
     Y_UNIT_TEST(Analyze) {
         Ydb::Table::FulltextIndexSettings::Analyzers analyzers;
-        TString text = "apple WaLLet  spaced-dog";
+        TString text = "apple WaLLet  spaced-dog_cat 0123,456@";
         
         analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::WHITESPACE);
-        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"apple", "WaLLet", "spaced-dog"}));
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"apple", "WaLLet", "spaced-dog_cat", "0123,456@"}));
 
         analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::STANDARD);
-        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"apple", "WaLLet", "spaced", "dog"}));
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"apple", "WaLLet", "spaced", "dog", "cat", "0123", "456"}));
 
         analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::KEYWORD);
         UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{text}));
 
         analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::WHITESPACE);
         analyzers.set_use_filter_lowercase(true);
-        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"apple", "wallet", "spaced-dog"}));
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"apple", "wallet", "spaced-dog_cat", "0123,456@"}));
+    }
+
+    Y_UNIT_TEST(AnalyzeRu) {
+        Ydb::Table::FulltextIndexSettings::Analyzers analyzers;
+        TString text = "Привет, это test123 и слово Ёлка   ёль!";
+        
+        analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::WHITESPACE);
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"Привет,", "это", "test123", "и", "слово", "Ёлка", "ёль!"}));
+
+        analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::STANDARD);
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"Привет", "это", "test123", "и", "слово", "Ёлка", "ёль"}));
+
+        analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::KEYWORD);
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{text}));
+
+        analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::STANDARD);
+        analyzers.set_use_filter_lowercase(true);
+        UNIT_ASSERT_VALUES_EQUAL(Analyze(text, analyzers), (TVector<TString>{"привет", "это", "test123", "и", "слово", "ёлка", "ёль"}));
+    }
+
+    Y_UNIT_TEST(AnalyzeInvalid) {
+        Ydb::Table::FulltextIndexSettings::Analyzers analyzers;
+
+        TVector<TString> texts = {
+            "\xC2\x41", // Invalid continuation byte
+            "\xC0\x81", // Overlong encoding
+            "\x80", // Lone continuation byte
+            "\xF4\x90\x80\x80", // Outside Unicode range
+            "\xE3\x81", // Truncated (incomplete)
+        };
+
+        for (auto i : xrange(texts.size())) {
+            TString testCase = TStringBuilder() << "case #" << i;
+            auto& text = texts[i];
+
+            analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::WHITESPACE);
+            UNIT_ASSERT_VALUES_EQUAL_C(Analyze(text, analyzers), (TVector<TString>{}), testCase);
+
+            analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::STANDARD);
+            UNIT_ASSERT_VALUES_EQUAL_C(Analyze(text, analyzers), (TVector<TString>{}), testCase);
+
+            analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::KEYWORD);
+            UNIT_ASSERT_VALUES_EQUAL_C(Analyze(text, analyzers), (TVector<TString>{}), testCase);
+
+            analyzers.set_tokenizer(Ydb::Table::FulltextIndexSettings::KEYWORD);
+            analyzers.set_use_filter_lowercase(true);
+            UNIT_ASSERT_VALUES_EQUAL_C(Analyze(text, analyzers), (TVector<TString>{}), testCase);
+        }
     }
 }