ok

code/datums/chatmessage.dm

@@ -114,7 +114,7 @@
 	// BYOND Bug #2563917
 	// Construct text
 	var/static/regex/html_metachars = new(@"&[A-Za-z]{1,7};", "g")
-	var/complete_text = "<span class='center maptext [extra_classes.Join(" ")]' style='color: [tgt_color]'>[text]</span>"
+	var/complete_text = "<span class='center maptext [extra_classes.Join(" ")]' style='color: [tgt_color]'>[owner.say_emphasis(text)]</span>"
 	var/mheight = WXH_TO_HEIGHT(owned_by.MeasureText(replacetext(complete_text, html_metachars, "m"), null, CHAT_MESSAGE_WIDTH))
 	approx_lines = max(1, mheight / CHAT_MESSAGE_APPROX_LHEIGHT)
 

code/game/say.dm

@@ -59,7 +59,7 @@ GLOBAL_LIST_INIT(freqtospan, list(
 	var/endspanpart = "</span>"
 
 	//Message
-	var/messagepart = " <span class='message'>[lang_treat(speaker, message_language, raw_message, spans, message_mode)]</span></span>"
+	var/messagepart = " <span class='message'>[say_emphasis(lang_treat(speaker, message_language, raw_message, spans, message_mode))]</span></span>"
 
 	var/languageicon = ""
 	var/datum/language/D = GLOB.language_datum_instances[message_language]
@@ -96,6 +96,21 @@ GLOBAL_LIST_INIT(freqtospan, list(
 	return "[say_mod(input, message_mode)][spanned ? ", \"[spanned]\"" : ""]"
 	// Citadel edit [spanned ? ", \"[spanned]\"" : ""]"
 
+/atom/movable/proc/say_emphasis(input)
+	input = encode_html_emphasis(input, "|", "i")
+	input = encode_html_emphasis(input, "_", "u")
+	input = encode_html_emphasis(input, "+", "b")
+	return input
+
+/**
+  * Replaces a character inside message with html tags.  Note that html var must not include brackets.
+  * Will not create an open html tag if it would not have a closing one.
+  */
+/atom/movable/proc/encode_html_emphasis(input, char, html)
+	var/regex/search = regex("([REGEX_QUOTE(char)])(\\b.*\\b)([REGEX_QUOTE(char)])", "g")
+	return search.Replace_char(message, "<[html]>$2</[html]>")
+
+
 /// Quirky citadel proc for our custom sayverbs to strip the verb out. Snowflakey as hell, say rewrite 3.0 when?
 /atom/movable/proc/quoteless_say_quote(input, list/spans = list(speech_span), message_mode)
 	var/pos = findtext(input, "*")