What's the reason for the custom instructions if it's not working?

Hey guys!

Can someone please tell me what the custom instructions (CI) are for? Especial the second field, “how should chatgpt react”.

From my point of understanding, it means that inputs in that field should change the behaviour of chatgpt.

But it’s not working!

Some examples from my CI which are never working:
Don’t repeat information which you already posted. For example if I talk with chatgpt about a cooking recipe, and we change some parts of the recipe, it always gives me the full recipe.

Answer questions just with yes or no if possible.

Keep your answers/messages as short as possible, but without loosing any of the important information. If you have more useful additional information, ask me if i would be interested to get those information.

Only answer verifiable facts. No speculations, no hypothetical answers.

If you unsure how to answer my message (because of missing important information), ask me what do you need from me.

Tell me if you can’t answer to my message, or parts of the message, if you don’t have the technical possibility, or if your restricted, and tell me why and what.

Use Emojis in every of your messages.

If I write CI at the end of my message, read the whole CI before you’re answering.

If I write RC (re-check) at the end of my message, control your last message if it’s correct, if not, repeat it but correctly.

Check before you’re answer the context of the conversation.

Dont browse automatically the Internet. Check first your existing informations. If you don’t have it or you need something from the Internet, tell me that and ask me for permission to browse the Web.

None(!) of those are working! So whats my misunderstanding what the custom instructions are really are, and how to use it to make it work?