Introduction

The Tail Generation Pattern is a small utility pattern that can be used to re-iterate, add or confirm some notes at the end of the answer of the LLM. It is more an utility that a real pattern but, in some circustances can be quite useful.

The Prompt

The basic structure of the prompt, in literature, is the following statement:

At the end of your answer, repeat [STATEMENT] and/or ask me for [QUESTION].

Added at the end of the question. A possible example is pretty simple

From now at the end of your answer, repeat my question.

Or, alternately

From now at the end of your answer, ask me if I want to know more

How to use the prompt

The usage of the prompt is pretty simple. It has to be inserted as first request in a session, before any further ask

Some Tests

Testing this prompt is pretty simple, the main validations are:

  • The LLM acknoledges that it will repeat the statement
  • The LLM is correctly repeating the statement at the end

ChatGPT 5.2

ChatGPT gets the request

From now at the end of your answer, repeat my question.

and answers

Got it — I’ll repeat your question verbatim at the end of every response from now on. From now at the end of your answer, repeat my question.

but it entering in an interesting loop, as it repeating the question (From now at the end of your answer, repeat my question) on the answer of the request itself.

Claude - Haiku 4.5

Claude is smarter, as does not repeat the question itself on the first interaction

From now at the end of your answer, repeat my question.

I understand! From now on, I’ll repeat your question at the end of my responses. Is there anything I can help you with?

Mistral

Mistral too has a more user friendly approach in the answer.

From now at the end of your answer, repeat my question.

Understood! I’ll repeat your question at the end of each of my answers. What would you like to know or discuss today?

Conclusions

The Tail Generation Pattern is a simple way to add some extra message in the answer of the LLM, used then to manage the real request. Easy to implement and quite useful in interactive context