OpenAI expands “custom instructions”, the user's details that ChatGPT should remember, to all users, after launching the ChatGPT Plus beta feature in July 2023
Context & Ripple Effects
Custom instructions began as an opt-in ChatGPT Plus beta, with availability limited outside the UK and EU in the earlier rollout. This expansion turns that subscriber beta for saved user preferences into a broadly available product behavior.
The move is an early step from one-time prompting toward persistent assistant context. It also precedes OpenAI's later test of default-on memory across conversations, which extends the same personalization arc.
First-order effects
- ChatGPT users can set details the assistant should retain when responding, reducing the need to restate preferences in new prompts.
- OpenAI removes a personalization capability that had been confined to the Plus beta, making it part of the wider ChatGPT experience.
Second-order effects
- Persistent preferences become a more visible baseline for conversational AI products, raising user expectations that an assistant will adapt across tasks rather than treat each prompt as isolated.
- The change shifts differentiation away from access to basic personalization and toward the breadth, reliability, and controls of the context an assistant can use.
Third-order effects
- If this progression continues, assistant competition will center increasingly on durable user context and the settings that govern it, not solely on model responses to individual prompts.
- More persistent context makes user control and transparency structurally important product requirements, because personalization depends on what an assistant retains and applies.
The trend: Consumer AI assistants are evolving from stateless chat interfaces into personalized operating layers that carry user preferences across interactions.