OpenAI has released a practical guide on its Academy platform showing users how to use custom instructions and memory to make ChatGPT responses more consistent and personally relevant.
The guide addresses one of the most common friction points in day-to-day ChatGPT use: the need to re-explain personal context, professional background, or preferred response styles at the start of every new conversation. According to OpenAI, the two features — custom instructions and memory — work together to reduce that repetition and produce more tailored outputs over time.
What Custom Instructions Actually Do
Custom instructions allow users to set persistent preferences that apply across all ChatGPT conversations. A user can specify, for example, that they are a software engineer who prefers concise answers without preamble, or a teacher who wants explanations written for a general audience. These instructions are stored at the account level and loaded automatically with each new chat.
The feature is available to all ChatGPT users, including those on the Free tier, according to OpenAI. Setup requires navigating to account settings and entering free-text preferences in two fields: one for what ChatGPT should know about the user, and one for how the user wants ChatGPT to behave.
Custom instructions and memory together shift ChatGPT from a stateless tool into something closer to a persistent working relationship.
The practical implication for professionals is significant. A developer who always wants code in Python 3, commented, and without markdown formatting can encode that once rather than specifying it in every prompt. A marketing professional who wants all copy written in a particular brand voice can do the same.
How Memory Changes the Dynamic
Memory operates differently. Rather than user-defined rules, memory allows ChatGPT to learn and store facts from conversations automatically — or when explicitly told to remember something. Over time, the model builds a profile of user preferences, projects, and recurring needs.
Users retain control: memories can be viewed, edited, or deleted at any time through the settings panel. OpenAI states that memory is currently available to ChatGPT Free and Plus users in most regions, with broader rollout ongoing. It is not available in the EU and UK due to regulatory considerations, according to the company.
The distinction between the two features matters for workflow design. Custom instructions are best suited for stable, known preferences — things unlikely to change. Memory handles dynamic, evolving context: ongoing projects, recent decisions, or preferences the user may not have thought to specify upfront.
Integration Complexity and Limitations
For professionals evaluating these tools for team or enterprise use, there are important caveats. Memory is currently a personal account feature and does not sync across team members or shared workspaces. Organisations using ChatGPT Enterprise or ChatGPT Team plans have access to separate administrative controls, but memory as described in this guide applies to individual accounts.
Custom instructions, similarly, are per-account rather than per-device or per-session, which means they apply consistently but cannot be easily switched between contexts — a user who wants different behaviour for work versus personal use must manually update their instructions each time, or use different accounts.
OpenAI's Academy guide does not address API usage. Developers building on the OpenAI API who want persistent personalisation must implement their own memory layer — typically using a vector database or structured prompt injection — since the API itself remains stateless by default. Tools like the Assistants API offer thread-based context persistence, but that is a separate product with its own pricing structure.
Pricing and Availability at a Glance
Custom instructions are free and available to all registered ChatGPT users. Memory is available on Free and Plus tiers in supported regions; ChatGPT Plus costs $20 per month. Neither feature requires technical setup beyond navigating account settings, making adoption accessible to non-technical users.
For enterprise teams, ChatGPT Enterprise pricing is negotiated directly with OpenAI and includes additional administrative and data governance controls not available on consumer tiers.
What This Means
For everyday users and professionals alike, OpenAI's personalisation features reduce the overhead of working with ChatGPT repeatedly — but developers building on the API will need to engineer their own persistence solutions until OpenAI extends these capabilities to its programmatic interfaces.