Personalization is the group of ChatGPT settings that lets you influence how ChatGPT communicates with you, what preferences and context it can remember, and how those preferences can be reused in future conversations.

Instead of explaining the same requirements every time you start a chat, personalization can provide ChatGPT with persistent guidance.

For example, you may regularly want ChatGPT to:

• provide detailed technical explanations;
• use professional language;
• provide step-by-step troubleshooting;
• explain commands before suggesting them;
• format articles with headings and FAQs;
• use a particular writing style;
• remember recurring preferences;
• respond in a friendly or professional tone.

Personalization helps establish these preferences once so that ChatGPT can apply relevant context automatically.

OpenAI describes Custom Instructions as a way to provide information that ChatGPT should consider when generating responses. On Web and Desktop, these settings are managed through Settings → Personalization.

Personalization should therefore be understood as a combination of several mechanisms rather than one single feature.

Depending on the ChatGPT version, account, model and rollout, the Personalization screen may include options such as:

Custom Instructions → Memory → Personality → Characteristics and related personalization controls.

The exact options can change as ChatGPT is updated.

Understanding the Personalization Screen

The screenshot shows several important settings:

1. Custom instructions
2. Memory
3. Enable local memories
4. Allow local memory generation from tool-assisted chats
5. Delete local memories
6. Personality
7. A notice concerning model support for personality settings

Each serves a different purpose.

Understanding the difference is important because Custom Instructions, Memory and Personality are not the same thing.

A simple way to remember them is:

Custom Instructions = rules you deliberately provide

Memory = useful context ChatGPT can retain

Personality = how ChatGPT generally sounds

1. Custom Instructions

The first section shown in the screenshot is:

Custom instructions

with the description:

Give ChatGPT extra instructions and context for all chats on this host.

Custom Instructions allow you to specify information or behavioral preferences that ChatGPT should take into account when responding.

OpenAI states that Custom Instructions are available on Web, Desktop, iOS and Android and can be edited or deleted.

What Can You Put in Custom Instructions?

You can specify things such as:

• your profession;
• your technical skill level;
• preferred answer length;
• preferred language;
• formatting requirements;
• coding preferences;
• preferred operating system;
• preferred writing style;
• terminology preferences;
• recurring business requirements;
• troubleshooting approach;
• units, date formats or conventions;
• article formatting requirements.

For example, an IT administrator might use instructions such as:

I work in IT infrastructure and Windows administration. When answering technical questions, provide step-by-step instructions. Mention important risks before suggesting Registry, PowerShell, Group Policy or system-level changes. Prefer practical troubleshooting procedures and explain how to verify whether each step worked.

A developer might instead specify:

I primarily develop Windows and PHP-based web applications. Provide complete working examples when practical. Explain important dependencies, security considerations and deployment requirements.

A knowledgebase writer could use:

When creating technical knowledgebase articles, use descriptive headings, step-by-step procedures, troubleshooting sections, best practices and an FAQ. Explain technical terminology so that both administrators and normal users can understand the article.

These instructions can substantially reduce repetitive prompting.

Custom Instructions vs Normal Prompts

Suppose you frequently type:

Explain everything step by step and include examples.

Without Custom Instructions, you may have to repeat that preference in many conversations.

With Custom Instructions, you can make it part of your default preference.

However, Custom Instructions do not prevent you from overriding them in an individual conversation.

For example, your Custom Instructions might request detailed explanations, but you could still ask:

Give me only the command. No explanation.

The specific request in the current conversation can influence the response appropriately.

Good Uses for Custom Instructions

Custom Instructions work particularly well for stable requirements.

Examples include:

Role

I am an IT administrator managing Windows desktops and servers.

Technical level

Assume I understand networking, Windows administration and basic scripting.

Response structure

For troubleshooting questions, provide: Cause → Diagnosis → Solution → Verification.

Safety

Warn me before commands that can delete files, change permissions, modify the Registry or interrupt services.

Coding

When giving PHP examples, use secure coding practices and explain database-security considerations.

Documentation

Technical articles should include introduction, requirements, procedures, troubleshooting, best practices and FAQ.

What Should NOT Be Put in Custom Instructions?

Avoid entering unnecessary confidential information.

For example, do not use Custom Instructions as a password manager.

Avoid storing:

• passwords;
• OTPs;
• API secrets;
• private keys;
• credit/debit card details;
• banking credentials;
• recovery codes;
• highly confidential customer information;
• sensitive authentication information.

Custom Instructions are designed for context and preferences, not secret storage.

2. Memory

The second major area in the screenshot is Memory.

Memory enables ChatGPT to use relevant information from previous interactions so that future conversations can be more personalized.

OpenAI explains that when memory is enabled, ChatGPT can remember useful context from chats, files and connected apps where supported, reducing the need to repeatedly provide the same information.

Examples might include preferences such as:

I normally use Windows 11.

or:

When recommending scripts, I prefer PowerShell.

or:

I prefer technical documentation written for system administrators.

The important difference is that Custom Instructions are deliberately configured instructions, while Memory can develop from information shared during conversations. OpenAI specifically distinguishes the two this way.

3. Enable Local Memories

The screenshot shows:

Enable local memories

with the explanation:

Create memories from chats on this computer and use them to personalize future chats on this computer.

This wording is particularly important.

It indicates that this Desktop personalization environment can maintain local memories associated with that computer.

When enabled, useful information learned from conversations on the computer can be used to improve later conversations on that computer.

Example

Suppose you repeatedly work on Windows Server troubleshooting and regularly specify:

Always give me PowerShell commands suitable for Windows Server.

Relevant preferences can potentially contribute to future personalization when memory functionality is enabled.

Instead of starting every conversation completely from scratch, ChatGPT can make appropriate use of remembered context.

Benefits of Local Memories

Local memory can be particularly useful when one computer is primarily used for a particular purpose.

For example:

Office Computer

ChatGPT may frequently be used for:

• technical documentation;
• quotations;
• business correspondence;
• Windows troubleshooting;
• server administration.

Development Computer

It may primarily be used for:

• PHP development;
• Python development;
• database troubleshooting;
• API development;
• debugging.

Personal Computer

It might instead contain personalization related to:

• learning;
• hobbies;
• travel planning;
• personal projects.

Device-oriented memory can therefore help keep the working context relevant to how ChatGPT is used on that machine.

Should You Enable Local Memories?

Enable it when:

• the computer is primarily yours;
• you regularly use ChatGPT for similar work;
• you want ChatGPT to remember recurring preferences;
• you want less repetitive prompting.

Consider disabling it when:

• the computer is shared by several people;
• it is a public computer;
• several unrelated users use the same ChatGPT environment;
• you deliberately want conversations to remain independent;
• organizational policy restricts persistent personalization.

4. Allow Local Memory Generation from Tool-Assisted Chats

The next setting shown is:

Allow local memory generation from tool-assisted chats

The screenshot describes it as:

Generate memories from chats that used MCP tools or web search.

This is a more specialized setting.

ChatGPT conversations can sometimes involve external tools, web search, MCP-based tools or connected services.

When this option is enabled, conversations involving those tools may also contribute relevant information to local memory.

What Are Tool-Assisted Chats?

A normal conversation may simply involve:

You → ChatGPT → Response

A tool-assisted conversation can involve:

You → ChatGPT → Tool/Search → Information → ChatGPT → Response

Examples may include conversations where ChatGPT:

• searches the web;
• accesses an MCP tool;
• works with a connected service;
• retrieves external information;
• uses another supported tool to complete a task.

This setting determines whether relevant context from such conversations can participate in local memory generation.

Why Is This a Separate Setting?

Tool-assisted conversations can contain different types of context compared with ordinary chats.

For example, suppose you repeatedly ask ChatGPT to research:

Windows Server performance optimization.

During those conversations, ChatGPT may use web search.

If tool-assisted memory generation is allowed, relevant recurring preferences discovered through those interactions may be considered for personalization.

This gives the user more granular control over which types of conversations can contribute to memory.

When Should You Enable Tool-Assisted Memory?

It can be useful if you regularly use ChatGPT for:

• web research;
• technical research;
• MCP integrations;
• connected applications;
• product comparisons;
• documentation research;
• business workflows involving tools.

Consider disabling it if you prefer tool-assisted research sessions to remain separate from persistent local personalization.

Important: Tool-Assisted Memory Does Not Mean Saving the Entire Search

A common misunderstanding would be to assume:

“If I enable this option, every website and every search result becomes permanent memory.”

That is not the correct way to think about memory.

Memory is intended to retain useful personalization context, not function as a complete archive of every conversation or every webpage.

OpenAI also notes that memory should be regarded as a mechanism for useful preferences and context rather than a place to store large blocks of exact text.

5. Delete Local Memories

The next option is:

Delete local memories

with a Delete button.

Its purpose is straightforward:

Delete all memories stored locally on this computer.

This is useful when you want to reset the personalization accumulated on that machine.

When Should You Delete Local Memories?

Consider doing this when:

• another person will start using the computer;
• the computer is being sold or transferred;
• old memories are producing irrelevant personalization;
• your role or workflow has changed significantly;
• you want to start with a clean personalization environment;
• incorrect information has accumulated.

Turning Memory Off vs Deleting Memory

These concepts should not be confused.

Turning memory off means:

Stop using or generating memory according to that control.

Deleting memory means:

Remove memory that has already been stored under that memory system.

OpenAI similarly notes for its account-level memory features that simply disabling memory does not necessarily delete previously remembered information. Memory management and deletion are separate actions.

Therefore, whenever privacy or a complete reset is the objective, carefully review both the enable/disable controls and the available deletion controls.

6. Personality

The screenshot also shows:

Personality

with:

Friendly

selected.

Personality controls the general communication style ChatGPT uses.

OpenAI describes Personality as the style and tone used in responses. Changing personality affects how ChatGPT communicates but does not change the underlying capabilities or safety rules.

For example, depending on the options available in your version of ChatGPT, a personality may make responses feel more:

• friendly;
• professional;
• concise;
• conversational;
• direct;
• expressive.

The exact available choices may change as ChatGPT evolves.

What Personality Does NOT Change

Changing personality does not give the AI a different technical capability.

For example:

Friendly personality

does not mean ChatGPT becomes more technically capable.

Likewise, a concise or professional personality does not necessarily change:

• model intelligence;
• reasoning capability;
• tool availability;
• account permissions;
• security restrictions;
• safety requirements.

It primarily changes presentation and communication style.

Personality vs Custom Instructions

This distinction is important.

Suppose Personality is set to:

Friendly

but Custom Instructions say:

Use professional technical language and avoid unnecessary conversational wording.

ChatGPT attempts to use these personalization mechanisms together.

OpenAI states that personality works alongside memories and Custom Instructions.

A useful mental model is:

Personality = broad communication style

while:

Custom Instructions = specific requirements

For professional use, Custom Instructions are therefore often more important than simply choosing a personality.

7. “Personality Settings Are Not Supported by Every Model”

The screenshot contains an important notice:

Personality settings are not supported by every model. Codex's tone can be customized in Custom instructions.

This means you should not assume that every ChatGPT model or specialized environment will respond identically to the selected Personality.

Some specialized models or tools may prioritize task-specific behavior.

For example, a coding-oriented environment may focus heavily on:

• accurate code;
• debugging;
• concise technical output;
• development workflow.

Therefore, if you require a particular response style consistently, Custom Instructions are often the better place to state explicit requirements.

For example:

Explain code changes before providing modified code.

or:

When fixing code, provide the complete corrected file rather than isolated snippets.

These are concrete instructions rather than simply tone preferences.

8. Custom Instructions vs Memory vs Personality

The easiest comparison is:

Feature	Primary Purpose	Example
Custom Instructions	Explicit rules	“Always provide step-by-step instructions.”
Memory	Remember useful context	“I usually work with Windows Server.”
Local Memory	Computer-specific remembered context	Preferences learned on this computer
Tool-Assisted Memory	Allows tool/search conversations to contribute to local memory	Preferences arising during web research
Personality	Communication style	Friendly, professional, etc.
Delete Local Memories	Reset locally stored memories	Remove accumulated local personalization

9. How to Use Personalization Effectively

The best results come from combining these features rather than expecting one setting to do everything.

A practical structure is:

Layer 1 – Personality

Choose the broad communication style you generally prefer.

For example:

Friendly

may be appropriate when you want explanations to remain approachable.

Layer 2 – Custom Instructions

Define important permanent rules.

For example:

I am an IT professional. For technical troubleshooting, provide detailed step-by-step procedures. Clearly identify commands, explain what they do and provide verification steps. Warn before destructive operations.

Layer 3 – Memory

Allow ChatGPT to learn relevant recurring context.

For example:

• technologies you commonly use;
• preferred documentation style;
• recurring project context;
• stable workflow preferences.

Layer 4 – Current Prompt

Give task-specific requirements.

For example:

Create a troubleshooting guide for Windows Server 2025 RDP disconnections. Include Event Viewer locations, PowerShell diagnostic commands and an FAQ.

The current prompt should still contain requirements that apply only to that specific task.

10. A Good Custom Instruction Template for Technical Users

A technical professional could use something similar to the following:

I work with Windows desktops, Windows Server, networking, cloud services, PHP, databases and business software.

For troubleshooting questions, provide detailed step-by-step instructions.

Structure troubleshooting answers as:

1. Possible Cause
2. How to Diagnose
3. Solution
4. Verification
5. Rollback or Safety Notes

Explain PowerShell, CMD, Registry and Group Policy changes before recommending them.

Clearly warn me before commands that delete files, reset configurations, change permissions or may cause downtime.

For software development questions, provide complete and secure code examples where practical.

For technical articles, use clear headings, examples, troubleshooting information, best practices and FAQ.

Avoid unnecessary repetition and clearly distinguish verified facts from assumptions.

This is far more effective than entering vague instructions such as:

Give me good answers.

Specific instructions produce more predictable results.

11. Avoid Overloading Custom Instructions

More instructions are not always better.

An extremely long list can contain:

• duplicated rules;
• conflicting requirements;
• obsolete information;
• task-specific information that does not belong globally.

For example:

Always give very detailed answers.

and later:

Always answer in two sentences.

These instructions conflict.

Instead, define priorities clearly.

For example:

Default to detailed technical explanations unless I specifically request a short answer.

That rule is easier to apply consistently.

12. Periodically Review Your Personalization

Personalization should not necessarily be configured once and forgotten forever.

Review it periodically.

Check whether:

• your job role has changed;
• technologies you use have changed;
• old preferences are no longer relevant;
• unwanted memories exist;
• Custom Instructions contain outdated requirements;
• your preferred response style has changed.

Think of Personalization as a configuration profile for your AI assistant.

Like any configuration, it benefits from occasional maintenance.

13. Privacy Considerations

Personalization provides convenience, but users should understand what information they intentionally provide.

Do not put secrets into personalization simply because it saves typing.

For sensitive one-off conversations where you do not want memory to be used or created, ChatGPT also provides Temporary Chat functionality. OpenAI states that Temporary Chats do not create memories and are not saved in normal chat history; they are deleted from OpenAI systems after 30 days, subject to the stated safety handling.

Also remember that Memory settings and model-training controls are separate concepts.

For account-level controls governing whether conversations help improve OpenAI's models, OpenAI directs users to Settings → Data Controls → Improve the model for everyone.

Therefore:

Personalization/Memory controls what context can help personalize your experience.

Data Controls govern other data-use choices such as whether conversations can be used to improve models.

Do not assume that changing one automatically changes every other privacy setting.

14. Personalization for Shared Computers

Extra care should be taken on a shared PC.

Suppose three employees use the same Windows computer and the same ChatGPT environment.

Person A may primarily work in:

Accounting

Person B may work in:

Networking

Person C may work in:

Marketing

If personalization is accumulated without separation, preferences may become less useful.

For shared environments, consider:

• separate Windows accounts;
• separate ChatGPT accounts where appropriate;
• disabling local memory;
• periodically reviewing/deleting local memories;
• avoiding confidential information in Custom Instructions.

15. Personalization for IT Administrators

IT administrators can benefit substantially from well-designed Custom Instructions.

For example:

Assume the environment may contain production systems. Do not recommend restarting a server, disabling security software, modifying firewall policies, changing permissions or editing the Registry without explaining impact and rollback procedures.

This makes the output more operationally useful.

Another helpful rule is:

When troubleshooting, begin with non-destructive diagnostic steps before recommending configuration changes.

This encourages a workflow such as:

Observe → Diagnose → Verify → Change → Test → Rollback if required

rather than immediately modifying the system.

16. Personalization for Developers

Developers might specify:

When providing code:

• use secure coding practices;
• validate user input;
• explain dependencies;
• include error handling;
• mention security risks;
• avoid deprecated functions;
• provide complete files where practical;
• preserve existing functionality when modifying code.

This can significantly reduce repetitive instructions during development conversations.

17. Personalization for Knowledgebase Writers

For users who regularly create technical documentation, a useful instruction could be:

Write technical articles for both intermediate users and IT professionals. Include an introduction, explanation, prerequisites, step-by-step instructions, examples, troubleshooting, security considerations, best practices, FAQ and conclusion. Use descriptive SEO-friendly headings.

This creates much more consistent documentation.

18. Characteristics and Additional Personalization Controls

OpenAI is also rolling out Characteristics, which provide more granular control over how responses are presented.

Characteristics can influence aspects such as:

• brevity;
• tone;
• formatting;
• emoji usage;
• general response style.

OpenAI explains that Characteristics work alongside Personality, Custom Instructions and saved memories. Availability can vary because the feature is being rolled out gradually.

Therefore, your Personalization screen may not look exactly the same as another user's screen.

This is normal.

ChatGPT features can vary according to:

• application version;
• account;
• plan;
• platform;
• model;
• geographic availability;
• feature rollout.

19. Recommended Personalization Strategy

For most professional users, a sensible configuration is:

Personality:
Choose whichever style is comfortable for everyday communication.

Custom Instructions:
Add stable professional, technical and formatting requirements.

Memory:
Enable it when recurring context is useful and appropriate.

Tool-Assisted Memory:
Enable it if research and tool-based workflows form an important part of your regular work and you want relevant context from those interactions to contribute to personalization.

Local Memory:
Best suited to a trusted personal/work computer rather than an openly shared machine.

Delete Local Memories:
Use when resetting the computer's personalization or when accumulated memory is no longer appropriate.

20. Common Mistakes to Avoid

Mistake 1: Putting passwords in Custom Instructions

Never treat Personalization as credential storage.

Mistake 2: Using temporary project details as permanent instructions

Put project-specific requirements in the project conversation rather than globally whenever possible.

Mistake 3: Creating conflicting instructions

Keep Custom Instructions logically consistent.

Mistake 4: Expecting Personality to change intelligence

Personality primarily affects communication style.

Mistake 5: Assuming Memory remembers everything

Memory is designed to retain useful context, not provide a perfect verbatim archive of every conversation.

Mistake 6: Assuming deleting a chat always deletes associated memory

OpenAI explains that saved memories and chat history can be managed separately. Fully removing remembered information may require removing it from the relevant memory system as well as its original source.

Mistake 7: Confusing Memory with model-training settings

Memory and Improve the model for everyone are separate controls.

Frequently Asked Questions (FAQ)

1. What is Personalization in ChatGPT?

Personalization is a collection of settings that allows ChatGPT to adapt its responses using your instructions, remembered context and preferred communication style.

2. Where can I find Personalization in the ChatGPT Desktop App?

Open ChatGPT settings and select Personalization. The exact interface can change with application updates.

3. What are Custom Instructions?

Custom Instructions are explicit directions you provide about what ChatGPT should know or how you prefer it to respond.

4. Are Custom Instructions applied to every conversation?

They are designed to provide persistent guidance across chats where the feature is supported. OpenAI states that updates to Custom Instructions apply immediately.

5. What is ChatGPT Memory?

Memory allows ChatGPT to retain useful context that can help personalize future conversations.

6. Is Memory the same as Custom Instructions?

No. Custom Instructions are explicit instructions you configure, whereas Memory can retain relevant information from your interactions.

7. What does Enable Local Memories mean?

In the Desktop interface shown here, it means allowing memories to be created from chats on that computer and used to personalize future chats on that computer.

8. What is tool-assisted memory generation?

According to the setting shown in the Desktop App, it allows memory generation from conversations that involved supported tools such as MCP tools or web search.

9. Should I enable tool-assisted memory?

It can be useful if web research and tool-assisted workflows are part of your regular ChatGPT use. Disable it if you do not want those interactions contributing to local personalization.

10. What happens when I click Delete Local Memories?

According to the interface, it deletes memories stored locally on that computer.

11. Is deleting memory the same as deleting chats?

No. Memory and chat history can be separate systems. OpenAI notes that deleting a chat does not necessarily remove saved memory created from that conversation.

12. What is Personality?

Personality controls the general style and tone ChatGPT uses when communicating with you.

13. Does Friendly personality make ChatGPT less technical?

Not necessarily. Personality affects communication style rather than the model's fundamental capabilities.

14. Can Custom Instructions override Personality?

They work together, but explicit task instructions and context can strongly influence the resulting response. For precise requirements, state them explicitly in Custom Instructions or the current prompt.

15. Why does ChatGPT say Personality is not supported by every model?

Because some models or specialized environments may not use Personality settings in exactly the same way. The screenshot specifically notes this limitation and recommends Custom Instructions for customizing Codex's tone.

16. Can I ask ChatGPT what it remembers about me?

Yes. OpenAI specifically suggests asking ChatGPT what it remembers if you want to inspect remembered information.

17. Should I store passwords in ChatGPT Memory or Custom Instructions?

No. Passwords, private keys, OTPs and similar authentication secrets should be stored in an appropriate password or secrets-management system.

18. Is Memory suitable for exact templates?

Memory is better suited to high-level preferences and useful context than storing large blocks of exact text.

19. What should I use when I don't want a conversation creating memories?

Use Temporary Chat where available. OpenAI states that Temporary Chats do not create memories.

20. Does disabling Memory stop my chats from being used for model improvement?

These are separate controls. Model-improvement preferences are managed through Data Controls → Improve the model for everyone.

21. Can Personalization settings change over time?

Yes. ChatGPT features and interfaces evolve, and some personalization features are gradually rolled out.

22. Why might another user's Personalization screen look different?

Different accounts, platforms, plans, application versions, models and staged feature rollouts can expose different controls.

23. What are Characteristics in ChatGPT?

Characteristics are additional controls that can fine-tune response qualities such as brevity, tone, formatting and emoji usage. The feature is being rolled out gradually.

24. Can Personality, Memory and Characteristics work together?

Yes. OpenAI says these personalization mechanisms can work alongside one another.

25. What is the best way to use ChatGPT Personalization professionally?

Use Personality for broad communication style, Custom Instructions for explicit stable requirements, Memory for useful recurring context and the current prompt for requirements unique to the task.

Conclusion

The Personalization section of the ChatGPT Desktop App can transform ChatGPT from a generic assistant into a more consistently configured working environment.

The most important distinction is:

Custom Instructions tell ChatGPT what you explicitly want.

Memory helps ChatGPT retain useful context over time.

Local Memory applies remembered context associated with the computer in the Desktop environment shown.

Tool-Assisted Memory allows supported tool/search conversations to contribute to that local personalization.

Personality determines the general style and tone of responses.

When configured carefully, these features can reduce repetitive prompting, produce more consistent answers and make ChatGPT substantially more useful for technical support, software development, research, business work and documentation.

At the same time, personalization should be treated as a configuration system—not as secure credential storage. Keep instructions relevant, periodically review remembered information, avoid unnecessary confidential data and use Temporary Chat when you deliberately want a conversation not to create memories.

For the latest official information, consult OpenAI's Custom Instructions documentation and OpenAI's Memory FAQ because Personalization features continue to evolve.

#Tags

#ChatGPT #ChatGPTDesktop #ChatGPTDesktopApp #ChatGPTPersonalization #Personalization #CustomInstructions #ChatGPTCustomInstructions #ChatGPTMemory #ChatGPTLocalMemory #LocalMemory #ChatGPTPersonality #ChatGPTSettings #ChatGPTGuide #ChatGPTTutorial #OpenAI #ArtificialIntelligence #AITools #AIProductivity #ChatGPTTips #ChatGPTTricks #ChatGPT2026 #MemorySettings #PersonalizedAI #ChatGPTPrivacy #ChatGPTSecurity #ChatGPTForBusiness #ChatGPTForIT #ChatGPTForDevelopers #ChatGPTForWindows #WindowsAI #TechnicalGuide #Knowledgebase #TechnicalArticle #AITutorial #ChatGPTConfiguration #ChatGPTCustomization #ChatGPTMemorySettings #ToolAssistedChat #MCPTools #WebSearch #ChatGPTWebSearch #ChatGPTTools #ChatGPTFriendly #AIPersonalization #ChatGPTProductivity #CustomAI #ChatGPTFAQ #DesktopApp #AITechnology #ChatGPTBestPractices