The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →The reliable way to limit chatbot personalization is to treat it as several separate jobs: disable or clear the memory source, remove old chats or saved information, disconnect any app that supplied the data, and use an unpersonalized temporary chat for a sensitive one-off question. A model-training switch, chat-history switch, and memory switch usually affect different systems, so changing one does not automatically erase the others.
This guide covers the documented controls for ChatGPT, Gemini, and Microsoft 365 Copilot. Menus, eligibility and administrator policies can change by plan, region, platform and date; the labels below reflect provider guidance available in October 2026.
As an Amazon Associate I earn from qualifying purchases.
If a recommendation is unexpectedly specific, first ask what could have supplied the detail: a saved memory, an earlier conversation, a profile field, a connected app, location, workplace communications or a usage signal. Then change the control tied to that source. A single “privacy” toggle rarely covers every path.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsFirst separate personalization, retention and model improvement
When a chatbot gives a recommendation that seems unusually specific, several data paths may be involved. A service might use a saved preference such as dietary restrictions, references to earlier chats, profile or account details, connected-app content, location, workplace communications, or usage signals. Separately, it may retain the conversation in history, and separately again it may use eligible conversations to improve its models.
#1 Best Overall
These are different purposes and require different actions:
| Goal | Control to look for | What it does not automatically do |
|---|---|---|
| Stop future answers using remembered preferences or earlier conversations | Memory, Personalization, Reference chat history or an enhanced-personalization setting | It does not necessarily delete the source chat or erase a memory already saved from it. |
| Remove an existing conversation or source record | Delete the chat, clear saved memories, remove activity, or submit an account data request | Deleting a chat may not remove a separately stored memory or data already reviewed for safety. |
| Stop eligible chats being used to improve models | Data Controls, Improve the model for everyone, Keep Activity or a comparable training option | It does not by itself disable memory, delete history or stop every safety and service process. |
| Prevent a connected service from supplying new information | Disconnect or revoke the connected app | It does not remove conversations that already contain the app’s data. |
| Ask one sensitive question without using existing personalization | Unpersonalized Temporary Chat (where offered) | It is not a promise of instant deletion or zero retention. |
Use the control that matches the data path you want to change. If you want a clean result, you may need more than one action.
Why the distinction matters
Suppose you told a chatbot about a food allergy six months ago. Turning off model improvement may prevent eligible chats from being used for training, but the allergy could still be in a saved-memory list or in the old conversation. Deleting that conversation may still leave a separately stored memory. Disconnecting a grocery app may stop future imports while leaving recommendations that already quoted its data in your history. Thinking in sources and purposes prevents a false sense of deletion.
How to stop ChatGPT using personal data for recommendations
1. Check which account and controls you have
ChatGPT’s available memory, chat-history and data controls vary by plan, region, platform and workspace. Open the account’s settings rather than relying on a menu path seen in an older screenshot. Personal accounts and managed workspaces can expose different options. If a control is missing, it may be unavailable for that account type rather than broken.
2. Turn off or clear memory and chat-history references
- Open Settings.
- Select Personalization, then Memory.
- Review the saved memories and any option that allows ChatGPT to reference chat history.
- Turn off the reference option that you do not want, and delete individual memories or clear the memory list when appropriate.
Disabling Reference chat history schedules information remembered from past chats for deletion within 30 days, according to OpenAI’s consumer guidance. The original conversations remain in your chat history unless you delete them separately. Turning off memory also does not retroactively remove details already present in a conversation.
Read the wording carefully: a control for “reference chat history” concerns what future responses can draw on, while a saved-memory list is a separate record. You can leave one enabled and remove selected entries from the other, or turn both off for a broader reduction in personalization.
3. Delete source chats and check for a separate saved memory
Delete old conversations that contain the preferences, health details, location, financial information or other facts you no longer want available as context. Then return to the memory panel and check it again: deleting a chat does not necessarily remove a saved memory derived from that chat. Clear both sources when your goal is to remove the detail from future personalization.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Deletion is also worth doing for practical reasons. A chat that remains in history can still be opened, searched or copied by someone who gains access to the account, even if it is no longer used to personalize a new answer. Remove the whole conversation when individual-message deletion is not offered, and review any exported or shared copy separately.
4. Disable model-improvement use separately
- Open Settings.
- Go to Data Controls.
- Turn off Improve the model for everyone (the exact label can vary).
This setting addresses eligible personal-account conversations being used to improve models. It does not delete chat history and does not turn off Personalization or Memory. Treat those as separate decisions. If your aim is simply to stop recommendations from reflecting old details, changing this setting alone is not the relevant fix.
5. Use an unpersonalized Temporary Chat for a one-off question
- Start a new Temporary Chat.
- Before sending the first message, choose the Unpersonalized option if the interface offers a choice between personalized and unpersonalized temporary chats.
- Ask the question without pasting identifying information that the answer does not require.
An unpersonalized temporary chat does not create or update memories. OpenAI says temporary chats are not used to improve models while they remain temporary, but a copy may be retained for up to 30 days for safety. If you save the conversation, it becomes a regular chat in your history and follows the account’s normal personalization and model-improvement settings. Saving therefore changes the privacy behavior you selected at the start.
Temporary Chat is most useful for a narrow, one-time question, not as a guarantee that the provider cannot process the prompt. The service still has to receive and analyze it to answer, and safety systems can inspect or retain a copy under the stated period.
Free tools Windows power users keep installed
One-click scans. No signup required.
6. Disconnect apps and review location sharing
If a recommendation used information from a connected app, disconnecting that app prevents future access but does not remove past conversations that already contain its content. Delete those conversations as a separate step. Location sharing for relevant local recommendations is optional where available; switch it off if you do not want location to influence results.
For example, revoking a calendar, shopping or travel connection can stop new data from being read, but it cannot rewrite a restaurant suggestion already generated from a previous import. Check both the chatbot’s connected-app page and the third-party account’s authorization page if you want to revoke access fully.
How to turn off Gemini memory and reduce activity use
Memory and Keep Activity are different controls
Gemini’s Memory controls memory-based personalization. Keep Activity controls storage of Gemini activity and its use for model improvement. Turning off one does not perform the other’s job. A person can turn Memory off while activity remains stored, or turn Keep Activity off while a previously saved personalization detail still needs to be removed.
1. Turn off Memory
- Open Gemini and select Settings and help.
- Open Personal Intelligence.
- Turn Memory off, or review and delete the individual remembered details offered there.
Google says memory-based personalization requires a personal Google Account, an account holder aged 18 or older, and Keep Activity turned on; availability can vary. A work or school account may not show the same control. If you do not see Personal Intelligence, confirm which account is active and whether an administrator manages it.
2. Delete chats that supplied unwanted details
Delete the Gemini conversations containing the preference or personal fact. Google notes that there may be a short delay before Gemini stops using deleted chats for personalization. If the information came through a connected app, delete the relevant Gemini chat and disconnect the app. Doing only one can leave the other source available.
Deleting activity is not identical to deleting a remembered item. After removing chats, revisit Personal Intelligence and confirm that the detail is no longer listed. Also check other Google services: Web & App Activity and Location History have their own controls and can continue saving information independently.
3. Change Keep Activity when model-improvement use is the concern
- Open Gemini’s activity controls, usually under Settings and Gemini Apps Activity.
- Change Keep Activity to the option that fits your preference, including turning it off where available.
- Review the auto-delete period; Google lists 18 months as the default, with options of 3 months, 36 months or no automatic deletion.
With Keep Activity off, future chats stop appearing in activity and are not used to train Google’s models unless you submit feedback, according to Google’s current Privacy Hub. That setting does not eliminate service, response-generation or safety processing. Separate Google controls, including Web & App Activity and Location History, can continue to save data independently. Turning Keep Activity off can also make connected apps unavailable.
The feedback exception is easy to miss. If you deliberately submit feedback on a chat, that submission can be handled for improvement even when Keep Activity is off. Avoid sending feedback from a conversation you are trying to keep outside that use unless you have reviewed the applicable notice.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall4. Use Gemini Temporary Chat for a sensitive one-off
- Start a new Temporary Chat from Gemini’s conversation controls.
- Ask only what is necessary, avoiding names, account numbers and other identifiers.
- Do not treat the temporary label as immediate deletion.
Google’s current Privacy Hub says temporary chats are not used to train Google AI and may be kept with the account for 72 hours for response, feedback and safety purposes. Keep-Activity-off chats can likewise be retained for 72 hours. Google also says chats reviewed by human reviewers and related data may remain for up to three years even after activity deletion. These provider-described periods are retention policies, not a guarantee that every copy disappears at the end of the period.
How to stop Microsoft 365 Copilot personalizing recommendations from work data
Understand the workplace scope
The documented enhanced personalization control applies to Microsoft 365 Copilot work data, not to every consumer chatbot setting. With consent and the tenant’s policy enabled, Copilot can use private Microsoft 365 communications, Microsoft 365 usage data, role and profile information, and data supplied through connectors to personalize the individual’s experience.
Rank #4
That means a work recommendation can reflect signals that are not visible in the chat itself, such as meetings, documents, job role or connector content. The relevant question is not only “what did I type?” but also “which organizational sources is Copilot authorized to use?”
Who can change the setting
A Microsoft 365 tenant administrator controls whether enhanced personalization is available for the tenant or a group. If the setting is missing from your Copilot account, ask the administrator to review it; an individual user may not have authority to change a centrally managed policy. Administrators can also limit the feature to selected groups, so two colleagues may see different options.
What disabling it changes
Turning off enhanced personalization stops that scenario from using the listed work data for personalization. It can also disable dependent features, including Copilot memory. The result is therefore a trade-off: less use of work signals for tailored responses, but potentially fewer personalized capabilities. Disabling the scenario does not erase existing Microsoft 365 records, mail, documents or connector data; retention of those sources remains governed by the organization’s Microsoft 365 policies.
If the concern is a confidential document, use the organization’s data-classification, sharing and retention policies as well as the Copilot setting. A personalization switch is not a substitute for removing a user’s access to the underlying file or mailbox.
A practical decision path for any chatbot
- Identify the account. Determine whether it is a personal, school, employer-managed or workspace account. Administrator controls can override what an individual sees.
- Identify the data source. Look for saved memory, prior chats, profile fields, connected apps, location, workplace communications and usage signals.
- Choose the immediate control. Turn off memory for future personalization; disable the model-improvement setting for training use; delete chats or activity for existing records; disconnect apps to stop new imports.
- Use a temporary unpersonalized chat. Select it before sending a sensitive one-off question, and remember the provider’s safety-retention period.
- Verify after the change. Reopen the memory or activity panel, confirm the setting, and test with a neutral question. A setting can be unavailable, delayed or limited to a particular account type.
Verification should be specific. Check that a saved item disappeared, that the old source chat is gone, that the app authorization shows disconnected and that the model-improvement control remains in the intended state. Do not test by repeating a sensitive fact in a new prompt; use a harmless preference or ask what personalization sources are enabled.
Comparison of the documented controls
| Service | Main personalization source | Individual control | Who controls access | What remains afterward |
|---|---|---|---|---|
| ChatGPT | Saved memories, referenced chat history, connected apps and optional location | Settings > Personalization > Memory; separate Data Controls setting; unpersonalized Temporary Chat | Usually the account holder, subject to plan, region and workspace availability | Original chats remain until deleted; temporary-chat copies may be retained up to 30 days for safety; disabling chat-history reference schedules remembered information for deletion within 30 days. |
| Gemini | Memory, Gemini chats, connected apps and related Google activity | Settings and help > Personal Intelligence; Gemini Apps Activity > Keep Activity; Temporary Chat | Personal-account eligibility and Google account controls; availability varies | Temporary and Keep-Activity-off chats may be kept 72 hours; service and safety processing continues; reviewed chats and related data may remain up to three years. |
| Microsoft 365 Copilot | Work communications, Microsoft 365 usage data, role/profile data and connector data | Enhanced-personalization scenario setting | Tenant administrator for the tenant or group | Turning it off may disable Copilot memory and other dependent features; underlying workplace records remain under organizational retention policies. |
Common mistakes and how to avoid them
- Assuming one privacy switch covers everything. Memory, history, training and app access are separate purposes. Map the switch to the purpose before changing it.
- Deleting a chat but not its saved memory. Check the memory panel after deleting conversations.
- Turning off model improvement and expecting a blank slate. A training opt-out does not erase history or stop personalization.
- Disconnecting an app and forgetting old content. Remove conversations that already copied the app’s data.
- Using a temporary chat and then saving it. Saving converts it to a regular chat with normal account settings.
- Interpreting a retention period as immediate deletion. ChatGPT may retain temporary copies up to 30 days; Gemini may retain temporary or Keep-Activity-off chats 72 hours, and reviewed Gemini data can persist up to three years.
- Expecting a workplace user menu to override an administrator. Ask the Microsoft 365 tenant administrator when enhanced personalization is centrally managed.
- Leaving identifying details in a question that does not need them. Even when personalization is disabled, providers still perform service, response and safety processing. Data minimization in the prompt reduces exposure at the source.
- Assuming account deletion and chat deletion are the same. A provider’s account-level request may have different scope and exceptions from deleting an individual conversation; follow the provider’s current data-request instructions when you need broader removal.
What a realistic privacy result looks like
After the right controls are applied, future recommendations should have fewer eligible personal signals, but no provider setting described here promises zero processing. Chatbots still need to process a prompt to answer it, and providers may retain limited copies for safety, feedback or service operations. Previously stored chats, memories, app records, workplace documents and data reviewed by people follow their own deletion or retention rules.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →The strongest practical routine is therefore layered: clear the memory, remove the source chats, disconnect the supplying app, adjust the separate model-improvement control, and use an unpersonalized temporary chat whenever the question is both sensitive and one-off. Recheck these settings after major product updates, account changes or a move between personal and managed work accounts.
Best Value
Frequently Asked Questions
How do I stop ChatGPT from using my chats for recommendations?
In ChatGPT, open Settings, then Personalization and Memory. Turn off chat-history reference and remove saved memories you do not want used. Delete the old source chats separately. If you also want to stop eligible personal-account chats being used to improve models, change Data Controls and turn off Improve the model for everyone; that setting does not replace the memory steps.
Does turning off chat history also stop training?
No. Chat-history reference, memory and model-improvement use are separate controls. On ChatGPT, change the Personalization setting for chat-history reference and the separate Data Controls setting for model improvement. On Gemini, Memory and Keep Activity likewise have different purposes.
Will a ChatGPT Temporary Chat be deleted immediately?
No. OpenAI says a temporary chat may be retained for up to 30 days for safety. It does not create or update memories and is not used to improve models while temporary. Saving it converts it to a regular chat with the account’s normal settings.
How do I turn off Gemini memory?
In Gemini, open Settings and help, choose Personal Intelligence, and turn Memory off or delete individual remembered details. Memory availability can depend on having a personal Google Account, being at least 18 and keeping Keep Activity on.
Does turning off Gemini Keep Activity erase old personalization?
Not necessarily. Keep Activity controls activity storage and model-improvement use; it does not automatically remove a saved Memory item or delete every old chat. Delete the relevant chats and review Personal Intelligence separately.
How long does Gemini keep a Temporary Chat?
Google’s Privacy Hub says temporary chats, and chats made with Keep Activity off, may be kept with the account for 72 hours for response, feedback and safety. Data reviewed by human reviewers and related information may remain for up to three years even after activity deletion.
How do I stop Copilot personalizing recommendations from my work data?
Microsoft 365 Copilot’s enhanced-personalization scenario is controlled by the tenant administrator. Ask the administrator to disable it for your tenant or group if the user setting is unavailable. Disabling it can also turn off dependent features such as Copilot memory, and it does not erase the underlying Microsoft 365 records.
Recommended Free Tools
Does disconnecting a connected app delete information it already supplied?
No. Disconnecting prevents future access, but prior conversations may still contain copied app data. Delete those chats or records separately, and check the connected app’s own authorization and retention controls.
Can I get a recommendation without any personal data being used?
Use an unpersonalized Temporary Chat when the service offers that choice, and omit names, account numbers and unnecessary identifiers. The provider still processes the prompt and may apply safety, service and short-term retention rules, so this reduces personalization rather than promising zero processing.
The Bottom Line
Bottom line: Match each privacy goal to its own control. Turn off and clear memory, delete the chats that supplied the detail, disconnect connected apps, change model-improvement settings separately, and use an unpersonalized temporary chat for a sensitive one-off. Then account for the provider’s stated safety retention, feedback exceptions and any administrator-controlled workplace policy.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
