Gemini now keeps chats with History off for about two months, up from 72 hours. Claude keeps incognito chats 30 days, Perplexity 24 hours
Google's Gemini app now keeps temporary chats and chats made with History turned off "for around two months", according to Google's Gemini Apps Privacy Hub as of 18 November 2026 (its privacy notice is dated 16 November 2026, its questions section 17 November). Before the update the page said 72 hours, according to an archived copy cited by Gizmodo on 17 November 2026; Ars Technica gave the same 72 hours that day. The longer window arrives with a new choice: Google's post of 16 November 2026 says users can now "keep your Gemini chat history, while opting out of generative AI model training". Google says the change is rolling out "in the coming weeks". Of the no-history modes in the table below, Gemini's is now the longest: Anthropic keeps Claude's incognito chats for 30 days, and Perplexity's incognito sessions expire within 24 hours.
How long each assistant keeps chats you keep out of history or training
As of 18 November 2026, from each vendor's own help or privacy page.
| Assistant and setting | How long the vendor keeps the chat | Used to train models | Source |
|---|---|---|---|
| Google Gemini, History off, before the update | 72 hours | Not stated in the reports | Privacy Hub version of 13 October 2026, as reported by Gizmodo |
| Google Gemini, History off or temporary chat, now | "around two months" | No, unless you send feedback | Gemini Apps Privacy Hub, questions section of 17 November 2026 |
| Google Gemini, History on, new training setting off | 18 months by default; 3 months, 36 months or indefinite by choice | No, unless you send feedback | Same page |
| Anthropic Claude, incognito chat | 30 days | No | Claude Help Center |
| Anthropic Claude, chat in history, training setting off | Until you delete it; then off back-end storage "within 30 days" | No | Anthropic Privacy Center |
| Microsoft Copilot app, personal account | The pages give no period and describe no incognito mode | Chats "aren't used to train foundation models" | Microsoft Support |
| Perplexity, incognito | "expire within 24 hours" | The page does not say | Perplexity Help Center |
| Perplexity, "AI data retention" toggle off | "as long as your account is active" | No | Perplexity Help Center |
The table does not cover OpenAI's ChatGPT.
Two figures sit outside every setting. As of 18 November 2026, the Gemini Apps Privacy Hub says Google keeps chats seen by human reviewers "for up to three years", even after you delete them. Anthropic's Privacy Center says it keeps chats flagged by its safety systems for up to two years.
Who this changes things for
The person most affected uses Gemini on a personal Google account for work and switched History off to keep client material out of Google's training data. Until now that was the only way to opt out, as Ars Technica notes, and it cost them their chat history. Once the change reaches them, they can turn History back on, switch off the new training setting and keep their chats. If they leave History off instead, each chat stays with their Google account for about two months, against 72 hours before, and the Privacy Hub says those chats are used "to respond to you, process your feedback, and protect Google, its users, and the public, with help from human reviewers". Temporary chats get the same two months. The Privacy Hub says work and school Google accounts "may be subject to different data handling terms", and Google's Gemini Apps Help says temporary chats are "not available for users signed in to a work or school Google Account".
Gizmodo reported on 17 November 2026 that the History setting inside the app still said 72 hours, and Google's post says the change rolls out "in the coming weeks". As of 18 November 2026, the Privacy Hub already gives two months.
Why two months
Google's post gives safety as the reason: "As AI systems get smarter, we're continuing to invest in safety and security measures". The Privacy Hub adds backup: chats are kept "so they are available in case of a potential system failure". Google declined to comment further to Gizmodo.
My reading: the two changes are one trade. The training setting removes the main reason people turned History off, so those who still do are a smaller group, and Google can now review what that group did for about two months. The timing fits. A day after the post, Google made Tasks in Gemini free, with an agent that calls businesses and pays through Google Pay. In my reading, 72 hours is too short to look into a complaint about a call or a purchase made by an agent.
The other explanation is that 72 hours was the outlier. Anthropic keeps incognito chats for 30 days and flagged ones for up to two years. On that view Google moved closer to the industry, and the agent launch is a coincidence.
My forecast: by 18 May 2027 the Gemini Apps Privacy Hub will name actions taken by Gemini's agent, such as Tasks, phone calls or purchases, among the reasons Google keeps chats made with History off and temporary chats. My reason: if the two months exist to look into what an agent did, Google will need to say so on the page it sends users to. As of 18 November 2026 the page gives four purposes: to respond to you, to process your feedback, to protect Google, its users and the public, and to have chats available after a system failure. If on 18 May 2027 the page still gives only those, I was wrong.