Google Gemini down as thousands report errors and missing chats

Google’s Gemini AI service suffered a sudden, widespread outage on Thursday, leaving thousands of users unable to generate responses, log in, or access their recent chat histories.

The disruption escalated quickly on the afternoon of 8 October 2026. Reports on the tracking site Downdetector shot up from about 1,700 to more than 5,000 within a matter of minutes, indicating a broad issue across the platform. Users trying to run simple queries were instead blocked by persistent error messages, with many seeing a prompt reading: "I encountered an error doing what you asked. Could you try again?"

What happened to Gemini?

According to report breakdowns, the vast majority of the issues—some 69 per cent—were tied directly to AI generation itself. This meant that while users could open the Gemini website or mobile app, the system failed the moment they tried to submit a prompt. Website issues accounted for 13 per cent of the reports, while app-specific problems made up another 12 per cent.

Some users encountered "error code 13" when typing in inputs. Google's own help documentation defines this error as a server-side handshake failure, browser conflict, or account-level desync that blocks text inputs or uploads. Others faced "error 1155," which simply told them that "something went wrong." For most, reloading the browser, clearing the cache, or logging out did nothing to resolve the issue.

Error codes and disappearing chats

Beyond failing to answer questions, Gemini also began hiding recent conversations. This left users temporarily unable to access previous work, ongoing research, or writing drafts. While frustrating, it is a problem Gemini users have seen before; Google had to fix a similar bug in February 2026 when a background process broke conversation metadata.

Did Google acknowledge the outage?

Google did not officially acknowledge the disruption. Throughout the afternoon, its Workspace and Gemini status dashboards remained green, showing "all systems operational." It is possible the hiccup was resolved before it triggered the company's automated public status updates, or that the scale of the error did not meet Google's internal threshold for a system-wide incident.

Indeed, the issue resolved nearly as quickly as it started. Within an hour of the peak, Downdetector reports had fallen back to baseline levels, and individual tests showed the service was back to responding quickly and accurately.

The outage came just a day before Google planned to shift its model access on 9 October, a change that will restrict free users to its Flash-Lite model. However, there is no evidence connecting those planned changes to the brief afternoon crash.