Google Search outage history and uptime record
Every outage below came from our own automated checks failing twice in a row. Nothing here is derived from user reports, so the record cannot be inflated by a brigade or deflated by a quiet crowd.
Google Search has not failed a single check in the 5 days we have been watching it.
Across 271 checks there has been no occasion where two consecutive probes failed, which is the bar we require before recording an outage. That is a strong record, though a short one is not the same as a good one - read the window length above before drawing conclusions.
Day by day, last 5 days
One bar per day. Green means every check that day succeeded; amber means at least one was slow; red means at least one failed.
Every recorded Google Search outage
Times are UTC. "Cause recorded" is what our probe saw at the moment it failed - It is a symptom, not Google Search's own explanation.
No outages recorded yet
We have not seen Google Search fail two consecutive checks since monitoring began. That is either a very reliable service or a very short history - Check the uptime window above to see which.
How to read these numbers
Uptime here is observed uptime: The proportion of measured time in which our check could reach google.com and get a usable response. It is not Google Search's internal availability figure and it will not match one.
What this record will miss
- Regional outages. Our check runs from one network. A failure confined to one country, one provider or one mobile app often will not appear here at all.
- Short outages. Anything shorter than the gap between two checks can pass unseen between them.
- Functional failures. A service that answers our request but cannot log anyone in, or serves an empty feed, is recorded as up. It is answering; it is just not working.
Why an outage needs two failed checks
A single failed request is not evidence. Packets are lost, routes flap, and a probe can fail for reasons that have nothing to do with Google Search. Requiring two consecutive failures costs us a few minutes of precision at the start of a real outage and saves us from publishing false ones, which is the right trade for a record people cite. It also means visitor reports can never put an outage on this page.
Why the reports chart exists next to this one
An automated check is precise and narrow. A crowd is broad and unverified. Between the two you get closer to the truth than either gives alone, which is why they are shown side by side and never merged. The full methodology is here, including what our data cannot see.
Comparing this with Google Search's own status page
They will differ, and the gap is usually explainable. Most published SLAs exclude announced maintenance; ours excludes nothing. If it did not answer, we recorded it as down. You can see what a given uptime percentage actually permits with our uptime calculator.