How reliable are the AI services? The last 30 days

This compares the services we watch over the 30 days to 23 October 2026. It updates as new checks come in, and a snapshot is saved at the start of each month.

Of the 10 services with enough data, Microsoft Copilot had the highest share of fully operational checks at 100% and ChatGPT the lowest at 93.9%. ChatGPT had the most incidents on its provider's status page, 20 in each case. Page-wide incidents on shared status pages are counted, so a count can include faults that did not touch the service itself. Every status change we saw on a service with a readable status page coincided with an incident the provider reported.

Services ranked by the share of our checks that found them fully operational.
ServiceFully operationalWeakest dayProvider incidentsStatus changes we saw
Microsoft Copilot100%7 Oct 2026 (100%)No readable feed0
DeepSeek100%7 Oct 2026 (100%)No readable feed0
Gemini100%7 Oct 2026 (100%)None0
GitHub Copilot100%7 Oct 2026 (100%)None0
Grok100%7 Oct 2026 (100%)No readable feed0
Mistral Le Chat100%7 Oct 2026 (100%)None0
Perplexity99.8%9 Oct 2026 (97.2%)1, 41 min in total1
Claude97.1%7 Oct 2026 (13.6%)3, 14 h 37 min in total2
Cursor94%16 Oct 2026 (77.4%)13, 24 h 25 min in total11
ChatGPT93.9%13 Oct 2026 (65.3%)20, 158 h 53 min in total7

How to read this

The percentage is the share of our five-minute checks that found the service fully operational, so it measures what our checks and the provider's status page showed, not what any one customer experienced. It is not a service-level figure from the provider. Differences of a few tenths of a percent are small enough to be noise. A service shows "Building history" until we hold at least three days of checks. Incidents that opened and closed between two of our checks may be missing. Services with no readable status page are compared on our own checks only, and the full method is in how we check.

The first monthly snapshot will appear here at the start of next month.