Buffer

Help CenterSubscribe to updates
Powered by
Privacy policy

·

Terms of service
Write-up
Issue with Free AI Tools
View the incident
INC-330: Free AI Tools Service Not Working

Date: February 2, 2026

Duration: ~2 hours 40 minutes (5:07 PM – 7:47 PM UTC)

Severity: Minor

Summary

On February 2, 2026, Buffer's free AI tools — including the social media post creator — were unable to generate responses for users. The issue was traced to a billing configuration issue with a third-party AI provider account. Once the billing configuration was corrected, the tools returned to normal operation.

Customer Impact

Users of Buffer's free AI tools on the marketing website were unable to generate AI-powered content during the affected period. No customer reports were received, and core Buffer publishing and scheduling features were entirely unaffected. The impact was limited to the free standalone tools on the website.

Root Cause

A billing configuration issue on our third-party AI provider account caused API requests from the free AI tools to be rejected, resulting in content generation failures. The error responses were generic, which initially made the root cause difficult to pinpoint and led the team to investigate other possible causes before identifying the provider-side issue.

Resolution

Correcting the billing configuration with our AI provider immediately restored functionality. All free AI tools returned to normal operation.

Key Learnings
  • Third-party account health needs proactive monitoring. Adding alerts for changes in third-party service status or billing configuration would help catch issues like this earlier.

  • Generic error messages slow down diagnosis. The API returned generic error codes that initially led the team to investigate the code rather than an external provider issue. Keeping external service health in mind as a possible cause can save time on investigations.

  • Low-traffic services need monitoring too. Because these tools receive relatively low traffic, basic health checks or synthetic monitoring would help surface issues more quickly, regardless of user report volume.