Gemini errors (deprecated model / empty output)

Symptom

Generation with the Gemini provider fails with a message like "gemini-2.0-flash is no longer available", or Gemini returns an empty response so no SEO is produced.

Cause

Google retired the gemini-2.0-flash model, so requests to it now error. Separately, newer Gemini Flash models can spend their entire output budget on internal "thinking" and return no visible text, which AISA reports as an empty response.

Fix

Update AISA to 1.99.826 or later. That release switches the Gemini default model to gemini-2.5-flash (the current available model) and sets thinkingBudget to 0, so the model spends its budget on the actual answer instead of hidden reasoning — eliminating the empty-output case. After updating, confirm in Settings → API keys that the Gemini model is gemini-2.5-flash (or another currently available model) and re-run generation.

Questions people ask

I'm already on a recent version but still see the old model — why? A model id saved earlier can persist. Set the Gemini model to gemini-2.5-flash in Settings → API keys.

Generation works but the result is blank — same fix? Yes. The empty-output case is the thinkingBudget issue, fixed in 1.99.826; updating resolves both.

Can I just switch providers? Yes — Claude, OpenAI or DeepSeek will work meanwhile, but updating fixes Gemini directly.

"Empty response from provider" / connection errors
Rate limit / 429 errors

Gemini errors (deprecated model / empty output)

Symptom

Cause

Fix

Questions people ask

Related