Generation Logs

The Generation Log is where you confirm what happened, what is still running, and why something failed.

The Generation Log Table

The full Generation Log is found in the Content > Generation Log sub-tab. It is the main place to monitor all AI activity on your site.

The table provides the following details for every job:

Status: Current state of the job (Queued, Running, Succeeded, or Failed).
Source: How the job was triggered (Manual, Bulk, Scheduled, or Transcription).
Model: Which specific AI model was used for the task.
Keywords: The topics provided to the AI.
Created/Runtime: When the job started and how long it took to complete.
Result: Direct links to “View” or “Edit” the generated post.

The Dashboard Overview

For a quick check, the Dashboard shows the last five generation jobs. This allows you to see at a glance if your recent posts were successful without digging into the full log.

Troubleshooting Failures

Since 4.4, failed jobs record a specific, actionable message instead of a generic “Content generation failed.” Each message tells you what to do:

Message	What to do
No API key configured for [provider]	Add a key in Settings → Connections.
[Provider] rejected the API key	The key is wrong, expired, or revoked. Check it under Settings → Connections → API Keys.
[Provider] is rate limiting requests	Too many requests too fast. Wait a minute and retry, or switch providers.
[Provider] is having server trouble	The outage is on the provider’s side. Try again shortly or use a different provider.
The prompt is longer than this model can accept	Use fewer keywords or lower the content length setting.
[Provider] does not generate text	You selected an image-only provider (e.g. Stability AI) for a text task. Pick a text model.
That provider is not available	The saved model no longer exists in the plugin. Pick a model from Settings → Connections.
Could not reach [provider]	Your server cannot connect to the provider — check firewalls or outbound connectivity.
[Provider] returned an empty response	Reasoning models can spend the whole token budget before writing. Raise the content length.
[Provider] returned something unexpected	A malformed response. Retry; switch providers if it repeats.
[Provider] returned SEO data that could not be read	The SEO metadata response wasn’t valid. Retry or switch providers.

Since 4.5, failed jobs also show a short “what to do next” line under the message — for example Check your key under Connections → API Keys or Rate limit — wait a minute, or set a fallback provider. The same line appears in the dashboard activity list.

Jobs that succeed with repair (the model hit its token limit and the plugin trimmed and closed the article) show an informational note in the Result column — the post is fine, just shorter than requested.

Fallback notes

When a fallback provider is configured, the Result column records what happened, tinted amber:

OpenAI failed (rate limit) → completed via Claude — the post was finished on the fallback provider.
Claude also failed (server error) — both providers failed; the job’s error is the primary’s.
Claude skipped (no key) / Claude skipped (prompt too long) — the fallback was not attempted.

The Model column always shows the model you asked for; the note tells you which one actually wrote the post.

Support: Failed jobs show the full error inline in the Result column, with a “Copy error” button for support requests. Dashboard activity entries show the error text directly — no hovering needed.

How to use the log operationally

For bloggers: Check that a manual draft actually finished before you start editing it.
For agencies: Use the source and keyword columns to confirm whether a post came from a client template, a schedule, or a bulk run.
For technical teams: Failed jobs are the fastest place to see whether the issue is credentials, provider quotas, or prompt/output formatting.