Aimogen ProDocumentation
Product page Support

Content quality problems

Fixing Aimogen Pro output that is too short, repetitive, generic, wrong, or in the wrong format.

Most quality complaints are prompt problems, not model problems. This page works through them by symptom.

Content is too short#

Likely causes and fixes#

  1. Max Tokens too low. Raise it, but not above the model output limit.
  2. The prompt did not ask for a length. Say "1,200 words" explicitly.
  3. AI Content Minimum Character Count not set. Set it, and the plugin asks the model to continue when output falls short.
  4. Section count and paragraphs per section too low. Both bound total length.
  5. The model output cap was reached. gpt-4o-mini will not produce 30,000 tokens whatever you set. Generate in sections.
  6. Use the Above Content Prompt To Create The Entire Article is on. Single-call generation is bounded by one output limit. Turn it off for long articles.

How to diagnose#

System & Logs › Usage Logs shows the request. A finish_reason of length means the output cap was hit. See Tokens, context and output length.

Content is cut off mid-sentence#

Same causes as above, plus PHP timed out before the provider finished. Check max_execution_time — see Recommended server configuration.

Sections repeat each other#

Likely cause#

The section prompt does not include %%article_so_far%%, so the model does not know what it already wrote.

Fix#

Write section %%current_section%% of an article titled "%%post_title%%".

Already written:
%%article_so_far%%

Do not repeat anything covered above. Do not re-introduce the topic.

This single change fixes most repetition.

Every post reads the same#

Likely causes#

One fixed prompt across a bulk run; temperature too low; prompts too prescriptive.

Fix#

Add variation with spintax and random sentences:

{Open with a concrete scenario the reader will recognise.|Open by stating the
single most common mistake people make with this.|Open with what changed
recently that makes this worth revisiting.}

Raise temperature to 0.8–1.0 for general articles.

Content is generic and ignores the topic#

Likely causes#

  1. Placeholders did not resolve. The model received the template, not your data.
  2. The prompt is vague. "Write an article about %%topic%%" produces a generic article.
  3. Source material was not supplied. A YouTube rule without captions, an Amazon rule without product data.

How to diagnose#

Enable Enable Detailed Usage Tracking and read the actual prompt in Usage Logs. If it still contains %%something%%, that placeholder is not available in that context. See Debugging prompts.

Content contains invented facts#

Models produce plausible statistics, dates, prices and citations that are wrong. This is inherent, not a configuration error.

Reduce it#

  1. Supply the facts. Retrieval, product data, captions, scraped source.
  2. Instruct explicitly:
Do not include statistics, dates, prices or named sources unless they appear in
the material provided above. If you would need a fact you do not have, write
around it rather than inventing one.
  1. Lower the temperature.
  2. Review before publishing. Set new rules to draft status.

Content has the wrong tone or style#

Set Writing Style, Writing Tone and Point Of View on the rule, and reference %%writing_style%%, %%writing_tone%% and %%point_of_view%% in the prompts.

For house style, use the global prepend field once rather than repeating it in every prompt.

Content is in the wrong language#

Set Content Language on the rule, and make sure %%language%% appears in the prompts. Instruct explicitly: "Write entirely in %%language%%."

Content contains AI tells#

The recognisable patterns: "In today's fast-paced world", "delve into", "It is important to note", "In conclusion", "landscape", "tapestry", "journey".

Fix#

Ban them explicitly in the global append field:

Do not use the words: delve, landscape, realm, tapestry, journey.
Do not open with "In today's" and do not close with "In conclusion".
Do not begin sentences with "It is important to note".

AI Undetectability › Parasite Phrases Removal also strips common filler automatically. That part of the feature is worth having on regardless of any detection question — see More AI tools.

Content contains model commentary#

Output beginning "Certainly! Here is the article you requested:" means the prompt did not forbid it:

Return only the article. No preamble, no commentary, no explanation of what you
have written.

Formatting is wrong#

No headings. Ask for HTML heading tags explicitly, and set Add Sections To Content As to h2.

Markdown instead of HTML. Ask for HTML, or enable markdown conversion.

Broken HTML. Enable Protect HTML Tags in Edited Text for editing, and keep Publish Edited Content Even if the AI Removed Parts of the HTML Text off.

Code blocks in prose. The model is mirroring your prompt formatting. Simplify the prompt.

Empty output with no error#

Almost always a safety block. The provider refused, and Google in particular returns an empty response with a SAFETY finish reason rather than an error.

Check Activity Logs for the finish reason, and rewrite the prompt. See API and provider errors.

Layouts broken after editing#

Block support was not enabled. Editing Gutenberg or Elementor content as one HTML string flattens it.

Enable Enable Gutenberg Block Support or Enable Elementor Editor Block Support before editing page-builder content, and test on a copy. See AI Content Editor.

Fixing quality, in order#

  1. Which stage produced it? Change that prompt, not another.
  2. Is the prompt specific about structure and length?
  3. Does it have the context it needs? Placeholders, retrieval, source data.
  4. Are model habits banned explicitly?
  5. Is the temperature right?
  6. Is the model capable enough? Test in Model Comparison.

Still stuck? Open a support ticket and include the diagnostics from Aimogen Pro › System & Logs › System Info.