Frequently asked questions

Answers to the questions Aimogen Pro support sees most, from setup and cost to content quality and legality.

Setup and licensing#

Do I need an OpenAI key?#

No. Any one supported provider is enough for text generation. OpenAI is only required for features that are OpenAI-specific: Assistants, Batch requests, fine-tuning, OpenAI Vector Stores, the realtime chatbot, Sora video and Whisper transcription.

Do I need the free Aimogen plugin?#

No. Aimogen Pro is standalone. The free plugin is a companion, and Pro shows a dismissible notice suggesting it. Nothing in Pro is disabled without it. See Aimogen Free vs Aimogen Pro.

My licence is an old Envato purchase code. Does it still work?#

Yes. Both the WPBay and legacy Envato formats are accepted, and no cutoff is currently enforced. See Migrating from Aiomatic.

Can I use it on multisite?#

Yes. Licensing is network-wide; settings and data are per-site; the updater runs on the main site only. See Requirements.

How many sites does my licence cover?#

It depends on your tier. Revoke a licence before moving a site, or the seat stays consumed. See Activation and licensing.

Cost#

How much does it cost to run?#

The plugin licence is fixed; the API cost is yours and depends entirely on usage. The variables are the model, the length, and how many requests each post takes — a typical bulk rule makes around fourteen requests per post before images.

Measure rather than estimate: run one post, read the cost in System & Logs › Usage Logs, and multiply.

How do I stop unexpected bills?#

Four layers, all of which you should use:

1. A monthly spend cap in each provider dashboard — the only one nothing can bypass
2. Usage limits, especially guest credits
3. Rate limiting on the chatbot
4. Sensible model choice and token budgets

See Controlling cost.

Why is my chatbot so expensive?#

Because context is resent on every turn. With 20 messages of history plus retrieved chunks, message 21 sends all of it again. Reduce Max Chat Messages To Send As API Context and the retrieval chunk count.

Which model is cheapest?#

For bulk work, a mid-tier model — gpt-5-mini, gemini-3.5-flash, or a Groq model. For titles and meta descriptions, the cheapest capable model, since they are a few dozen tokens. Cloudflare Workers images and royalty-free sources have no per-unit cost at all.

Content#

Why is my content generic?#

Usually because the placeholders did not resolve, so the model received a template rather than your data. Enable detailed usage tracking and read the actual prompt. See Debugging prompts.

Why do sections repeat each other?#

%%article_so_far%% is not in the section prompt. Adding it fixes most repetition. See Content quality problems.

Why is my content too short?#

Max Tokens too low, the prompt did not ask for a length, or the model output cap was reached. See Tokens, context and output length.

Can I make the content sound less like AI?#

Partly. Ban the recognisable phrases explicitly in your prompts, use spintax for structural variation, and enable parasite phrase removal.

The larger answer is that content which is genuinely useful, specific and grounded in real information reads differently from content which is not — and that is a prompting and sourcing problem, not a post-processing one.

Will Google penalise AI content?#

Google policy targets content produced primarily to manipulate rankings, not automation as such. Publishing large volumes of thin, unedited AI content is a known way to have a site demoted. Using AI to produce genuinely useful content, reviewed before publication, is not.

The practical position: generate to draft, have a person read it, and be honest about whether the result is worth publishing.

Should I disclose that content is AI-generated?#

There is no universal legal requirement, but there are specific ones — product reviews, endorsements and some regulated sectors. Beyond compliance, disclosure is straightforward to add: put it in HTML Text To Prepend To AI Created Content.

Can I generate content in other languages?#

Yes. Set Content Language on the rule. Generating natively in a language usually produces better results than generating in English and translating. See More AI tools.

Chatbot#

How do I make the chatbot answer from my own content?#

Index your content with embeddings, then tick Chatbot Shortcodes under Settings › Embeddings › Enable Embeddings For. That second step is the one people miss.

Why does the chatbot invent answers?#

It has no knowledge of your site unless you give it some, and nothing in its instructions telling it to admit ignorance. Add both. See Knowledge and context.

Can I have different chatbots on different pages?#

Yes, with personas and per-placement shortcodes, or with global injection plus conditions. See Global and conditional chatbots.

Can visitors upload files?#

Yes — images for vision, and PDFs for PDF chat. Set expiry on both. See File uploads, vision and PDF chat.

Is the chatbot GDPR compliant?#

That depends on your configuration and your obligations, not on the plugin. The plugin provides the controls: a consent gate, retention settings, a provider retention toggle and compliance text. What you do with them is your assessment. See Data, GDPR and retention.

Automation#

What is the difference between OmniBlocks and agents?#

You define the steps in an OmniBlock; the AI decides them in an agent. Use a workflow when you can write the steps down — it is cheaper, faster and predictable.

Why does nothing run on schedule?#

Almost always no real cron trigger. See Scheduled tasks do not run.

A rule stopped working and there is no error#

A stale entry in the running list. System & Logs › Maintenance › Clear Running Rules List.

Are agents safe to use?#

They are as safe as the tools you give them. An agent with read-only tools is low risk; one with database write access, file writing and social publishing is not. Read Agent security before enabling anything beyond the read-only set.

Technical#

Which providers are supported?#

Twelve for text: OpenAI, AiomaticAPI, Azure OpenAI, Anthropic, Google AI Studio, xAI, Perplexity, Groq, NVIDIA NIM, OpenRouter, Hugging Face and Ollama — plus any OpenAI-compatible endpoint. Separate sets for images, video, speech, search and vector storage. See AI provider configuration.

Can I run it entirely on my own hardware?#

Close to it. Ollama for text and embeddings, plus a self-hosted Qdrant or Chroma, keeps content on your infrastructure. Image generation still needs a provider unless you run your own.

Where are my API keys stored?#

Plain text in the WordPress options table. Anyone with manage_options or database access can read them. They can be supplied from code instead. See Where credentials are stored.

Why do the option names still say "aiomatic"?#

Deliberate backward compatibility. Renaming options, shortcodes, hooks and tables would break every existing site and every custom snippet. See Legacy Aiomatic identifiers.

Can I use it from code?#

Yes: a PHP API, a REST API, WP-CLI commands, the Abilities API, webhooks and 146 hooks.

My settings keep reverting when I save#

max_input_vars is too low. The settings screens post many hundreds of fields, and PHP silently drops the overflow. Raise it to 5000. See Recommended server configuration.

Getting help#

Where do I get support?#

Open a support ticket. Include your System Info, the error from Activity Logs, and what you did.

What does mean?#

See the Glossary for the terms used throughout this documentation.

Is there a changelog?#

Recent versions are in the plugin readme.txt. The full history is in the Knowledge Base.

Where can I see what changed in this version?#

This documentation describes Aimogen Pro. See Updating Aimogen Pro.