Writing placeholder copy twice — once to see the layout, once for real — is one of the slower parts of building a page. The AI text generator writes directly into the block you are editing.
It connects to your own OpenAI or Google Gemini account. Nexter does not provide the model or the credit; you bring the key and pay the provider.
Before You Start
- An API key of your own, from OpenAI or Google Gemini.
- You have to be an administrator. Measured across four roles: the AI endpoints require the
manage_optionscapability, which administrators have and editors, authors and subscribers do not. An editor cannot generate AI content. - The AI integration switched on in Nexter’s settings before anything appears in the editor.
Required Setup
Turn the AI integration on, then add the key for whichever provider you are using.
How Your Key Is Stored
- Your API key is stored encrypted. Measured: the settings are written as an encrypted blob and decrypted when used, so the key is not sitting in the options table in readable form.
- You have to be an administrator. Measured across four roles: the AI endpoints require the
manage_optionscapability, which administrators have and editors, authors and subscribers do not. An editor cannot generate AI content. - It costs your visitors nothing. Measured: the AI layer is loaded only inside wp-admin or on its own REST endpoints, so no AI code runs on a front-end page view.
Adding Context
A business description and a target audience can be saved once and applied to every generation, so you do not have to repeat them in each prompt.
- Both are stored with your other AI settings and sent with requests.
- They are a starting point, not a style guide. They shape tone; they will not stop the model producing something you would not have written.
Generating Text
- If you enable both providers, Gemini is used. Measured from the selection logic: Gemini is checked first, and OpenAI is only reached when Gemini is off or has no key. Enabling both does not load-balance or fall back.
- The default models are recent ones — measured as
gpt-5.2for OpenAI andgemini-3-pro-previewfor Gemini — and both are selectable. - A maximum token setting caps each response, which is also what caps your cost per generation.
- Requests go directly from your site to the provider. Measured destinations: OpenAI’s chat completions endpoint and Google’s generative language endpoint. Nothing is routed through Nexter.
Modifying Generated Text
Generated text can be regenerated or adjusted without starting again.
Repeater Blocks And Code
Repeater-based blocks can be filled item by item, and the code block can generate code as well as prose.
Limitations
- What this page verified, and what it did not. Setup, permissions, storage, provider selection and the exact failure messages were measured on a site with no API key. Generation itself was not exercised, so nothing here describes output quality, token cost or rate limits.
- You pay the provider directly. Every generation is billed to your own account, and nothing in the editor shows a running cost.
- Enabling both providers does not give you a fallback. See the Gemini precedence above; if Gemini’s key is wrong, it fails rather than trying OpenAI.
- Administrators only. A site where editors write the content cannot use this without giving them administrator access, which is usually the wrong trade.
- Generated text is a first draft. It does not know your product, your claims or your legal constraints, and it will state things confidently that are not true.
- Model names change. The defaults measured here are current to this version; providers retire models on their own schedule.
Troubleshooting
The AI options are not in the editor
The integration is not enabled in Nexter’s settings, or you are not logged in as an administrator.
Generation fails immediately
The message you get with nothing configured is No AI service is enabled or API key is missing for text and prompt enhancement, and Settings not found for image generation — two different messages for the same cause, so do not read the second as a separate fault.
Responses are being cut off
The maximum token setting is capping them. Raise it, and expect the cost per generation to rise with it.
I set an OpenAI key but Gemini is being used
Gemini takes precedence when both are enabled. Switch Gemini off to use OpenAI.
