AI providers

Writrex connects directly to the provider you choose, using your own API key. You pay the provider for usage; Writrex adds no markup and no per-word charge.

Overview

Provider	Good for	Plan
Google Gemini	Balanced quality, speed and cost. The recommended starting point.	All plans
OpenAI	Strong long-form and detailed FAQ output.	Starter+
Anthropic Claude	Natural, human-sounding copy for client-facing pages.	Starter+
Perplexity	Web-grounded research before writing.	Starter+

Configure providers in Writrex → Settings → AI Provider Configuration and Additional AI Providers. Typical generation cost is a few cents per page, depending on the model and page length.

Live model discovery

Writrex doesn't hard-code a model list. After you save a key, it asks the provider which text models that key can use and lets you pick one. Use the Refresh models buttons to fetch the latest list.

Gemini, OpenAI and Claude — models are discovered live and cached for a few hours.
Perplexity — choose a model such as sonar-pro in Settings.
If discovery fails, Writrex falls back to a built-in list so generation keeps working.

Availability depends on your account. Some models need billing enabled or aren't available in every region. If a model doesn't appear, check the provider console.

Getting an API key

Google Gemini

Open Google AI Studio and sign in.
Choose Get API key → Create API key.
Paste it into Settings → Gemini API key and save.

OpenAI

Open the OpenAI platform and create a secret key. It is shown only once.
Add billing credit — OpenAI needs a funded account.
Paste it into Settings → OpenAI API key.

Anthropic Claude

Open the Anthropic console and create a key.
Top up billing credit.
Paste it into Settings → Anthropic API key.

Perplexity

Open Perplexity → Settings → API and generate a key.
Paste it into Settings → Perplexity API key.

Primary and fallback provider

Choose a primary provider for pages, blogs, WooCommerce, templates, editor buttons and SEO suggestions. Enable AI fallback and pick a different provider so a failed request or outage doesn't stop a campaign. The fallback only runs on failure.

Research-first generation

With Perplexity research enabled, Writrex sends a research query first, then hands the findings to your writer model to turn into finished copy. This helps location pages where current local facts improve quality. It requires a Perplexity key. The Structured brief option in the generation form adds a research brief that's stored with the page and shown in its Research Brief panel.

Temperature

Lower values are safer and more predictable; higher values add variation. In Pages Automator, the Writing temperature (recommended 0.75–0.9) and Refinement temperature (recommended 0.45–0.65) can be set per generation.

Last updated: 19 September 2026