AI Content Generator supports three AI providers natively: OpenAI (GPT models), Anthropic Claude and Google Gemini. You set up the API keys once in AI Content Generator → Settings, pick a global default provider, and from then on every campaign uses that default unless you override it. The override happens in the AI Provider metabox at the top right of the campaign editor — and that override is where most of the editorial flexibility lives, because different campaigns can use different providers without any extra setup.
This tutorial covers the entire flow: how the global default works, how to override it per campaign from the AI Provider metabox, when to use each provider for which content type, how to mix providers across campaigns running on the same site, and the operational rules of thumb that help you keep the editorial output consistent even when you’re routing different topical clusters to different AI vendors.
By the end you’ll have a clear mental model of “OpenAI for X, Claude for Y, Gemini for Z” inside your own site, and the muscle memory to flip a campaign to a different provider in two clicks whenever you need to.
The global default lives in Settings
Open AI Content Generator → Settings. You’ll see three provider cards stacked vertically: OpenAI (GPT), Anthropic Claude, Google Gemini. Each card has its own API key field and its own model selector. Once a card has a valid key, it’s “configured”; once it’s the default, it’s the one new campaigns use unless overridden.
Below the three cards there’s a Default provider section with a single radio choice: pick OpenAI, Claude or Gemini. That choice cascades to every new campaign — but it does not retroactively change existing campaigns, and it can be overridden per campaign from the campaign editor.
The override happens in the campaign editor
Inside any campaign (new or existing), look at the right sidebar. The top metabox is titled AI Provider. The current selection shows up as a small dropdown labelled “(Global default: …)” with the provider name in parentheses. Click the dropdown to swap to OpenAI, Claude or Gemini explicitly for this campaign.
Below the dropdown the metabox shows the configuration status of all three providers with coloured dots: green dot means “configured and working”, grey dot means “no key set up”. This is the fastest sanity check before launching a generation — if your target provider is not green, the campaign will fail.
When to use which provider
The three providers are not interchangeable. They have different strengths, different price points and different rate limits. A rough decision matrix that holds for most editorial workflows:
- OpenAI (GPT) — broadest model family, fastest on shorter content with the Mini variants (GPT-4o Mini), highest ecosystem maturity. Use it as the default for general-purpose blog content, news summaries, listicles. The cheapest option per article in most cases.
- Anthropic Claude — best long-form quality, very low rate of hallucinations, excellent at following structural instructions (“write 8 H2 sections”). Use it for cornerstone articles, deep-dive tutorials, anything where retention matters more than cost.
- Google Gemini — extremely fast on short content with the Flash variants, generous free tier, the lowest cost per article when configured well. Use it for high-volume short-form content, news roundups, and as a fallback when OpenAI or Claude hit rate limits.
None of this is set in stone — models evolve every quarter and the right choice depends on your specific niche and editorial standards. But these guidelines are a reasonable starting point that has held across many sites.
A practical multi-campaign setup
The most efficient configuration for a multi-cluster blog uses different providers per campaign. A concrete example for a blog covering WordPress, AI plugins and SEO:
- Campaign “Long-form tutorials” → AI Provider: Claude. Target audience: power users; expected reading time: 6–10 minutes; tone: deep and explanatory.
- Campaign “Daily news brief” → AI Provider: Gemini (Flash variant). Volume: 5–10 short posts per day; cost target: under $10/month.
- Campaign “Listicles & comparisons” → AI Provider: OpenAI. Sweet spot of cost vs structure quality; readers expect clean H3 sub-sections with bullet points.
Each campaign overrides the global default explicitly, so changing the global default later doesn’t accidentally retarget the long-form pillar content from Claude to whatever new default you picked.
How to flip a campaign in two clicks
Open the campaign in the editor. In the AI Provider metabox on the right, pick the new provider from the dropdown. Click the Update button at the top of the page. That’s it. The next time the campaign runs (either manually via the Generation & controls metabox or via the scheduled cron) it uses the new provider.
The change does not retroactively re-generate published articles. They stay attached to whichever provider wrote them. If you need to regenerate, that’s a separate operation: pick the article in the Calendar view (or open the post and use the AI Content Generator panel), and explicitly request regeneration with the new provider.
Mixed-provider campaigns are not supported (and that’s fine)
One campaign always uses one provider. You cannot tell AI Content Generator “use Claude for the body and OpenAI for the title”. If you want that level of granularity, the right pattern is to split into two campaigns: one with Claude as the writer, another with OpenAI handling the surrounding metadata. In practice the editorial value of mixing providers within a single article is low — the consistency you lose by mixing usually costs more than the marginal quality you gain.
The exception is fallback. When the primary provider returns a rate-limit error or a transient API failure, AI Content Generator does not automatically fail over to a different provider — that behaviour would be unpredictable for editorial planning. Instead, you manually retry the campaign once the primary provider recovers, or you re-launch it under a different provider explicitly for a specific batch of titles.
Operational rules of thumb
- Configure all three providers even if you only use one. The five minutes you spend pasting in the other two keys today pays off the day your primary provider has an outage and you need to swap campaigns over.
- Never change the global default after the first month of operation. Once campaigns are running with explicit overrides, changing the default has zero effect — but if any campaign was left on the default, it suddenly swaps writer mid-stream. Set the default once, then override per campaign.
- Tag campaigns with the provider in the campaign name. “Blog tutorials — Claude”, “Daily brief — Gemini Flash”. When you have ten campaigns and need to debug a quality regression, the name tells you which AI is responsible in zero clicks.
- Keep a “fallback campaign” template. Once you’ve configured a campaign on each provider, save the campaign as a clone template. Spinning up a fallback the day a vendor has an outage takes two minutes instead of fifteen.
Troubleshooting
- The dropdown shows only “(Global default)” with no other options — the page didn’t finish loading the provider list. Refresh and try again.
- The campaign saves but generation fails with “no provider configured” — the provider you selected has no API key in Settings. Either switch back to the default or add the missing key.
- Provider override doesn’t persist after Update — check the campaign Publish metabox: if the campaign is in “Pending review”, certain fields can be locked. Switch to Draft or Scheduled, update, then switch back.
- Generation cost suddenly doubled — somebody flipped a campaign from Gemini Flash to GPT-4o or Claude Opus. Audit the AI Provider metabox on every active campaign.
- Provider returns a 401 immediately on every call — the API key in Settings expired or was rotated externally. Re-paste the key under the matching provider card in Settings; the campaign picks up the fresh key on the next run with no other change required.
A note on costs
AI Content Generator doesn’t track API spend itself — each provider’s billing dashboard is the source of truth. Open Platform OpenAI’s usage page, Anthropic Console’s usage page, and Google AI Studio’s quota page once a week, scan the daily cost line, and look for unexpected spikes. A spike on one provider while the others stay flat is the clearest signal that a campaign got accidentally swapped.
For predictable budgets, configure a monthly hard limit at each provider directly. OpenAI lets you cap at a dollar amount per month, Anthropic has per-key spending caps, Google Cloud has billing budget alerts. Pair these with AI Content Generator’s per-campaign provider override and the worst case is “a misconfigured campaign hits the provider’s cap and stops” — never “a campaign drains the account overnight”.
You’re done
That’s the full provider model: three providers (OpenAI, Claude, Gemini), a global default in Settings, a per-campaign override in the AI Provider metabox, an explicit choice on every active campaign. Keep all three providers configured, override per campaign explicitly, tag campaigns with the provider name, audit weekly via each vendor’s billing page.
From here, switching providers becomes a routine operation rather than a project: pick the right vendor for the editorial intent of each campaign, leave the rest unchanged, and rely on the matrix of vendor strengths to compose a publishing pipeline that’s more resilient and more efficient than any single-provider setup.