Setting up AI: OpenAI vs. Claude vs. Gemini for authors

Three AI providers are supported for text: OpenAI (the ChatGPT people), Anthropic (the Claude people), and Google Gemini.

They all work. They all produce acceptable captions. The difference is in voice, price, and which one best matches your style. Here's the honest comparison.

The one-sentence verdict

New to AI, want the safe default: Start with OpenAI.
Want prose that reads more like a human wrote it: Claude.
Want the cheapest + most generous free tier: Gemini.
Running multiple pen names with very different voices: Claude handles tone-switching best.
Want to split drafting vs editing roles: use Claude for drafting, OpenAI for short punchy rewrites.

Those are rules of thumb. Your mileage may vary based on your specific brand voice guides.

How they sound (same prompt, same guides, three different drafts)

Prompt: "Draft an Instagram caption announcing book 3 of the Driftwood Harbor series. Book is enemies-to-lovers with a lighthouse keeper and a climate scientist. Launching May 10. Tone is warm and slightly wry."

OpenAI (GPT-5):

*Book 3 of Driftwood Harbor is coming May 10 — and this one's going to hurt in the best way.

Enemies-to-lovers. A lighthouse keeper who'd rather you didn't. A climate scientist who came to decommission her lighthouse and stayed for something she didn't plan for.

Pre-order at storylink.to/salt-sky. Tissues optional, encouraged.*

Claude (Sonnet):

*May 10.

Book 3 of Driftwood Harbor. Enemies to lovers. A lighthouse that's been in her family for a hundred years, and the climate scientist sent to prove it shouldn't be.

Somewhere between she'd push him off the cliff and she might push him off the cliff gently, Salt & Sky happens.

Pre-order: storylink.to/salt-sky*

Gemini:

*Salt & Sky — book 3 of the Driftwood Harbor series — launches May 10th.

This one's enemies-to-lovers, set at a Maine lighthouse. The heroine is the lighthouse keeper. The hero is the climate scientist trying to decommission her lighthouse. They have very different feelings about that.

I'm really proud of this one. Pre-order at storylink.to/salt-sky.*

All three are usable. They hit different beats. OpenAI leans punchy-sales-copy. Claude leans writerly and uses rhythm. Gemini is straightforward and informational.

Pick based on which voice is closest to yours — or use one and then tell the AI to rewrite it in the style of the other if you want to experiment.

Pricing (per million tokens)

Here's what each provider actually charges. A "token" is roughly 3/4 of a word, so 1M tokens ≈ 750,000 words in and out.

For a normal author workload (14-day campaign, 30-40 posts, 3 platforms = maybe 500k tokens total input + output):

OpenAI GPT-5: ~$2.50 per 1M input tokens, ~$10 per 1M output. A full campaign ≈ $3-6.
Claude Sonnet: ~$3 per 1M input, ~$15 per 1M output. Full campaign ≈ $4-8.
Gemini Pro: ~$1.25 per 1M input, ~$5 per 1M output. Full campaign ≈ $1.50-3.

These are approximate and the providers change them monthly. Check each provider's pricing page for current numbers.

For most authors generating 1-2 campaigns per month, the difference is under $10. Not worth optimizing on price alone unless you're running very high volume.

Free tier comparison

Each provider offers a free tier for testing:

OpenAI: $5 in credit for 3 months with a new API account. Burns down with use.
Anthropic: $5 in credit for new API accounts.
Google Gemini: Most generous — substantial monthly free quota at the free tier, renewed each month. Many authors never hit paid usage.

If you want to try AI before committing to any paid usage, Gemini is the easiest on-ramp. Create an account at ai.google.dev, grab the key, use it for a month, see if you like having AI in your workflow at all.

What each one is particularly good at

OpenAI strengths

Sales copy and CTAs. GPT models were trained on a lot of marketing copy. They're good at punchy calls-to-action and clear value propositions.
Carousels and structured lists. When you need a clear 5-slide carousel ("5 tropes I used to hate"), OpenAI outputs clean, numbered, parallel structure.
Ecosystem. OpenAI has the most tools built on top of it (custom GPTs, ChatGPT Plus integrations). If you already use ChatGPT heavily, OpenAI will feel familiar.

Claude strengths

Tone-matching and voice consistency. Claude is the best at reading a brand voice guide carefully and writing captions that hit specific, idiosyncratic rhythms. If you have strong individual voice, Claude preserves it better.
Longer-form thinking. For campaigns where each post is thinking-out-loud-style or essayistic, Claude handles it gracefully. OpenAI tends to punchify.
Multi-pen-name workflows. Claude remembers which persona you told it to be better across a campaign generation.
The Claude plugin. Author Automations Social's official plugin is for Claude (Claude Code + Claude Cowork). Using Claude as your text provider and the plugin for conversational workflow is the integrated experience.

Gemini strengths

Cost. Cheapest per token by a good margin.
Free tier generosity. You can test extensively without billing.
Speed. Generally the fastest response times of the three.
Google integration. If you're deep in Google Workspace, Gemini has native integrations elsewhere in your stack.

What each one is not great at (for author work)

OpenAI: Can drift into generic marketing-speak if your brand voice guides are weak. "Don't miss out!" energy if you're not careful.

Claude: Occasionally verbose — Claude will write 200 words when you asked for 100. Tell it "keep under 80 words" explicitly.

Gemini: Formal tone by default. If your brand is casual, Gemini needs stronger voice guides to lean in. Sometimes refuses prompts about fictional violence (thrillers, horror genres) on safety grounds.

Picking your default

If you genuinely don't know, here's the decision tree:

Are you doing a lot of long-form, voice-driven content like blog-adjacent posts or substack-style newsletter captions? → Claude.
Are you doing sales-driven, CTA-heavy launch campaigns? → OpenAI.
Do you want to minimize cost and try AI without committing to a paid provider for a few months? → Gemini.
Are you planning to use the Claude plugin (conversational workflow inside Claude Code or Claude Cowork)? → Claude, so everything is in the same ecosystem.

If you still don't know: OpenAI. It's the safest default and the easiest to find help for online.

Can I use all three?

Yes. In Settings → AI, you can add keys for all three providers and pick your "preferred" one. The tool uses your preferred for new drafts, but you can switch per-campaign or per-post if you want a specific vibe.

Some authors I know:

Use Claude for their main brand (voice-heavy)
Use OpenAI for their pen-name where voice matters less
Keep Gemini on standby as a cheap option for bulk evergreen

There's no penalty for keeping all three keys active. You only pay for what you actually generate.

Switching providers mid-campaign

You can, but it's visible. If you've generated 10 posts with Claude and then switch to OpenAI for the last 4, those 4 read slightly differently. Fine if you're intentional; confusing to readers if you're not.

For launches, pick a provider and stick with it. For evergreen content, switching doesn't really matter.

What to do next

If you haven't set up any AI: pick Gemini for the easiest free test, or OpenAI for the safest default, and paste the key in Settings → AI.
For the mechanics of API keys in general, what's an API key covers it.
For image and video (a different AI provider, with a different pricing model), FreePik, explained.
Once AI is set up, 14 days of content is the first campaign to try.