Routing — which model does what.
Three rows: Text for Compose, Blog for BlogCraft, Image for rendering. Blank means our included AI — except on the Image row, where blank means no picture at all.
A connected key does nothing on its own. Routing is where you say which model does which job — and it's the step people skip. It lives directly under the provider cards on the AI Models page.
The three rows
Three jobs, three rows. That's the whole model.
| Row | What it powers | Blank means |
|---|---|---|
| Text | Compose — the posts it drafts, and the image prompts it writes | ScoutRival's included AI |
| Blog | BlogCraft — the body of a generated article | ScoutRival's included AI |
| Image | Rendering an actual picture from an image prompt | Nothing renders |
Text and Blog are separate on purpose. Most people want a fast, cheap model drafting social posts all day and a heavier one writing the occasional 2,000-word article. Splitting them is the entire point of having two rows.
Assigning a model
Each row is a single dropdown listing every connection that can do that job. Pick one, and it saves immediately — there's no separate save button.
- The model comes with the connection. Selecting a connection automatically uses the default model you chose for it when you connected. The row shows you which one — for example OpenAI · GPT-5.5.
- To change the model, not the provider, click Manage on that provider's card and change its default. The routing row follows.
- A green “Your key” chip appears on any row running on your own account. A grey Included chip means it's on ours.
If you connected a provider but left its default model blank, there's no model to send the request to — so the row shows your key, and generation quietly runs on ScoutRival's included AI anyway. If your key doesn't seem to be doing anything, this is the first thing to check: Manage the connection and make sure a model is actually selected.
Blank means our included AI
Leaving a row on ScoutRival (included) is a perfectly good answer, and it's the default. You can mix freely — Blog on your Claude key because you care about long-form quality, Text left on ours because the posts are fine and you'd rather not think about it.
Setting a row back to blank at any time returns that job to the included AI, immediately. Nothing else is affected.
The Image row has no included fallback
This is the one thing on this page that catches everybody, so it gets a callout of its own.
Text and Blog fall back to our included AI when the row is blank. Image does not. ScoutRival has no included image model, so with the Image row unset you will only ever get the image prompt — never a picture. The row shows Not set rather than Included, and in Compose the button reads “Connect an image model to render”. That is the product working as designed, not a bug. The full explanation →
To get pictures inside ScoutRival, connect OpenRouter, OpenAI or Google (Gemini), give it a default image model, and point the Image row at it. Then Generate with my model appears under every image prompt in Compose, and the render is free — you're paying your provider, not us.
Anthropic never appears in the Image row
Claude is a text model. Anthropic has no image-generation API at all, so an Anthropic connection is simply not offered as an option on the Image row — the dropdown only lists connections that can actually render. If Anthropic is the only key you've connected, the Image row will have nothing to choose but blank. Connect a second provider for images.
What routing doesn't cover
Only those three jobs run on your key. The rest of ScoutRival — the Daily Brief, competitor analysis, SEO audits and AI-visibility checks — always runs on our included AI, regardless of what you've connected. Those are the intelligence layer, and they're not routable.
Within BlogCraft, it's the writing of the article that goes to your Blog model. The surrounding steps — title suggestions, the outline, the SEO pack — run on our AI.
When your key can't be used
ScoutRival tries your key first and, if it can't use it, falls back to the included AI so your generation still completes. It falls back when:
- The row is blank — the normal case.
- No model is set on the connection (see the callout above).
- Your provider rejects or fails the request — a dead key, an empty balance, a rate limit, a model your account can't access, or a timeout.
When your key fails, you are not told. The post or article arrives looking completely normal, written by our AI, and your credits are spent as usual. If you connected a key and your balance is still moving the way it did before, start with AI Models troubleshooting.