Blog Writer

Generic "write me a blog post" prompts produce generic posts — the model has nothing to work from but its own training. Blog Writer changes the inputs: it fetches the pages you cite, hands their actual text to the model alongside your brief, and drafts from sources instead of from thin air.

How it works

Two streams of input converge on one prompt:

1. URL fetches the pages you list — competitor posts, documentation, a research note, whatever you want the draft to draw on.
2. Parse Data turns the fetched pages into clean text, stripping markup down to the content the model should actually read.
3. Instructions is a free-text input for your brief: the angle, audience, tone, length, must-hit points.
4. Prompt binds the two together — reference text in one field, your instructions in another — so the model writes your piece from those sources.
5. Language Model drafts the article, and Chat Output returns it.

The design point is that references and brief stay separate inputs. You can keep the same instructions and swap the sources, or hold the sources and change the angle, without rewriting a monolithic prompt each time.

When to reach for it

Reach for Blog Writer when a draft should be anchored to specific material: first drafts that synthesize a few sources, briefs expanded into structure, content that has to reflect a real document rather than the model's best guess. It gets you to an editable draft fast — a human still edits before anything ships.

When to reach for something else

If you don't have sources and just want a single shaped response, that's Basic Prompting. If the job is answering questions against a document rather than writing from it, use Document Q&A. And if the sources number in the hundreds, retrieval (Vector Store RAG) scales better than pasting URLs.

Try it

Give it two or three URLs on one topic and this brief:

Audience: technical buyers evaluating workflow automation.
Goal: a 600-word explainer that synthesizes the linked sources into a neutral
overview. Use H2 sections. End with three honest trade-offs. No marketing claims
you can't support from the sources.

The draft will reflect the pages you linked — swap one URL and rerun to see the output shift with its inputs.