Skip to main content
Create Content is the manual path through the content agent: you pick the prompt you want to win, choose the content type and settings, and the agent researches and writes the draft while you watch. The Content Agent automates the same pipeline from your gap backlog; this page is for when you decide what gets written. Open Create Content in the sidebar and click Run content agent. The same dialog also opens from a content gap recommendation, where the content type comes prefilled and the recommendation rides along as context for the agent.

The configuration dialog

The dialog is titled Content Generation Agent. Its sidebar has two groups: Overview, the core settings you scroll through, and Advanced, optional sections that open on click. Prompt and content type are required; everything else has a sensible default, most of them inherited from your project defaults in Brand Book → AI Agent (the Edit project defaults link in the footer takes you there; see Settings for what lives where).

Prompt

Select the prompt where you want to improve your visibility. The picker searches your active monitors’ prompts and shows each prompt’s visibility score, so you can confirm you’re targeting a prompt where you’re actually weak. Prompts from inactive monitors don’t appear.

Content type

Sixteen types, with six shown first (Article, Listicle, Blog post, Product page, Comparison, Landing page) and the rest behind Show more (How-to guide, Review, Opinion, Case study, Interview, Documentation, Wiki, Press release, and two legacy types). The default is Article. Pick by the intent of the prompt: a “best X for Y” prompt wants a listicle or comparison, a “how to” prompt wants a how-to guide.

Content length

Three options: Short (200-500 words), Medium (500-1,000 words, the default), and Long (1,000-3,000 words).

Persona

The audience the piece is written for. A persona is preselected for you (usually your project’s first persona), so check it matches the audience you’re targeting and switch it when it doesn’t.

Language

The language the AI agent writes in. Defaults to your project’s content language, set in Brand Book → AI Agent.

Tone of voice

Ten presets (Professional, Casual, Technical, Conversational, Neutral, Authoritative, Friendly, Persuasive, Formal, Witty) plus Custom, which requires a description of the tone you want. Defaults to the project tone from Brand Book → AI Agent.

Sources

Optionally select up to 5 competitor references: pages that are currently cited in AI responses to your selected prompt. The agent analyzes their content patterns, topics, and structure to write something that targets the same prompt more effectively. Adding references automatically enables the Analyze Reference Pages tool.

The Advanced sections

  • Content brief: give the agent direction instead of a blank slate: a Title, description, objective, call to action, up to 10 key points, and additional context. You can also upload an existing brief (PDF, DOCX, TXT, or MD) and have the fields extracted from it.
  • Image settings: control the generated featured image: an artistic style (Photo Realistic, Flat Illustration, Watercolor, 3D Render, Minimalist, Isometric, Editorial Photography, Abstract, or Auto), custom image instructions, brand adherence, and Include text on image to overlay the blog title and subtitle. Only applies when the Generate Featured Image tool is enabled.
  • Knowledge base: mark Knowledge Base items the agent should prefer when it searches during research. Requires the Search Knowledge Base tool.
  • Restrictions: blocked words the draft must never use, seeded from your project’s blocked words.
  • Author: the byline used when the document is published to a CMS. Leave it untouched to inherit the project default author.
  • AI tools: the eight tools the agent may use during research and writing: Google Search, Google News Search, Find Internal Links, Crawl Website, Generate Featured Image, Take Screenshot, Analyze Reference Pages, and Search Knowledge Base. Each tool is a switch plus a behavior choice: Allow AI to decide or Always use. Defaults come from Brand Book → AI Agent.

Run it, or park it

The footer has two buttons. Save as draft stores the configuration without generating; it appears in the Create Content list as a draft you can reopen and finish later. Run Agent starts generation and takes you straight to the document. Each run counts against your organization’s content generation allowance for the billing period, the same budget the automated Content Agent draws from (its share is capped by the budget share setting). When the allowance is used up, running the agent prompts an upgrade instead of generating.

After you hit Run Agent

You land on the document page while the agent works. The Agent Workflow panel streams the run: the Research agent searches and reads (every tool call is listed), condenses its findings into a Research brief, and hands off to the Writing agent, which drafts the piece section by section. Once it finishes, the document is a normal editable draft: rewrite anything in the editor, and check the outline and content stats in the sidebar. From there, publishing works like every other Promptwatch document: push it to your CMS as a draft or live page from the publication card, or record the URL if you publish elsewhere, and the live page lands in the Page Tracker so you can watch it earn citations. The full last leg is in Publishing and tracking results.