Skip to content

Create an article and review its brief

Copy for agent Copied

The New article flow turns a topic into a plan before writing the article. It asks what the article should answer, which project facts and documents are relevant, and what style to use. PromptEye then prepares a brief for review. You can save the brief for later or continue to an article draft.

Choose New article on Content → My Content. The menu has three starting routes:

  • Write for a prompt you monitor is the direct path when you already track the question you want to answer. Select the prompt from the project’s monitored prompts. The prompt acts as the article’s target and supplies context for fan-out, title, and FAQ planning.
  • Write from your own title is for a subject you want to cover when there is no suitable monitored prompt. Enter an article title and ask PromptEye to suggest a measurable prompt based on it. The suggested wording removes the brand name and promotional phrasing to make the question sound like something a person might ask. You can edit the suggestion, switch back to choose a tracked prompt, or skip adding the new prompt to monitoring.
  • Optimize existing content starts with the text of an article you already have. You can paste the text, fetch it from a public URL, or upload DOCX, PDF, or TXT. Continue to Optimize existing content for the import and fit-review steps.

For a monitored prompt, the selection view can include its group or category, average visibility, citations, quality, priority, article counts, and recent changes. The helper text explains these values: visibility is the average presence of the brand in AI answers over the recent period; citations count when pages are used as sources; quality describes the role of content in answers; and priority estimates how worthwhile it may be to act. Use the values to choose a target, not as a prediction that one article will change the score.

If you start with your own title and choose to add the suggested prompt to monitoring, PromptEye adds it as a monitored prompt, which uses one of the workspace’s prompt slots. Adding it is optional. You can still generate an article without monitoring the prompt, but the later Content workflow cannot show visibility impact for a question that is not monitored.

The Knowledge step lets you choose which project materials should inform the article. The source picker can include documents from both the project’s private base and the organization’s shared base, plus Brand input, the project’s company profile. Select material that directly helps answer the target prompt. A broad catalog is not automatically useful for every article; irrelevant material can introduce facts that do not belong in the piece. See Company knowledge and document sources to understand the difference between project context and selected document sources.

The screen shows how much source material is available, selected, or missing for the available knowledge areas. The wizard also displays the selected character count and does not let the source selection exceed its supported size. On a first visit, the interface may preselect some available documents; review those choices instead of assuming they all apply. You can change the selection before continuing.

Brand input is the project’s company profile: industry, product or service category, target audience, ideal customer, operating area, and company description. It is not a Knowledge document selected for just this article. You can fill in or edit it from this step or on the Company profile tab of Knowledge. Changes apply to the whole project, not only this article. See Company profile for Content.

In Style, select the sound and shape of the article. Available choices include tone of voice, writing style, brand presence, narrative perspective, audience level, structure, and call to action. Examples range from professional or friendly, to educational or narrative, from beginner to expert, and from a FAQ structure to continuous prose.

These preferences influence presentation. They do not replace the always-applied writing rules shown in the brief, such as giving a direct answer early, using concrete evidence, or keeping sections useful. You can use the default educational and informational style or make a more specific choice. If you need an instruction that is not covered by a setting, add optional editorial guidelines. They are followed when the article is written, so write them clearly and keep them relevant to the article.

The Brief step is a plan, not the finished article. It proposes related questions, a title, and a heading outline for the target prompt. It usually takes several seconds. If the brief request fails, use its retry action or return to the previous step to adjust the prompt or source selection.

A query fan-out is a group of related search questions and phrases derived from one target prompt. The target prompt remains the main question. Fan-out explores useful nearby questions: comparisons, costs, requirements, how-to steps, use cases, or narrower audiences. These variants help PromptEye plan what an answer should cover.

For example, a target prompt about selecting accounting software for a small company may produce related phrases about comparing tools, tax features, number of users, or setting up a first account. The brief can use relevant variants as sections or FAQ questions. A query about payroll for a large enterprise may be a different topic and can become an idea for a separate article instead.

Fan-out is a planning aid. It is not a report of verified monthly search volume, a record of questions that actual people asked, or a promise that every AI model runs those exact searches. The brief generator’s progress text may show a phrase count, but the final phrases available to the article depend on the generated result and its classification. Use the list and counts in your own brief.

Fan-out can contain phrases in more than one language. If the phrases point to the same subject but readers use different languages, the interface guidance recommends planning two separate articles, one for each language, so each article matches its readers’ wording and intent. In the article wizard, the article’s language follows the project’s market; check it in Project settings before generating the brief. The market is set when the project is created and cannot be changed, and the wizard has no separate language switch. Review the phrases grouped into the brief and exclude those that do not fit the article’s language. To produce a version for another language, use a project created for that market.

The brief shows the target prompt and current visibility context where available. The proposed article title should sound natural and include the target question closely enough that the relationship is clear. The brief helper identifies the prompt and a fan-out phrase plus brand as title context. Edit the title if it is misleading, over-promotional, too broad, or does not match the facts you can support.

The phrases section shows how many phrases are selected from the available set. The brief describes a natural use target of roughly two to three occurrences per phrase, but says to paraphrase when an exact phrase would read awkwardly. Use Exclude phrase when it does not belong in this article; Restore phrase makes it available again.

The outline groups phrases into the current article. PromptEye can also split off phrases that deserve separate articles. The Next article ideas table lists these topics, what prompt they relate to, why they are suggested now, and a priority. Common reasons include missing coverage, a side topic, a niche product, a different segment, or a cluster gap. These are suggestions generated from the brief; review whether each one fits the business before turning it into more work.

Removing a heading from the outline does not delete its phrase. The phrase returns to the pool and may be available in future suggestions. This lets you keep one article focused without discarding the related topic entirely.

The outline controls the order and focus of the article. You can drag headings to reorder them, add headings, edit section notes, or remove a heading. The brief says its H2 order is binding and allows H3 subsections; preserve the intended heading hierarchy when you hand the brief to a writer.

A section note can ask for a direct answer under the heading or mark FAQ questions derived from fan-out. The FAQ questions are intended to be woven into the article content. The generator does not need a detached block of generic questions appended after unrelated sections.

When optimizing an existing article, the brief can show how many fan-out phrases have coverage in the source text. A low source-fit warning means the selected prompt and current article may not align well. The screen can warn when coverage falls below its threshold and give you a choice: return to select a better prompt, or continue knowing the generated text may be largely new. The fit percentage is a comparison signal, not an editorial grade for the original article.

The brief includes writing rules that apply across style choices. They ask for a direct answer in the opening of each section, the most important information first, and concrete numbers or dates when selected sources support them. The rules also encourage short paragraphs, lists, selective bolding, a useful comparison table where appropriate, fan-out FAQ content in the body, and a natural brand reference in a heading.

These rules are guideposts for the generated text, not proof that every output is factually correct. Check every factual claim, number, source, and product statement before publication. The brief explicitly ties numbers to selected sources; if a needed fact is missing, add an authoritative source or change the brief rather than letting the article invent one.

If you want a writer to produce the article, use Export brief. Brief only downloads a DOCX with the goal, writing rules, style, outline, and phrases. Full package downloads a ZIP with the brief and selected source documents. The export also carries project and article metadata. Keep the source files with the brief when the writer needs their supporting facts.

You can also finish on the brief. PromptEye saves it in the library, where you can reopen it, review the plan, export it, or generate an article later. Saving a brief does not publish a page or start citation tracking. The saved screen offers two next steps: write the article from the brief or, after the article has been written and published, add its URL to track citations.

Choose Generate article when the plan is ready. The article is written section by section from the brief, with selected company knowledge and the configured writing rules. The progress screen gives an estimated duration; generation can take longer or fail. If the request errors, the interface reports that it could not generate the article and lets you return to the brief to retry or revise it.

The article view shows what was applied, such as direct answers, FAQ questions from fan-out, a comparison table when the brief calls for one, selected-source numbers, and phrase use. It also shows what the article was built from: target prompt, fan-out phrases, sources, and brief. Treat these as an explanation of the generation inputs, not a final independent fact-check or citation result.

The generated article is a draft. Review its accuracy, tone, links, and usefulness. Content generation does not publish it, does not guarantee indexing, and does not guarantee that an AI answer will cite it. Mark it Ready when you consider it prepared for your publishing process. See Library, publishing, indexing, and citations for the next steps.

Can I generate a brief and write the article later?

Section titled “Can I generate a brief and write the article later?”

Yes. Finish on the brief to save it in the library. Open it later to review or edit the plan and start article generation when you are ready.

What if a fan-out phrase does not fit my audience?

Section titled “What if a fan-out phrase does not fit my audience?”

Exclude it from the current brief. The phrase remains separate from the article, and removing its heading returns it to the phrase pool. If it represents a valuable distinct audience or topic, consider it as a separate article idea.

Can I add a monitored prompt after starting from a title?

Section titled “Can I add a monitored prompt after starting from a title?”

Yes, when a suitable prompt is suggested. Adding it to monitoring is optional and uses one of the workspace’s prompt slots. Without monitoring, the writing workflow continues, but later monitored visibility impact is not available for that title.

Does changing Brand input affect only this brief?

Section titled “Does changing Brand input affect only this brief?”

No. Brand input is the project’s company profile. A substantial change can affect monitored prompt evaluations across that project, so review them in Prompt Intelligence after changing core company facts.

Is the brief a guarantee that the generated article will be cited?

Section titled “Is the brief a guarantee that the generated article will be cited?”

No. It structures content around a target question and related phrases. Indexing, selection by AI systems, and citations are separate outcomes.