Documentation

Content Editor

Research a page, write a draft, and use review tools before publishing.

The first video shows how to start a draft. The second tours the editor workspace.

Content Editor, the PirateSERP research-backed drafting workspace, builds a page around one primary entity or keyword. It compares your draft with the entities, attributes, relationships, terms, and reader questions found in your research, and marks each one as you use it.

Search data is included. AI writing requires an OpenRouter API key saved in Account Settings. Outline, Review answers, and both rewrite actions run on the model that you select in Model for all AI tools there.

Start a draft

Select the project, then choose one of two research methods.

Start from industry research. Use research you already ran in Service Atlas. This method loads terms, entities, outlines, and questions at once, with no search and no API key.

In Start from Industry Research, select an industry. To filter the list, type in the field.
To narrow the research to one service, select that service. To use the whole industry, keep All services (optional).
Optional: In the Post title (optional) field, enter a title.
Click Create Post.

Research a keyword. Use the pages that rank for a keyword as your references.

In the Keyword you want to rank for field, enter a keyword.
Review Search Location and Search Lang.
Click Submit.
Select the ranking articles to compare against. Content Editor recommends 5 to 10 articles and requires at least two.
Click Continue. Keep the page open while Content Editor analyzes your chosen articles; the editor opens when the analysis is saved.

The results list also shows badges for search features on the results page, such as Featured Snippet, People Also Ask, or Local Pack. Your business profile can supply starting values for either method, so review them before you continue.

Each draft saves to Saved drafts in the active project.

Set the main subject

The main subject is the keyword or entity that question research and generated writing are about.

In the Main keyword / entity field, enter the keyword or entity.
Expand Settings, and review Country, Language, and Search intent.
Click Save target.

The status beside Save target reads Target saved or Unsaved target. Generation tools stay unavailable until you save the target.

In the Search intent list, select the intent that matches the page's purpose:

Choice	Writing guidance
Informational	Explain the subject, answer questions, and connect relevant entities and attributes.
Commercial investigation	Compare alternatives through supported criteria, differences, and tradeoffs.
Transactional	Explain the offer, requirements, and next steps for a purchase, booking, or other action.
Navigational	Help readers identify and reach the named destination, organization, or resource.
Not specified	Preserve the draft's purpose without assuming commercial intent.

The saved intent guides Review answers, Outline, and Rewrite. Changing intent keeps your saved questions. A draft without a chosen intent shows Not specified.

Changing the target keeps the original research and your draft, and it does not collect replacement research. Existing reference analysis belongs to the previous target, so Content Editor excludes it from generated writing. After you change the subject, review the existing entities and references for relevance.

Write or import text

Use the toolbar to add headings (Heading 1 to Heading 3), text formatting, lists, quotes, code blocks, tables, and images. Click Save before you leave the page or open another task.

To import a public page, enter its URL and click Import URL, then confirm Replace the current draft?. The imported article replaces the editor content; click Save to keep it.

Add URL in Outline adds a reference page to your research and merges its terms. It does not import that page into the draft.

Draft history

Click History to open Draft history. Each save that changes the text records a revision; saving unchanged text does not.

In the Revision list, select a revision to compare Selected revision with Current editor.
Optional: To label the revision, enter a name in the Revision name field, and then click Name revision.
Optional: To restore the revision, click Restore revision, review the confirmation, and then click Confirm restore.

Restoring saves the selected content at once and keeps your current text in history, so undo can reverse the change. Click Save to keep a reversal.

If another session changed the draft, save and restore reject the outdated version. Copy your unsaved edits, reload the saved draft, and compare the two before you save again.

Revisions belong to your account, project, and draft. Deleting a draft also deletes its revisions and its saved review.

Choose an editor tool

The editor tools sit in three groups. For more writing space, click Collapse editor tools. To bring the tools back, click Show editor tools.

Group	Tools
Research	Questions, Competitor Gap, and Topic Clusters
Entities & Terms	Entities, Fact status, Attributes, Relations, Terms, Verbs, and Linguistics
Write & Review	Outline, Meta Description, Diversity, Readability, Sentiment, Style Review, and Rewrite

Open tools keep their results while the editor stays open. Save draft changes before you navigate to another page.

If reference analysis fails, click Retry reference analysis.

Review entities and terms

Entities and terms are marked Phrase observed as the draft uses them. Marking updates while you type, delete, paste, or undo, with no button or save required.

Phrase observed is a neutral note. It means the words appear in the draft. It does not confirm that a sentence explains an entity, relationship, or answer. To check a fact, use Fact status below.

Click All, Not observed, or Phrase observed to filter entries by whether the draft uses them.
Click a phrase observed entry to select its matching text in the draft; click again to move to the next match.
To remove an irrelevant suggestion from the research, click the entry's delete icon.

Add only relevant information. Matching the research vocabulary does not require repeated mentions.

Each tool covers one kind of research:

Entities lists named things. Each one shows how many sites back it. For entities from the reference pages you chose, the label reads Core when at least half of those pages mention the entity, and Supporting when fewer do, for example "Core · 3 of 5 sites". For entities from industry research, you see only the site count. Start with Core entities. The summary above the list shows, for example, "12 of 40 researched entity phrases observed in your draft". Click Not observed to see the missing entities, best-supported first. Click Competitor headings on an entity to list the headings on the researched pages that contain it. Click Export CSV to download each entity with its site counts and whether your draft contains it. Identity shows an existing Wikidata identifier or Wikipedia link when one is stored; it does not run another lookup.
Attributes lists entity characteristics as phrases that describe something, such as "tankless water heaters" or "failing anode rods".
Terms lists words and phrases from the research, including hyphenated phrases such as "move-in cleaning" when the research supports them.
Verbs lists actions, and Linguistics lists multi-word phrases from the research, such as "sediment buildup" or "repair transmission". Linguistics never lists a single word. It also leaves out a phrase that the Terms or Attributes tab already shows, so each phrase appears in one place.

What reference pages say

The Attributes tool ends with What reference pages say, a list of the attributes and descriptions your reference pages state, in their own words. Examples include a product weight, a warranty length such as "1-year warranty," an arrival window such as "2-hour window," or "licensed and insured." A row appears only when at least two sites state it.

Each row shows how many sites state it, and the most-stated rows come first.
Open a row to read each site's own wording, with the site's domain. Hold the pointer over an example to read the full source sentence.
Examples marked possible or conditional were hedged or qualified on the page, and a negated claim keeps its "not."
To remove a row, click its delete icon.

When the list is empty, it reads No shared statements found in your references yet. If research that you analyzed earlier shows no rows, analyze your references again.

When research panels are empty

For a draft started from industry research, reopen the editor to load the available research for the saved industry and service. Your draft and removed suggestions stay intact. If the panels remain empty, confirm that the selected service has completed research in Service Atlas, or add reference pages with Add URL.

Check facts against the draft

Fact status lists facts that your research says other pages state about the main entity or service, and checks each one against the draft. Open it from Entities & Terms. The panel updates as you type.

Each fact has one status:

Stated: a sentence in the draft states the fact. The panel shows that sentence and its heading. Click Show in draft to jump to it. A stated fact is not necessarily true.
Not observed: the draft does not state the fact. Other pages in the industry research state it. That is not a requirement, so write the fact only when it fits your page.
Conflicting: the draft gives a different figure for the same attribute, such as another warranty period. Click Show conflicting sentence in draft to see it.
Not evaluated: the check could not run. The reason says why, for example that the draft is empty or that the source sentence names no object to compare.

Each row names its source, such as "Industry research: Plumber, 3 pages", and when the check ran. Fact checks cover English drafts only; other languages show Not evaluated.

Read relationships and topic groups

Relations separates four kinds of evidence:

What sources say lists, for each entity, statements that two or more industry sources make about it, each with the source sentence and a source link. Entities your draft already mentions come first and carry an In draft label, so you see supporting detail for what you are writing about. Expand an entity to read its statements, and click Show all to list every entity. Treat a statement as what sources say, not as a verified fact about your business.
Service connections come from the service research. Each connection carries a label: Appears together, Entity / action, Hierarchy, or Stated relationship. A Stated relationship reads as subject, relation, and object, in the direction the source stated it. Entity and action pairs show associations, not who performs an action.
Reference co-occurrence lists entity pairs that appear together in your reference pages. Co-occurrence suggests a connection; it does not verify a relationship.
Topic groups repeat the group memberships from Topic Clusters. A topic group organizes related subjects for planning; membership does not establish a factual relationship between every pair of entities.

An empty Service connections list reads No service connections are available yet. Service connections and What sources say depend on completed research for the selected industry, service, and language.

To group your research terms into topics, in Topic Clusters, click Build Clusters. To rebuild the groups, click Refresh. Use the groups and their graph to plan coverage and related pages. For a draft started from industry research, Entity groups from industry research also appears: groups of entities that sources confirm and mention together, such as a water heater with its installation, maintenance, and repair. Each group is a candidate section, and the first name is the group's most central entity. The content brief includes these groups as subtopics.

If no entity groups appear, open the industry in Service Atlas and click Research More.

Find competitor keyword gaps

Competitor Gap lists keywords that a competitor domain ranks for and your domain does not.

In the Your domain and Competitor domain fields, enter the two domains.
Click Find Gaps.
Review each Keyword with its Volume and Competitor position.

Domain gaps can span many subjects. Use a phrase on this page only when it serves the saved target, and plan a separate page for a distinct reader need. Competitor Gap does not support Catalan.

Fetch Google questions

Save Main keyword / entity and its market settings.
Open Research, then Questions.
Click Fetch questions.

Questions lists Google People Also Ask (PAA) questions for the saved keyword, country, and language. Typing a keyword or opening the tab does not fetch questions, and clicking Fetch questions also saves the target. Results are saved for the keyword, so other drafts with the same keyword and market reuse them. Capitalization and extra spaces do not count as a different keyword.

After you change the target, click Fetch questions again. Questions from the previous target stay marked as earlier research.

Click Insert to add a question as a heading. If the question already appears in the draft, Find selects it instead; repeated clicks cycle through matches. Finding a question does not confirm that the draft answers it.
Click Dismiss to hide an irrelevant question. Click Show dismissed questions, then Restore, to return it. Dismissals belong to the draft.

Review answers and relationships

Content review, in the Questions tool, judges how well your open draft, including unsaved text, answers its questions and explains its relationships. Save your target and search intent before you click Review answers.

Review uses the model that you select in Model for all AI tools and the existing research. It does not fetch questions, and it excludes dismissed questions.

What to address lists prioritized question gaps, relationship gaps, search-intent differences, and topic placement suggestions. Expand the list to read more findings.

Open Review details to read each judgment:

Answered marks a substantive answer in body text. A heading or a repeated question alone is not an answer.
Explained marks a relationship that the draft explains, as opposed to one it only mentions.
Partial marks incomplete coverage, Not found marks absent coverage, and Uncertain marks an inconclusive judgment.
Reference pairs do not establish business capabilities or causal claims.
Topic suggestions separate subjects for this page from subjects for a separate page.

Open Supporting passages and click Find passage to select the exact quotation in the draft. Coverage judgments do not verify factual accuracy or predict rankings.

Insert writing guidance adds a labeled instruction to the suggested section. When no section fits, it adds a heading with the instruction. Guidance stays out of the draft HTML and exports until you insert it. Use undo to reverse an insertion, then click Save to keep your changes.

The most recent review saves privately with its draft and project, and reopening the editor loads it without another model request. Changes to the text, intent, target, active questions, or references mark the result out of date with Review is out of date. Click Review answers again before you use outdated passages or insert guidance. Restoring the exact reviewed text makes those actions available again when the other inputs still match.

Overlapping findings in the same section are combined, and their supporting passages and writing requirements remain available. Separate-page suggestions stay distinct. A failed review shows an error instead of unanswered labels; retry when your model provider is available.

Review has a size limit on questions and draft length, and it tells you when a draft or question list is too large. Dismiss out-of-scope questions or shorten the draft to fit. Each review reports how many research items it reviewed out of those available. The limit never truncates your draft text, and the detailed panels keep the full research.

Generate an outline with writing guidance

Open Write & Review, then Outline. Every generated outline includes writing guidance.

Current draft lists the draft's headings. Click a heading to locate it in the draft.
Suggested shows generated headings with section guidance.
Competitor headings shows headings from your reference pages. Click Insert heading to add one to the draft. If a page's outline is off-topic or boilerplate, click Remove outline under that page and confirm. Its headings then stay out of this list, Generate outline, and coverage reviews for the post, while its terms and entities remain.

To combine competitor headings with your entities, active reader questions, topic groups, and a suggested length, click Generate outline. Generation follows the saved search intent and excludes references from an earlier target.

Review the suggestion before you apply it. Insert adds one section with its writing guidance. Replace draft with suggestion replaces the whole draft after confirmation. Use undo to reverse an insertion, and click Save to keep changes. Generation alone does not replace or save draft content.

Rewrite the draft

Rewrite proposes a full replacement draft that works in the facts that Fact status shows as not observed and the reader questions the sidebar shows as not observed. Rewrites aim to name entities and state their attributes instead of using vague descriptions.

Open Write & Review, then Rewrite.
Keep Default Rewrite, or select another listed rewrite style.
Click Rewrite article (or Rewrite with followed by the style name).
Read the Preview. The rewrite works from the facts already in the draft, so add any missing facts yourself.
Click Apply rewrite, and then confirm. The rewrite replaces the whole draft.
Click Save.

To undo an applied rewrite, press Control+Z (or Command+Z on macOS).

If the target or research changes after a rewrite, run Rewrite again before you apply the result.

To revise one passage, select its text, and then on the toolbar, click Rewrite selection. Review the change before you save.

Review style and wording

For wording suggestions, open Style Review, and then click Run style review. Read the proposed text and its Changelog, then click Apply rewrite and confirm to use it. Applying the style fix replaces the draft.

Meta Description also has a Title tag box. Type or paste a title, or click Use H1 to copy the first H1 in the draft. The counter shows the character count and how much of the space search results give a title is used; a title that is too short or too wide shows in red. Content Editor does not save the title. Meta Description generates a description from the draft in the editor. Click Generate, edit the text until the counter turns green and shows no "aim for" note, and then click Copy. The counter shows the character count and how much of the space search results give a description is used; a description that is too short or too wide shows in red. Content Editor does not save the description, so paste it into your page's meta tag.

Compare vocabulary diversity

To score your draft's vocabulary range, open Diversity, and then click Score my draft. Reference median MTLD shows the median score of your reference articles; with an even number of references, the median is the average of the two middle scores. The panel states whether your draft is at or above, or below, that median, and lists each reference with its word count and score. Read Lexical Diversity for what each score means.

Check sentiment

Open Write & Review, then Sentiment. Click Analyze sentiment to review the tone of the current draft.

The panel shows Entire document, each section, and each sentence. A section contains one heading or subheading and its text, ending at the next heading; text before the first heading forms its own section. Expand a section to see its sentences, and click a sentence to select it in the editor.

Measurement	What it reports
Score	Tone from −1 (negative) to +1 (positive) for the document, each section, and each sentence.
Magnitude estimate	Estimated emotional intensity. Longer documents can have larger values, and the estimate can miss mixed sentiment within one sentence.

These measurements are PirateSERP's own, not Google Natural Language scores, and the magnitude estimate is not a confidence value. Green marks positive tone, red negative tone, and amber near-neutral tone; blue badges show magnitude estimates. Each badge includes a text label.

Language coverage is experimental. You can analyze any supported project language, but sentiment accuracy is not validated for each one. Sentiment can misread factual service descriptions, negation, and mixed sentiment, so review the source text before you change it.

Editing the draft or changing its language clears the results. Click Clear sentiment to remove results without changing the draft. Sentiment tells you when a draft is too long to analyze.

Check readability

Open Write & Review, then Readability. Click Analyze to review an English draft. Readability supports English only; drafts in other languages show Readability is English-only.

To set the long-sentence guideline, select a level in the Target audience list. Expand a category, then click an example to select that passage in your draft.

Readability highlights never become part of saved content. Click Clear to remove the analysis and highlights without changing your text. Editing the draft clears the analysis, so click Analyze again after you edit. Switching tools keeps the results until the draft changes.

Choose an audience level

The Common European Framework of Reference for Languages (CEFR) describes language proficiency from A1 to C2. A1 and A2 describe basic proficiency, B1 and B2 independent proficiency, and C1 and C2 proficient language use. These levels describe readers, not writing quality.

Level	Reading ability	Sentence guideline in PirateSERP
A1, beginner	Familiar words and short, simple sentences.	10 words
A2, elementary	Short texts about everyday needs and familiar topics.	15 words
B1, intermediate	Straightforward factual texts about familiar subjects.	20 words
B2, upper intermediate	Complex texts and technical discussions in a familiar field.	25 words
C1, advanced	Long, demanding texts with meaning that is not stated directly.	30 words
C2, proficient	Dense, abstract, or structurally complex texts.	35 words

PirateSERP flags sentences longer than the selected guideline. These word limits are editorial presets, not official CEFR thresholds or a validated proficiency assessment.

B1 is the default. Choose the level of your intended audience, keep technical terms your readers need, and explain unfamiliar terms. Changing the level adjusts only the long-sentence check; your text, measured counts, and other checks stay the same. Choosing C2 does not make a draft better.

Expand What do A1 to C2 mean? in the panel to compare levels without leaving the editor. For the framework's own definitions, read the Council of Europe level descriptions.

Interpret the checks

Check	What it reports	How to use it
Word and sentence counts	Total words, total sentences, and average words per sentence.	Review the draft's length without treating length as a quality score.
Long sentences	Sentences that exceed your selected audience guideline.	Split sentences where splitting clarifies meaning.
Complex sentences	Sentences that are hard to read because they are long or use long words.	Review highlighted sentences in context.
Vocabulary to review	Words with three or more syllables, excluding proper nouns.	Keep necessary terms or explain them. This check does not assign CEFR vocabulary levels.
Possible passive constructions	Constructions identified through grammatical analysis.	Decide whether the sentence needs to name who performs the action.
Adjectives	Adjective occurrences and their percentage of total words.	Review descriptions for precision. The percentage does not measure credibility or factual accuracy.

The highlight counts also report adverbs and qualifiers. Grammatical estimates require review in context; a flagged passage does not require a rewrite.

Readability analyzes draft text, including headings. Formatting labels such as H1 and PARAGRAPH are not part of that text. Readability sets no total sentence limit and no target adjective percentage.

Check that each section answers its heading

Analyze also runs an Answer-first check on the draft, including imported pages. It lists each section whose first paragraph doesn't answer its heading, with the problem and a suggested fix. The check flags these problems:

A question heading answered with another question.
A section that opens with a contrast word, such as "However".
A heading that asks how much, how long, or how often, with no figure in the first two sentences.
A how-to section whose steps run as prose instead of a numbered list.

Click a heading in the list to select it in your draft. The check runs in every language. In languages other than English, it checks only for question headings answered with another question.

Check that the draft leads with your keyword

In English, Analyze also runs a Target keyword check when the post has a target keyword. It lists these problems:

An H1 that doesn't name the keyword.
A first sentence that doesn't name the keyword.
Sections that mostly open on a different subject than the keyword.

The section check is skipped on a how-to page, where sections open with steps for the reader instead of the topic. Content Editor recognizes a how-to page by its wording, and the panel says when it skips the section check. The H1 and first-sentence checks still run.

Click a flagged heading to select it in your draft. A keyword that isn't a name or a topic, such as "best roofers near me", is flagged even when the page is on topic. In that case, ignore the check.

Send the draft for another check

Send to Entity SEO Auditor saves the draft and opens Entity SEO Auditor with the draft selected.
Check Semantic Alignment saves the draft and opens Content Alignment with the draft selected.
In E-E-A-T Analysis, select the saved draft in the Or a saved draft list.
In Entity Graph Viewer, select My Articles to see the draft's entities as a graph.

None of these actions publishes the page. Review the final text and copy it into your website editor when it is ready.

Find saved work

Drafts belong to the active project. If an expected draft is missing from Saved drafts, check the project selected in the header.

Research you contribute through Content Editor can enter shared industry research. Submit public reference material, not confidential documents.

Read Languages for language-specific analysis limits.