How to use Topical Map

A first project takes you from a blank workspace to a set of approved pages. The Overview screen always shows where you are and what to do next.

1. 1Create an account and open the app.
2. 2In Settings, add an AI provider key for writing and, if you have one, a search-data provider key.
3. 3Start a new project: enter the brand, services, cities, languages and site model.
4. 4Upload a few brand documents, such as a price list and warranty terms.
5. 5Run topical research, group the results and build the topical map.
6. 6Generate briefs and drafts, review them and approve the pages you want.
7. 7Publish approved pages to WordPress, or export them as files.

The brand foundation holds who you are: name, services, service areas, contact details, tone of voice and the terms you always or never use. Every later step reads from it.

The knowledge base holds your documents: price lists, warranty terms, service descriptions, FAQs. They are split into passages so a page can quote them. A price, warranty or response time appears on a page only when a document states it; otherwise the page leaves it out and you get a to-do.

Tip: Upload the documents you would give a new salesperson. They are what keeps the pages specific and accurate.

Personas describe the people you serve: their role, what worries them, the objections they raise and the words they use. Research uses them to find questions, and briefs use them to decide what each page must answer.

• A homeowner in Riyadh comparing AC maintenance contracts before summer.
• A facilities manager in Dubai who needs a supplier with fast response times.
• A patient in Cairo who wants to know what a dental implant costs and how long it takes.

The site model decides how service and location pages are planned. Choose the one that matches how you sell:

• One service in one city: a single focused site.
• One service in many cities: a page per city, written to avoid thin near-duplicates.
• Many services in one city: a hub per service.
• Many services in many cities: services and cities combined only where real demand exists.
• Agency: hubs for services, industries, locations and case studies, with a guard against thin combination pages.

One research run collects the topics people search for around your brand, and shows where each one came from.

• Google and YouTube autocomplete, including question and comparison modifiers in Arabic and English.
• People Also Ask questions, followed up to five levels deep (needs a search-data provider).
• Dialect and spelling variants, for example «مكيف» and «تكييف».
• Questions drawn from your personas.
• AI search gaps: topics that AI assistants recommend but your plan does not cover yet. You choose which assistants to ask.
• Your own keyword lists, imported from Excel or CSV. Volumes are kept exactly as supplied and stay blank when no source gave one.

Tip: Start with autocomplete, which is free, to see the shape of the topic before you spend provider credit on People Also Ask.

Two searches belong on the same page when Google shows mostly the same top-10 results for both. Clustering groups your keywords that way, so each page targets one intent and your pages do not compete with each other.

The similarity slider runs from 20% to 80%. A lower value groups more keywords together; a higher value splits them into more pages. Regrouping saved results costs nothing, so try a few values. The same threshold decides which pages the map merges (default 30%).

The map arranges your pages in four levels: a pillar for the main topic, hubs for each service or major subtopic, branches for narrower groups, and articles that answer one question each. Each page shows its journey stage and its top-10 search results.

Internal links follow the hierarchy: each page links up to its parent and across to close siblings. Edits you make are matched by URL, so they survive when you rebuild the map. A build summary lists what was added, merged or kept.

A brief tells the writer exactly what a page must do before anything is written. Each one contains:

• The main question the page answers and ranked title options.
• Headings drawn from real People Also Ask questions, and an FAQ section.
• The facts to use from your knowledge base.
• What the page should add that the current top results do not.
• The internal links the page must carry.

Drafts are written from the brief and your brand facts by the AI provider you connected. Arabic pages are written as Arabic, with their own slug, headings and questions, not translated from the English page.

You can pin a different model to each task in Settings, for example one model for research and another for writing.

Every page moves from Draft to In review to Approved. Only the approved version can be exported or published. Any edit after approval sends the page back to Draft, so what ships is always what someone approved.

23 automatic checks run on every page before it can be exported. They look at the heading structure, the direct answer under each heading, facts against your documents, internal links, repeated wording and structured data.

A failing check blocks export and tells you what to fix. Warnings are shown but do not block.

Connect a WordPress site with an application password. Approved pages can go out as drafts, as published posts or on a schedule, in the order the planner set. Pillar, hub and branch pages publish as WordPress pages, and articles as posts.

You can also export files: the topical map as CSV, pages as Markdown, a sitemap, the internal link map, the check results, or everything in one ZIP.

Tip: Send the first batch as WordPress drafts so you can check the layout in your theme before anything goes live.

Search data comes from DataForSEO, Serper or SerpApi, or from a free built-in tool that covers Google and YouTube autocomplete only. Writing uses your own OpenAI, Anthropic, Google or OpenAI-compatible key. Each provider bills you directly.

Keys are encrypted at rest and never sent back to your browser. Larger tasks show an estimated credit cost before they run, and every task is recorded in a usage log. During early access, credits are not charged.