From Blank Page to Published Authority: How to Get the Most Out of ATLAS

Author: Roumi Gop, CEO & Co-founder, Kretell Published: December 14, 2025

Two Users, Two Outcomes
Before You Start: What ATLAS Needs From You
The Free Phase: What You Get Before You Pay Anything
Step 1: Choose Your Format Deliberately
Step 2: Write a Real Brief — This Is Your Most Important Decision
Step 3: Read the Topic Intelligence — Don't Skip It
Step 4: Source Approval — Your Quality Gate
Step 5: Building Section by Section
Step 6: The Voice Consistency Check
Step 7: The Research Confidence Score and Export
Common Mistakes and How to Avoid Them
Frequently Asked Questions

Two Users, Two Outcomes

Two professionals opened ATLAS on the same day to write white papers on the same broad topic: digital transformation in financial services.

The first user spent four minutes on the brief. "White paper on digital transformation in banking." Clicked through the topic intelligence screen without reading it. Approved all sources at once. Hit "Write for Me" on every section without reviewing the research panel.

The second user spent twelve minutes on the brief. Described her specific audience — CFOs at mid-market West African banks — her thesis, her intended publication destination, and the evidence standard she needed. Read the topic intelligence carefully and pivoted her angle when ATLAS showed her an undersaturated territory. Rejected two sources she didn't recognise. Used "Polish My Draft" for the section where she had proprietary client data no AI could know.

Both received complete documents at the end.

One was generic enough that its author could have been anyone. The other was so specifically hers — her voice, her sources, her argument — that it was already shortlisted for an industry publication before it was formally submitted.

ATLAS is not a vending machine. It is a structured, collaborative workflow. What you put in at each decision point is exactly what determines what comes out.

Before You Start: What ATLAS Needs From You

Before you open your first ATLAS project, understand two things.

First: your voice profile matters. ATLAS writes in your voice using the same 100-marker system that powers your Kretell LinkedIn post generation. The more complete your voice profile, the more accurately ATLAS will reflect how you actually write. If you have not completed the Kretell voice profile wizard or generated a significant number of posts, do that first. You can start an ATLAS project at any completeness level — but a richer profile produces a more distinctively yours document.

Second: ATLAS rewards deliberate users. Every decision gate in the workflow is an opportunity to shape the final document. Users who engage seriously at each gate produce documents they are proud to publish. Users who click through without reading produce documents that require extensive revision. The difference in time is rarely more than twenty minutes. The difference in output quality is significant.

The Free Phase: What You Get Before You Pay Anything

ATLAS does not ask you to pay before you have seen what it will build.

The first five phases of every ATLAS project are free. Format selection. Brief entry. Topic intelligence — where ATLAS researches the landscape and shows you which angles are saturated and which have white space. Source brief — where the oracle introduces the research universe it has assembled specifically for your professional context. And the full Table of Contents — your complete document structure, section by section, with rationale for each.

You see what the document will be before you commit to building it. You see the sources before a single word is written.

The paywall activates at Phase 6: Build. If you choose not to continue, your project stays in the sidebar as a dormant Atlas — visible every time you log in, ready to resume if you return.

The free phase is not a teaser. It is substantive. Many users refine their angle, change their format, or redirect their brief entirely based on what they learn in the topic intelligence step. That is the system working as intended.

Step 1: Choose Your Format Deliberately

Format selection is the first decision you make in ATLAS and it cannot be changed after the project starts. This is not an arbitrary restriction — the format shapes everything that follows. Source selection. Section structure. Word targets. Citation style. Whether a Story Bible is activated.

Twelve formats are available across three categories:

Non-Fiction (Research Mode): Blog Post (800–1,500 words), News Article (500–1,200), Published Article (1,500–3,000), Op-Ed / Opinion Column (600–1,000), Academic Paper (2,000–8,000), Technical Documentation (1,000–5,000), White Paper (3,000–8,000).

Fiction and Creative (Fiction Mode): Screenplay, Children's Book, Novel.

Flexible: Mini Book (8,000–20,000 words), Book (20,000–60,000).

How to choose correctly

Ask yourself two questions: What is the destination? And what is the required evidence standard?

If you are writing for a trade journal, the destination is clear — Published Article. If you are writing definitive industry analysis for a senior executive audience, that is a White Paper. If you have a provocative argument to make without needing extensive citations, that is an Op-Ed.

The most common mistake is choosing Blog Post because it feels lower stakes. A Blog Post in ATLAS produces 800–1,500 words with proportionate research depth. If you actually need 4,000 words and a formal bibliography, you need a White Paper. Start with the right format.

Step 2: Write a Real Brief — This Is Your Most Important Decision

The brief is the most leveraged input in ATLAS. Everything downstream — source selection, topic angle analysis, table of contents structure — flows from it.

The brief is a plain language description of what you want to write. No forms. No dropdowns. No structured fields. You describe it the way you would explain it to a knowledgeable colleague.

The difference a real brief makes

| Weak Brief | Strong Brief | |-----------|-------------| | "Something about leadership" | "An op-ed arguing that traditional command-and-control leadership models are actively destroying retention in Indian tech companies post-pandemic, aimed at CHROs at 500+ person companies, with supporting data from 2020–2025" | | "AI in business" | "A white paper for mid-market West African banks exploring AI-driven risk modelling ROI — primarily skeptical audience, needs case study evidence and regional financial data" | | "Health article" | "A published article for a trade journal on Ayurvedic integration into corporate wellness programmes in Indian multinationals, peer-reviewed tone, citations required" |

The difference in output quality between a weak and strong brief is not incremental. It is the difference between a generic document that could have been written by anyone and a specific, targeted piece of work that reflects genuine expertise.

A strong brief answers:

What is the specific argument or thesis?
Who is the audience?
What is the evidence standard required?
What is the publication destination?
Any constraints on tone, length, or subject matter?

Step 3: Read the Topic Intelligence — Don't Skip It

After you submit your brief, ATLAS researches the topic landscape before building anything. This step produces two things: a map of the territory and a first preview of research sources.

Oversaturated angles are the territory everyone is already covering. Your document will compete against many existing pieces if you publish here. It is not impossible — if you have a genuinely differentiated perspective, saturated territory can still work. But ATLAS is telling you something when it flags an angle as saturated.

White space angles are underserved territory. High-quality content here faces less competition and tends to attract more durable attention from specialist readers who have not seen the topic addressed from this direction.

Source cards give you a first taste of the research system. You will see a handful of sources the oracle has already found, with credibility tier badges. This preview lets you sense-check whether the research universe being assembled is appropriate before you commit to a direction.

How to use topic intelligence well

Do not treat your original brief as locked. The intelligence is showing you what the landscape looks like. You can confirm your direction, pivot to white space, or refine the angle based on what sources exist.

The oracle found your sources based on who you are. If the sources in the preview do not match the conversation you want to have, that is information worth acting on before you build a table of contents.

Step 4: Source Approval — Your Quality Gate

Source approval is the most consequential decision in the paid phase of ATLAS. It is not a formality.

When you approve a source, you are telling ATLAS: use this as factual foundation for the section it supports. When you reject a source, it is removed from context entirely — it cannot influence a single sentence of the generated text.

Understanding the credibility tiers

| Tier | Badge | Meaning | |------|-------|---------| | Verified | 🟢 | Direct API source with confirmed URL and live data | | Reliable | 🟡 | Credible publication retrieved via Perplexity with citation | | Referenced | 🟠 | URL available but date or domain uncertain | | Unverified | 🔴 | Claude knowledge only — no URL, no confirmed source |

Practical rules for source approval

Reject any 🔴 source unless you have a specific reason to include it. Unverified sources have no URL — you cannot check them. If the claim they support turns out to be wrong, you have no way of knowing before publishing.

Do not approve sources from domains you do not recognise without opening the URL first. Source cards show you the link. Open it. Confirm it is what it claims to be.

Prefer 🟢 and 🟡 sources throughout. Your Research Confidence Score at export is a weighted average of the tiers of every approved source. A document built on verified and reliable sources scores high. A document with significant unverified sources will score low — and you will see that clearly before you export.

Step 5: Building Section by Section

The build phase is the core of ATLAS. For each section, you move through four micro-steps.

Research Panel

Source cards for this specific section appear. Review them. Approve the ones you trust and find relevant. Reject the ones you do not.

After approval, ATLAS is ready to build.

Write for Me

ATLAS generates the section using only your approved sources, in your voice, following the section's word count target and the format's citation style. Every factual claim is grounded in an approved source card. Every section begins from your voice profile — the same system that makes your LinkedIn posts sound like you, applied at document scale.

Polish My Draft

This is the most powerful mode in ATLAS — and the most underused.

If you have proprietary data, personal case studies, or specific arguments that only you could know, write those yourself first. Paste your draft into Polish My Draft. ATLAS rewrites it to match your voice profile and weaves in citations from the approved sources, without altering the substance of what you actually wrote.

The result is the most authentically yours content ATLAS can produce. You bring the expertise and the evidence that no AI can access. ATLAS brings the voice consistency and the research integration. The combination is the closest thing to how serious professional writing has always worked — an author with expertise, supported by a research team.

If you have genuine expertise to contribute to any section, use Polish My Draft for that section instead of Write for Me.

Review and Approve

Read the draft. You can request specific revisions — "make this shorter," "the second paragraph is too formal," "add more about the India context" — or approve it as written. Approval writes the section to the database immediately and advances to the next.

Managing multi-session work

ATLAS is built for documents that take days or weeks. If you step away, ATLAS generates a session summary capturing the last decisions you made and a single clear next action. When you return, you see exactly where you left off.

At the end of each working session, you also receive a briefing email — a concise record of what was built, what was decided, and what comes next. For white papers and books that may span multiple weeks, this email is your re-entry point. It means you never return to a project wondering where you were.

Step 6: The Voice Consistency Check

When every section is approved, ATLAS assembles the full document and runs a voice consistency check before you export.

This step exists because long documents generated across multiple sessions can drift. A section written in one context can feel like a different person wrote it than a section written three days later. The consistency check flags sections where the voice score has dropped — patterns inconsistent with your profile detected by the system.

Do not skip this. A white paper that shifts register halfway through is one of the most detectable signs of AI-assisted writing. Address flagged sections before exporting. This is the step that makes the difference between a document that reads as professionally authored and one that reads as obviously assembled.

Step 7: The Research Confidence Score and Export

Before export, you see the Research Confidence Score — a weighted average of the credibility tiers of all sources used throughout the document.

A high score means your document is built on verified, reliable research. You can publish with confidence.

A low score means significant portions relied on unverified or knowledge-only sources. You need to verify those claims independently before publishing under your name.

Export options:

PDF — formatted document with title, metadata line, all sections, references, and visual credits
Clipboard — clean plain text, formatted for pasting into any platform

Common Mistakes and How to Avoid Them

| Mistake | Why It Happens | How to Avoid It | |---------|---------------|----------------| | Weak brief | Feels faster | Spend ten minutes on the brief. It saves hours downstream. | | Ignoring topic intelligence | Seems like an extra step | Read it. It often changes what you should build. | | Approving all sources without checking | Feels like a formality | Reject every 🔴 source by default. Open URLs you do not recognise. | | Using Write for Me for every section | Easier | Use Polish My Draft for any section where you have expertise or data only you can provide. | | Skipping voice consistency check | Impatience at the finish line | Run it every time. Always. | | Wrong format choice | Feels low-stakes at the start | Choose format based on destination and evidence standard, not word count preference. |

Frequently Asked Questions

How long does a typical ATLAS project take?

It depends on the format and your pace. A Blog Post might take one focused session of 45–60 minutes. A White Paper typically takes multiple sessions across several days. A Novel is designed for weeks of sustained multi-session work. ATLAS saves everything immediately — there is no pressure to finish in one sitting.

Can I change my table of contents after I start building?

You can edit section titles and add or remove sections before you approve the TOC. Once building begins, structural changes require starting a new project. This is why engaging with the topic intelligence before committing to a direction matters.

What if I disagree with the sources ATLAS selected?

Reject them. Source rejection removes the source entirely from the context for that section. If the source universe feels systematically wrong for your topic, that is usually a signal that the brief needs to be more specific — the source resolver uses your brief and professional identity to assemble sources, so sharper briefs produce better-matched sources.

Can I use ATLAS for a topic outside my professional background?

Yes. The oracle personalisation is based on your professional identity, but it does not restrict the topics you can write about. If you write outside your professional domain, sources may be drawn from a broader pool rather than domain-specific databases. The voice system applies regardless of topic.

Is there a word limit per project?

Word targets are governed by the format you select. The system handles the full range — from a 500-word News Article to a 100,000-word Novel. There is no artificial cap below the format's natural ceiling.

Related Articles:

Last Updated: February 2026 Word Count: ~2,100 Reading Time: 8 minutes