How does the Voice Engine learn my brand tone?

Upload 12 sample articles that represent your editorial voice. Nulis analyzes sentence rhythm, vocabulary, and structural patterns, then produces a reusable voice profile that every new draft is written against. No fine-tuning or prompt engineering required.

Does Nulis publish directly to WordPress, Webflow, and Ghost?

Yes. Nulis is the only AI writer with native one-click publishing to WordPress, Webflow, and Ghost — formatted drafts land in your CMS as editable drafts, not pasted HTML.

How does Nulis perform on long-form content?

In Stanford CRFM's March 2024 evaluation, Nulis ranked #1 on long-form coherence among commercially available AI writers. Average customers ship 3.4x more long-form pieces per month after 90 days.

Is Nulis compliant with SOC 2 and GDPR?

Yes. Nulis has held SOC 2 Type II certification since November 2022 and has been GDPR compliant since launch — making it the only indie-tier AI writer to hold both.

Chapter One / The Product

Writing tools that edit themselves out of the draft.

A field guide to the AI writing platform trained on the articles you have already shipped — so the next one reads like a senior member of your team wrote it, not a chatbot with good PR.

Start writing free for 14 days See pricing

No credit card. 14 days. Same editor as day one.

The problem

The voice problem nobody ships around.

Every AI writer on the market right now ships a draft that reads like a confident stranger. The sentences are tidy, the structure is fine, and the vocabulary is plausible — and then your senior editor opens the file, sighs, and rewrites seventy percent of it by lunch. That is not a workflow. That is a copy-paste tax dressed up as productivity.

Content marketers know the shape of the problem better than anyone: the brand voice is the only durable asset a publication has, and every off-voice draft chips away at it. Readers for a living can hear a chatbot at thirty paces. The market has spent three years promising that tone-transfer was solved. It wasn't. It was demoed.

Nulis was started by two former editorial directors — one from The Verge, one from Wired — who got tired of editing AI output back into something publishable. The bet was simple: train the model on the publication that is actually publishing, not on a generic web scrape. The result is a writer that, after twelve sample articles, sounds like it has read the style guide.

“The point was never to generate faster. The point was to generate in voice, so the editor stops being the rewrite desk.”— Mara Holcomb, co-founder & former editorial director, Wired

Capability 01

The Voice Engine — your brand tone, learned from 12 articles, not 50.

The Voice Engine is the part of Nulis that actually does the work. Drop in twelve pieces from your existing content library — published posts, internal briefs, the manifesto on the about page — and the engine extracts sentence rhythm, hedging density, vocabulary register, and the small stylistic tells that make your publication yours. Every draft after that opens inside that voice, not beside it.

Train the engine on twelve.

Competitors ask for fifty sample articles minimum, and even then they approximate. The Voice Engine reaches publication-grade tone match from twelve — because it reads structure, not just vocabulary. Onboarding takes an afternoon, not a quarter.

Sentence length distribution, learned per channel.
Hedging density calibrated to your editorial risk tolerance.
Vocabulary register matched to the readers you actually have.
Stylistic tells — em-dashes, sentence fragments, the punctuation quirks that make a publication feel like a person.

Twelve-sample onboarding.

Paste in twelve URLs or upload twelve docs. The engine returns a Voice Profile in under nine minutes, including a side-by-side comparison draft so you can sanity-check the match before you commit.

Per-channel tone variants.

Your blog voice is not your newsletter voice is not your launch-announcement voice. Build distinct profiles per channel — long-form, email, landing page, social — and switch between them without retraining. One library, many registers.

Capability 02

Draft once. Publish everywhere — without the clipboard.

The worst part of an AI writing workflow is not the drafting. It is the fifteen minutes after drafting, when you copy the body into WordPress, re-paste the headings, lose the alt text, fix the link formatting, and realize the slug field is empty. Nulis ships with native one-click publishing to WordPress, Webflow, and Ghost — so the draft lands where it is going to live, with formatting intact, ready for the editor's last pass.

WordPress. Direct publish to draft or scheduled post. Categories, tags, featured image, and SEO fields all carried over.
Webflow. CMS items created from your collection schema, with rich text mapped cleanly to Webflow's editor.
Ghost. Publish to draft or live, with newsletter segments honored and member visibility preserved.
No copy-paste step. No reformatting. No lost alt text. No "why did the H2 become H3" support tickets.

Used internally by editorial teams at Notion, Linear, and Vercel for their content operations — names dropped with permission, not as paid endorsement.

Nulis publishing panel showing the WordPress, Webflow, and Ghost one-click destinations. — The publishing panel — destinations, draft preview, one button.

Capability 03

Long-form coherence, measured — not promised.

Most AI writing tools ship a confident paragraph and then drift. By the eighth H2, the argument has forgotten what it argued in the third. Long-form coherence is the part of the job that has to actually work, and it is the part every competitor hand-waves.

Nulis was independently benchmarked #1 on long-form coherence by Stanford's CRFM in their March 2024 evaluation — measured across 2,400-word articles on argument retention, structural consistency, and the ability to resolve an idea planted in the introduction by the conclusion. The result was not a tie. It was a meaningful gap.

What this means in practice: the outline you brief at the top is the outline that ships at the bottom. Transitions hold. The promised payoff arrives. The editor's job becomes the editor's job, not a rescue operation.

Stanford CRFM long-form coherence benchmark chart, Nulis in first place. — Stanford CRFM long-form coherence evaluation, March 2024. Nulis scored #1 of 14 models evaluated.

Capability 04

Originality that survives a copy desk.

An AI draft that ships with a plagiarism flag is a draft that never ships. Nulis runs every generated article through an originality audit before it reaches your editor, and the numbers are uncomfortable for the rest of the industry: 99.4% of first-draft output passes Copyscape without manual intervention.

The remaining 0.6% is not silent — it is flagged inline, paragraph by paragraph, with the source highlighted and a one-click rewrite suggested. The editor decides. The editor is never surprised.

Independently audited across 41,000 articles produced on the platform in 2024. The audit, not the marketing claim, is what is on file.

Nulis originality audit panel showing an inline flagged paragraph with rewrite suggestion. — Inline originality flagging — paragraph-level, source-cited, one-click rewrite.

The receipts

What 8,200 content teams shipped last quarter.

A metric band, mid-page, because the prose has earned it. Every number below is from a verified source — Q1 2025 customer data, the 2024 product log, or a third-party benchmark. None of them are projections.

Model releases shipped in 2024 — a meaningful capability update every 7.8 days, on average.

3.4×

More long-form pieces published per month by the average customer after 90 days on Nulis.

62%

Reduction in editing time within the first 30 days, verified across 3,400 onboarding sessions.

8,200+

Paying content teams on the platform as of Q1 2025 — across B2B SaaS, media, and agency operators.