The brief
Most AI writers solve the wrong half of the problem.
They produce volume, not articles you would publish under your own name. Publizo had to fix that without turning into a thirty-step wizard.
Generic output that reads like every other tool. Site owners rewrote most of it before publishing.
Solo site owners, niche site operators and small editorial teams who care about traffic and their reputation.
Editor first, not prompt first. If the first article looks publishable, people stay. If it looks generic, they leave within a minute.
What I committed to: a free tier that delivers a finished, publishable article on the first try; a quality gate that holds weaker drafts back instead of pushing them live; a visual system that doesn't look like a SaaS template; and pricing you can understand on one screen.
As a team of one, every decision had to remove future work, not add it. That ruled out custom illustration, mascots and a marketing site that needs weekly upkeep. Four rules guided the rest:
- Quiet over loud. The product is the proof.
- One clear path before three clever ones.
- Words on screen are part of the design.
- Never ship a feature that needs a tooltip to make sense.
Research
I listened to people complain in their own words.
Instead of formal interviews, I spent two weeks in SEO forums, Reddit threads and YouTube comment sections. Two kinds of users came out of it.
The Operator
Runs three to ten niche sites. Wants speed, but has been burned by content that sank their rankings.
The Editor
Writes well and hates the busywork. Wants a strong first draft, full control and no plugin clutter.
People don't want more articles. They want fewer articles they have to apologise for.
That line shaped every decision after it, from the quality gate to the free tier.
Information architecture
Four jobs. Everything else lives behind a setting or doesn't exist.
Free and paid plans share the same shell, so nobody relearns the product when they upgrade.
Generate
The desk: topic, source, options, publish. Single articles, or bulk from the Growth plan up.
Sites
Connect WordPress, Webflow, Wix, Squarespace or a custom CMS, with a health check per site.
History
Published articles and drafts, with status, quality score and a retry option.
Settings
Voice presets, billing, plan and API keys. One click away, never buried.
The topic field is the template
Templates suggest you need to learn the tool, and Publizo promises the opposite: name a topic, get a finished article. Most competitors also group articles into projects. I tested that layer and it added clicks without adding meaning, because a site already carries the project context.
Visual system
Software made by someone who cares about the written word.
The palette is deliberately restrained: warm paper, deep ink and a single indigo accent. Paired with Fraunces and Inter, it reads like an editorial tool, not a dashboard.
Typography
Fraunces at 600 with tight tracking for headlines and editorial moments. Inter at 400 and 500 for everything else, so hierarchy stays calm.
Layout
A twelve-column grid capped at 1120px with 24px gutters. Borders only where something needs a frame, one card elevation, warm shadows.
Motion
One easing curve, cubic-bezier(0.22, 1, 0.36, 1). Page transitions under 240ms. Nothing bounces or travels further than it needs to.
Key flows
Show the value before asking for the work.
Two flows decide whether someone stays: onboarding and the first published article. I rebuilt both around that one principle.
- Step 1Sign upGoogle or email. No marketing modal, no tour.
- Step 2Try a demoA sample article with no site connected and no credits used.
- Step 3Connect a siteWordPress, Webflow, Wix, Squarespace or a custom CMS.
- Step 4PublishThe quality gate runs. The article goes live or lands as a draft.
What changed
New users now see a finished sample article before any setup, and the path to a first published article fits on a single screen. The funnel data behind these changes is on the problems I found and fixed page.
Editorial system
The real design problem was making AI not read like AI.
Not the UI, not the dashboard: whether the article that lands on someone's site reads like a person wrote it. Most of this work was prompt design, evaluation design and house style rules.
- Em dashes everywhere, colon titles and "in conclusion" endings
- Generic intros with no point of view
- The same listicle structure for every topic
- Articles published live whatever their quality score
- Banned punctuation and phrasing handled at the model layer
- Opinionated leads that take a side in the first 80 words
- Structure decided per topic, not per template
- A quality gate that sends weaker drafts to draft status
Two retries, then draft
When a piece scores below the publish threshold, the system retries quietly with a stricter prompt. Only after a second miss does it land as a draft, which protects the user's credits and their reputation at the same time.
Voice presets. Each site picks a voice. News reads like a wire desk, astrology like a thoughtful column, commercial topics like a careful buyer's guide. Same engine, completely different rhythm.
Pricing
Pricing is a design surface too.
It usually gets the least design attention of any page. I gave it the same type, grid and restraint as the rest of the app, and no comparison table that needs a magnifying glass.
Free
Five articles a month, for good. Enough to feel the product, not to run a site on.
Creator
A monthly plan for solo writers who need steady output.
Growth and Pro
For operators with several sites: image fetching, more credits and automation.
Founding Member
One payment, lifetime access, capped at 20 seats. Scarcity that is real.
Paid tiers list what they include, never what cheaper tiers lack. Negative selling makes a pricing page feel hostile.
Intent survives sign-up
Choosing a paid plan now carries through account creation and lands on checkout, instead of dropping people on the dashboard to find billing on their own.
Reflection
Why it exists, and what I would change.
Between design projects, products and research work, my own blogs kept slipping. Writing, optimising, formatting and publishing every article took more time than I had. Publizo automates that busywork so the time goes into content people want to read.
What I would do differently: ship the quality gate on day one. Most early complaints were about articles that should never have been published. The fix was the gate, not better copy on the marketing page.
Style packs per niche
House rules that ship with each vertical.
A light CMS layer
A minimal publishing surface for people without WordPress or Webflow.
A public quality benchmark
One that doesn't flatter the tool publishing it.