Sign up and write.
The rest is the newsroom.
Not the brochure version - the whole editorial process, and what actually happens in it between asking for an article and publishing one. It is the content production process end to end, with the editorial standards applied at each step written down rather than implied. If you have been looking for content workflow software, this is the newsroom that replaces it. No site required to start. New to the category? Start with what an autonomous AI newsroom is.
I built this to run my own sites, so this page is the walkthrough I would have wanted: what you set up once, what happens to every article, and where you stay in charge.
From nothing to a published article.
- Sign up
Google, email and password, or a one-time link.
One screen and you are inside, on a free 7-day trial with 10 articles. None of the three doors asks for a card, and nothing is billed unless you later choose a plan.
- Tell it who you are
Have a website? We read it. Don't? You write anyway.
Give it a web address and we read the site, then set the newsroom up from what we found - the name, the language and country, the topics it already covers, even the CMS it runs on. You correct all of it before anything is created. Without a website there is nothing to read, so you set the language, the length, the voice and whether articles carry an image, and write from your profile from day one.
- Ask for an article
One field. Type a line, get an article.
There is no mode to pick first - the desk reads what you typed. One line is one article; add a row and it is one article per row. Whichever way in you use, it is the same nine desks and the same checks behind it.
- Read what it did
The sources, the checks, and the checks it failed.
Every article carries its own record: the sources it was written from, named, and each check that ran - including any it did not pass, marked in amber with the measurement written out. An article can arrive with a known flaw attached rather than being quietly held back or quietly passed. None of it is asserted; it is what the desks recorded.
- Publish or export
You press Publish. That press is the review.
Send it to your own CMS - WordPress.org, WordPress.com, Ghost, Webflow, Shopify, HubSpot or GoHighLevel - with its byline, its image and its metadata - or connect nothing at all and take it as Markdown or HTML. A connected site is never required to get your work out.
Seven ways to ask, one newsroom behind them.
They differ only in how the content brief is put together. After that it is one set of content operations - the same desks, the same sourcing and the same checks, whichever door you came through. There is no editorial calendar software to keep in step beside it and no separate editorial management software: the schedule, the standards and the publishing are the same newsroom. The content approval workflow is part of it rather than bolted on - review before publishing is the default and auto-publish is opt-in per site, so the editorial workflow has a stop in it unless you remove it. One newsroom workflow, not a content workflow stitched across four tools.
A topicone keyword or headline → one article
Type one line. Give it a keyword rather than a headline and the desk settles on the headline the article will be sourced to before a word is written - a real editorial decision, taken up front.
A listone article per row
Add rows and each becomes its own article. Any row can carry its own outline and the writer keeps it, sub-headings included. Worth knowing: a run of ten is ten full generations, not one cheap batch - the desks run per article.
A structure you reuseyour shape × your values → the set
Save a shape once - the sections, what each is for, roughly how long - then cross it with a table of your own values and generate the whole set. The value tables are saved separately and load into any run.
A URL you publishedrefreshed, rewritten, or beaten
Point it at a page you already have. Pages are fetched and shown to you before a generation is spent - the headline we found, the domain, the word count, the date - and one we could not read is named and left out rather than written against. You can pull the list straight from your sitemap, narrowed by path and by how long since a page was touched.
Two subjectsone comparison that picks a side
Name two things and get one comparison. This is one of the two ways in that has been run all the way to a finished article rather than verified up to dispatch.
A subject and a countone ranked list
A rundown, in the shape a ranked list wants, over the same sourcing and the same checks as everything else.
A term to explainone definition piece per term
Type one term and it is one piece; add a line and it is one article per term. There is nothing to classify first - the surface reads what you typed, the same way the front door does.
There is an eighth door that is not a shape at all: Find sources searches the web first and shows you what survived, with the reason each one ranked - then you tick the ones you want and write from those. And nothing you typed is thrown away by navigating: leaving a surface with work on it offers to save the list. See how it works
It can also run without being asked.
The nine desks run once when you ask, then stop. A domain's autonomous newsroom runs on a schedule once you finish its setup in Newsroom and turn it on.
It is a setup you finish rather than a switch you flip, and it is built around a real website - which is why the on-demand side above needs no site at all and this side does.
- A website we can readthe scan is how a newsroom is born - no hand-typed sites
- An editor of recorda named human who is accountable, because nobody reads these before they publish
- The topics you coverread off your own site, corrected by you
- At least one scheduled timewhen it should file
A CMS connection is never required, even here - drafts-only is a full mode.
Everything it does while you sleep.
ai fact checkingAI fact checking against its own sources.+
Unsourced, unchecked text is what gives AI writing its bad name. Every article here is checked against the sources it was written from, and if something could not be put right, the article says so plainly instead of showing you a green tick.
- Claims cross-checked against source material
- Format and SEO rules enforced before delivery
- Anything unresolved is named, with the fix it needs
inline citationsInline citations, and every source listed.+
Real inline citations in the body, and every source listed beside them. Anything a check could not settle is named on the article rather than quietly passed. It waits as a draft until you press Publish.
- Real inline citations in the body, and the article's own source list
- What could not be settled is named, not silently passed
- It waits as a draft until you press Publish
content briefA content brief per article, not a template.+
Unless you pin a fixed structure yourself, the brief is planned from the subject, article by article, so two pieces on one topic do not arrive built the same way. How hard each one reads is measured against the level you asked for, and the number is shown on the article.
- Sections planned per article, unless you pin a structure
- Readability measured, and a miss shown in amber rather than hidden
- No house-style tic repeated across your archive
topic researchTopic research, before you ask for it.+
Set what your publication covers, once. From then on the newsroom brings you the stories worth writing in it - your niche, your geography, your moment, already sourced. You never stare at a blank page.
- Aligned to your niche and country focus
- Domains you never want cited are never fetched
- Up to five keywords per topic, steering what it searches
brand voiceDrafts in your voice.+
Set the voice per publication - tone, reading level, point of view and audience, plus your own banned words and standing instructions. Switch CMS later and the voice stays as you set it. Every article carries a real, named byline.
- Set tone, reading level, point of view, and audience
- Your own banned words and standing instructions
- Named authors and one editor of record per publication
multilingual contentMultilingual content, in the local register.+
Write in any language, for any country. The newsroom register is tuned per country - German quotation marks, Swedish that ends on the last fact - not translated English.
- Any language, any country
- Native newsroom register, curated per market
- Single subscription covers every language
article imagesAn article illustration drawn for the story.+
Every article gets its own illustration, drawn for that piece rather than pulled from a stock library, and stored on our CDN. Ask for another one and it draws a fresh cover.
- One original illustration per article, never stock
- Drawn for that piece and stored on our CDN
- Ask for another one, and it draws a fresh cover
content automationContent automation you set once.+
Set the times and a domain's newsroom files on its own schedule, in your time zone. Turn a domain off and it stops; turn it back on and it resumes.
- Daily cadence and time-of-day under your control
- Enable or disable a domain whenever you want
- Cancel any time, data retained 30 days
cms integrationCMS integration, configured once.+
Connect WordPress.org, WordPress.com, Ghost, Webflow, Shopify, HubSpot or GoHighLevel, or choose draft-only and skip the connection entirely. Credentials are encrypted at rest and the connection is tested at setup.
- Seven CMS platforms, or draft-only
- Article body, image and metadata handled
- Credentials encrypted at rest; connection tested at setup
articles on demandOn demand content, from one brief.+
Drop in a brief, or paste your own links, and the newsroom writes, illustrates, fact-checks and ships it through the same checks as everything else.
- Write from your own brief, or just paste links
- Bring your own sources, or let it find them
- Runs every check - counts as bonus output
content apiA content API, and no lock-in.+
Read your domains, your articles and their images from your own code through the /v1 API, take any article as Markdown or HTML, or download one archive of everything the account owns, self-serve from Settings. WordPress WXR and Ghost JSON exports are still coming.
- Read it all from your own code - the v1 API is live
- Any article as Markdown or HTML, no request, no waiting
- Whole-account archive - self-serve from Settings; WordPress WXR and Ghost JSON still coming
ai disclosureBuilt for the AI disclosure rules.+
A named editor of record on every domain, a visible AI credit on every generated image, and an article nobody read before publishing carries both a machine-readable marker and a visible label. An article you publish yourself does not - your click is the review. Mastheads is a registered U.S. Copyright Office DMCA agent with a published takedown route.
- A named editor of record, accountable per domain
- Generated images always carry a visible AI credit
- Unattended publishing labels itself; your own publish is your review
- Registered U.S. Copyright Office DMCA agent, with a published takedown route
You set the direction. It does the rest.
You are never managing a tool. Tell Mastheads what your publication is about and how it should sound - then ask, and read back exactly what it did.
- Your topics, your angle - you set the direction
- Your voice, and your writers' names on every piece
- The sources you trust, and the ones to avoid
- How many sources an article has to stand on
- Whether it publishes to your site or you export it
- Need a specific story today? Just ask for it
- ◢Reads around the topic and writes it in your voice
- ◢Checks the facts against the sources it used
- ◢Draws a cover that fits the piece
- ◢Writes the metadata your CMS wants
- ◢Shows you where every story came from, and what it could not fix
We do not bypass paywalls - paywalled and aggregator domains are skipped; we target publicly accessible content and operate at low volume
We do not publish content we identify as prohibited (see Terms)
We do not retain rights in your published articles
We do not require attribution back to Mastheads on your site
How long does setup take?
+
There is nothing to set up before your first article - sign up and write. Building a newsroom around your own website takes longer, because we read the site first and you confirm what we found.
Can I just ask for a specific article?
+
Yes - that is the front door. The dashboard opens on one field. Type a topic and you get one article; add rows and you get one per row; point it at a URL you already published and it comes back refreshed, rewritten, or written to beat it. Same desks, same sources, same checks.
Do I need my own website?
+
Not to write. You can sign up with no site at all, write from your profile and export what comes back. A website is only required for the autonomous newsroom, because that is built around a real site and we have to be able to read it.
Can I review everything before it publishes?
+
Yes, and that is the default. You press Publish on an article and that press is the review. Auto-publish is opt-in per domain and belongs to the autonomous newsroom, which any account can switch on.
Which CMS do you support?
+
WordPress (self-hosted and WordPress.com), Ghost, Webflow, Shopify, HubSpot and GoHighLevel, plus a draft-only option with no site connected at all.
Ask it for one article.
Your first week is free - 10 articles, no card required. Pick your tier whenever you are ready.