Writing blog posts from your own material with AI

How the bot works

Nine steps. Collecting the material, building the brief, checking every claim against a source and moving the draft through your approval workflow are ordinary code and cost almost nothing to run; the model is only asked to write prose over material you already own.

1. It starts from your material, not from the topic. Your documentation, your past posts, the recording of your own webinar, the notes from a job you did, the spec of the thing you sell. A topic and a word count is not a source.

2. The brief comes before the draft. What the article is for, who it is for, the questions it answers, the sections, and which piece of your material backs each one. A brief you can reject is cheaper than a draft you have to rewrite.

3. Every claim names its source. This sentence came from that document, that page, that recording. The citation travels with the draft, so a reviewer can check a claim without going looking for where it might have come from.

4. What your material does not support becomes a question. A number nobody wrote down, an outcome nobody recorded, a comparison nobody made. The draft asks the author instead of filling the gap, because an invented detail in your own voice is the expensive kind of mistake.

5. Your voice comes from your own writing. The examples are your published posts, not a description of a tone. Sentence length, vocabulary and structure are held to what you have already published, and a draft that drifts off it is flagged rather than shipped.

6. Your claims rules are applied, clause by clause. The phrases you have taken out before, the comparisons you do not make, the disclosures a regulated subject needs, the words your industry will not let you use. Each one is a rule, and the rule that fired is named.

7. Nothing publishes without a named approval. The approval workflow is yours: who reads it, who signs off, what happens when somebody asks for a change. An unapproved draft is never published, and there is no configuration in which it is.

8. The calendar is a queue of briefs, not an autopilot. A schedule tells the bot when a brief is due, not that an article may go live. Nothing is written on a topic nobody chose, and nothing is published because the calendar said Tuesday.

9. It tells you which claims nobody can source. The statements your material does not back, the pages that contradict each other, the posts resting on a figure nobody can find. That list is the useful output in the first month, because it is already on your site.

What you configure

It can be configured to run in any business. This is what that actually means — the list of things that are yours, not ours.

Your material

What the bot may draft from — which documents, which past posts, which recordings — and which sources win when two of them disagree.

Your briefs

Who writes them, what a brief has to contain before a draft may start, and which article kinds you produce at all.

Your voice

The published writing it is held to, the length and structure you want, and the things you never say.

Your claims rules

Banned phrases, comparisons you do not make, required disclosures, and which subjects need a named reviewer every time.

Your approval workflow

Who reviews, who approves, what a rejection does to the draft, and which article kinds need a second reader.

Your calendar

When briefs are due, how many are in flight, and which weeks are quiet. Publishing is not on this calendar.

Your links and images

Which of your own pages and assets may be linked or used, and the fact that no image is generated and no outside source is quoted without your say.

Your reviewers

Who checks a claim in each subject, who owns a question when the material comes up short, and who may approve on their behalf.

Your release switches

Which article kinds may be drafted automatically from an approved brief, which wait for somebody to ask, and whether a draft is delivered anywhere at all. Publishing is off, and it is not a switch.

Transparent, and configurable by you

The dark band below is what the system will not do. This is what you can see and change — and it is how we build everything, not something invented for this page.

You see what it did, and why

Every factual sentence names the source behind it, every question says what the material did not cover, and every draft records the material it read, the brief it worked from, the rules that fired, the questions it raised and who approved it. A post can be checked against your own material rather than taken on trust.

Versioned prompts you can tune yourself

The material list, the brief template, the voice examples, the claims rules and the prompts behind the drafting are yours to tune. A change is a reviewable diff, replayed against posts you have already published so you can see what it would have written differently, and rolled back the same way.

A human in the loop wherever judgment actually matters

Collecting the material, building the brief, drafting and checking each claim against a source are automatic. Deciding what is worth saying, whether the advice is good, what your business is willing to claim and whether this is ready to publish are a person's — that boundary is a property of the system, not a promise about how carefully we'll use it.

Every run logged, along with what it cost

One row per draft and one per question: the material it read, the brief, every claim with its citation, the rules that fired, the questions it raised, who approved the post and where the file went — plus token usage and cost for that run.

It can be configured to run in any business. The material and the rules are yours: your sources, your voice, your claims rules, your approvers. The articles on this page are one configuration of it, not the limit of it.

Works with your tools

This drafts into the tools your team already writes in, from the material you already have. We're not selling you a content platform, and your own writing stays the record.

• A website or CMS you control
• Notion or Confluence
• Google Drive, SharePoint or Dropbox
• Microsoft 365 and Word
• Google Workspace and Docs
• Mailchimp, Klaviyo or a similar email platform
• A social media scheduler
• A spreadsheet of briefs

Works well for

• Law firms — The expertise is already written down in memos and filings, and none of it is written for a reader outside the firm.
• Accounting firms — Seasonal questions come around every year, and the answers live in advice already given to clients.
• Dental practices — The same patient questions repeat, and the answers have to stay inside what a practice may claim.
• Healthcare providers — Plain-language explanations of your own services, where a claim has to be traceable to something approved.
• Manufacturers — Specifications, application notes and installation guides already exist and read like documentation.

How a rollout works

Nothing is published on day one and publishing never turns on. The first milestone is posts you have already published, redrafted from your own material, so you can compare them paragraph by paragraph with what is live today.

1. 01Discovery

   The quote call, then a working session. We go through what you publish, which material the bot may draft from, where your voice examples come from, your claims rules and required disclosures, who reviews which subject and who approves. You get the material map, the rules list and a written scope.

2. 02Dry run

   Posts you have already published, redrafted. Nothing is delivered. You compare them with what is live and read the unsourced list — and the unsourced list is where the work is, because a claim nobody can back is already on your site.

3. 03Assisted mode

   Live on one article kind. The bot drafts from an approved brief, your reviewer checks the claims and their citations, and the post goes live the way it always has, by somebody's own hand.

4. 04Optional limited automation

   Optional, and narrower than it sounds. A short list of article kinds you name, drafted from approved briefs on your calendar, still unapproved and unpublished, with every hard stop in force and the kill switch. Plenty of teams stop at step 03, and that's a good place to stop.

What we don't claim

• No numbers we haven't earned. No traffic claim, no ranking claim, no posts-per-week figure, no hours-saved number. We have not written your articles. What a run takes and what it costs come out of the dry run, and they will be your numbers.
• No customer claims on this page. No names, no case studies, no logos.
• We make no promise about search results. What ranks is not ours to control and the rules change without notice. This page is about where a sentence on your site is allowed to come from, which is a different question from how it performs.
• A draft is not a point of view. The bot can assemble what you already know into an article. It cannot have the opinion that makes an article worth reading, and whether there is one is a judgment it does not make.
• If you have no material, this is the wrong order. Drafting from your own sources assumes your sources exist. Writing down what you know is real work, and it is worth doing whether or not anything is automated afterward.

Questions people ask

Can AI write our blog posts?

Yes, from your own material. An agent can build a brief, draft from your documentation and your past writing, name the source behind every factual sentence, turn anything your material does not support into a question for the author, and move the draft through your approval workflow. What it will not do is publish, or claim something your sources do not contain.

Where does the content come from?

Your own material: documentation, past posts, recordings you own, notes from your own work. Each claim in the draft names the document and the page behind it, and nothing is drafted from a topic alone.

Will it publish automatically?

No. Publishing is not a switch on this bot in any configuration. The approved file is handed to a person, and a draft nobody approved never reaches your site.

Will this help us rank?

We make no claim either way, and we are not going to put a number on it. Search outcomes depend on things nobody here controls; what this page promises is that every sentence can be traced to something you own.

Does it sound like us?

It is held to your own published writing rather than to a description of a tone, and a draft that drifts off it is flagged. Whether a particular draft actually sounds like you is a judgment your reviewer makes, not one the bot gets to make for you.

Do we have to say it was written with AI?

That is your decision and your industry's, not ours. The workflow records exactly what the bot drafted, which sources it used and who edited it, so whatever you decide to disclose, you can say it accurately.

Is this the same as your other writing bots?

No, and they are deliberately separate. This writes long-form articles on your own site from your own material. Listings from product data, posts on somebody else's platform, campaign emails to your list and procedures for your own staff are each a different bot with its own page.

What if our material contradicts itself?

The draft stops on the contradiction and names both sources. Deciding which one is current is a person's call, and the bot does not pick the more convenient version.

Is there a record of how a post was drafted?

Yes, per draft: the material it read, the brief it worked from, every claim with its citation, the rules that fired, the questions it raised, who approved the post and where the file went - plus the token usage and cost for that run.