Blog automation
SEO content automation that stops when it should.
The repetitive half of content — the planning, the metadata, the remembering to publish — runs on a schedule. The half that decides what a business says stays with you, because that is the half a pipeline cannot be trusted with.
Plan, research, write, publishGates before anything goes liveApproval mode when you want it
What is SEO content automation?
Worth being precise about, because most of what is sold under the name is a text generator with a schedule bolted on.
The short version
- SEO content automation is the repeatable half of content work — choosing queries, researching them, writing to a brief, producing images and metadata, publishing on a date, reporting the result — running on a schedule rather than on someone’s memory.
- It is not unattended publishing. Every post passes checks, and an approval mode can hold each one until you release it.
- Xt4b publishes to a blog it hosts on your own domain, on the days you choose, per site and per language, and reads Search Console back per site.
- It costs 20€ per website, per language, per month. Nothing is charged per article.
SEO content automation is the repeatable part of content work running on a schedule: choosing which queries to go after, researching them, writing to a brief, producing the images and metadata, publishing on a date, and reporting what the posts earned. The judgement — what the business should say, and whether a draft is good enough — is the part that stays with a person.
Xt4b is that pipeline: it plans per site and per language, writes each article against an approved entry, publishes to a blog it hosts on your own domain, and reads Search Console back per site. It is not a prompt box that returns a draft and leaves the planning, the internal links and the publishing to you.
Automation is only safe when three conditions hold. If any one of them is missing, the result is either content you do not recognise or a calendar that quietly fills with things nobody would publish.
See what one planned entry contains before it is written- A plan you can read and discard from before a word is written, so a wrong assumption costs an entry rather than a published page.
- Checks between the draft and your domain: numbers it can source, links that resolve against your sitemap, exactly one H1, and a review that can reject the post.
- A destination you own — your domain, your sitemap, your canonical URLs — so thirty articles become an asset rather than thirty pieces of borrowed traffic.
Where the time actually goes.
Content work is mostly the same six tasks repeated, whatever the niche. That is what makes it worth automating — and what makes the strategy half so valuable once it is automated.
- Choosing the query
- Deciding what a page should rank for is the same decision every week, and it is answerable: your sitemap and headings show what you already cover, so the gap is arithmetic.
- Reading the results
- Opening the top ten, noting what they cover and what they miss, stealing nothing but the shape of the answer. Repetitive, and it is where the specifics come from.
- Writing to a brief
- Once the query, intent, outline and word count are fixed, the writing is execution. This is the one part a model does well enough to be useful; it is not the part that decides whether the page is worth publishing.
- The metadata nobody enjoys
- Meta title, description, slug, image alt text, internal links, the canonical URL. Twenty minutes of tedium per post, and the difference between a page that gets indexed properly and one that does not.
- The images
- A cover and an in-article graphic, on brand, sized and named correctly — repeated for every post, by hand, forever, in most teams.
- Remembering to publish
- The most expensive small task on the list. A finished draft that sits in a folder for three weeks has cost the same as a published one and returned nothing.
The loop, in the order it runs.
Seven steps, and only one of them needs a person. The rest is the part of content work that is identical every week, which is exactly the part worth automating.
- 01
Plan the next few openings
The planner reads the site — pages, headings, existing URLs — and dates the next three publishing days. Not a year of filler: a short horizon that refills as posts go out.
- 02
Research the query, for that market
A live SERP for the post's keyword and market, plus the top results' coverage and excerpts from the pages worth quoting. Facts come from somewhere real or they do not ship.
- 03
Write against the brief
The draft follows the plan entry: query, intent, format, outline, word count, metadata. Nothing is reinterpreted between approval and writing, which is the whole reason a plan exists.
- 04
Produce the parts nobody enjoys
Cover and inline image, meta title and description, internal links from your real sitemap, the conversion link from your own call-to-action list.
- 05
Run the gates
Unsupported numbers are deleted, a second H1 is dropped, linkless sources are removed, and the final review asks whether the piece is concrete. Thin work goes back — or does not publish.
- 06
Publish on your own domain
Live on your blog on the date, with canonical URLs, structured data and the sitemap updated in the same run. A failed attempt is retried rather than lost.
- 07
Report, then re-aim
Search Console is read-only per site, so the queries that earned impressions are visible — and the next plan is aimed at the gaps rather than the same topics again.
A pipeline with gates is not a button with a prompt.
Both produce text. What differs is what happens between the text and your domain.
A prompt, run by hand
- You remember to do it, and you remember the metadata afterwards
- One generic article, with no idea which queries your site already answers
- Numbers appear that nobody can source, and nothing catches them
- Links point at pages that may not exist
- Publishing is a separate afternoon of copy-paste
- Nothing tells you whether any of it worked
A planning pipeline
- It runs on your publishing days, with dates already assigned
- Each post answers a query the site does not cover yet, in the order that matters
- Every figure is checked against the research or your own pages, or deleted
- Internal links are resolved against your real sitemap
- The blog, the canonical URL, the schema and the sitemap update in the same run
- Search Console sits beside the calendar, per site
The calendar is the control panel.
Automation without a view of it is a black box. Every decision the pipeline makes is visible as a dated entry you can act on before or after the fact.
xt4b.com
EN · DE · FR · ES · PTNext publish in 2 daysPlanned
12
Live
07
Languages
3
Publishing calendar
September
- Mon 09How long does SEO take to show results?EN · 1,840 wordsPublished
- Wed 11How much does SEO content cost per article?EN · 1,840 words · publishing nextWriting
- Fri 13Do AI-written posts rank in 2026?EN · 1,840 wordsQueued
- Mon 16SEO-Content auslagern: Kosten und AblaufDE · 1,840 wordsQueued
Planned, writing, live, waiting for approval, failed review — the states are on the entry, not buried in a log. A post that failed its review is shown in red rather than disappearing, because a silent failure is worse than a rejected draft.
Moving a date or discarding an entry changes the calendar without touching an article. Discarding is the cheap operation in a planned pipeline: you throw away a plan entry, not a finished piece of writing.
When something upstream breaks — the site briefly unreachable, the image model refusing a request — the job is retried on a backoff rather than dropped, so a transient failure does not leave a hole in your calendar.
Where it stops and asks you.
The interesting part of automation is not what it does alone. It is what it refuses to do alone.
Every one of these is a switch in the dashboard, not a support ticket. The default is the one that publishes on schedule — the others exist because a business changes its mind, and a pipeline that cannot be interrupted is one nobody wants to leave running.
- Discard a planned post before it is written, and nothing is wasted — the entry, not an article, is what you throw away.
- Turn on approval and every new post waits as Needs approval until you press Approve & publish.
- Set the days you publish on, per site: a Monday–Wednesday–Friday site gets those dates and no others.
- Write the list of pages you want readers sent to. Each post links the one that fits, and we never invent a destination.
- Choose the shape of the article per site: table of contents, sources, FAQ — on, off, or on only where it belongs.
- Request a post in your own words, with an audience, an intent and a goal, and it is dated today instead of waiting for the planner.
- Publish a planned post early, discard one you do not want, unpublish a live one, or take a language off the calendar without touching the others.
What it refuses to publish.
These are the failure modes that make automated content embarrassing. Each one has a check behind it, and the check runs on every post rather than on a sample.
- A figure it cannot support
- A sentence stating a market rate, a percentage or a price is checked against the research and your own pages. If the number is not in the material, the sentence is deleted rather than softened.
- A second H1
- The article has exactly one top-level heading. The renderer, the JSON-LD and the table of contents all assume it, so the gate enforces it instead of trusting the writer.
- A link to a page that does not exist
- Internal links are resolved against your sitemap. A target that is not a real page is dropped rather than guessed — a 404 in a published article costs more than a missing link.
- Thin advice
- “Be consistent, measure it, publish regularly” is not content. A piece that reads as advice instead of material is sent back for repair, and discarded if it stays that way.
- A cover in the wrong style
- If you chose photorealistic covers and the image generation fails, the post waits and retries. It never publishes with the abstract gradient as a substitute, because that is not what you asked for.
- A post on a day you do not publish
- Dates come from the site's publishing days, not from whenever the pipeline happened to run. With every day selected you get the old behaviour; with one day a week you get that day.
What it cannot do.
Worth reading before the FAQ, because the limits define what you are buying.
The honest boundaries
It cannot promise rankings, traffic or revenue — nobody can, and the terms of service say as much. It cannot know your positioning, your margins or which client is worth annoying: that judgement is what the approval step is for.
It cannot write from nothing. Research comes from the live SERP and from your own pages, and if neither yields anything specific the article gets thinner rather than more confident.
It cannot run on a site it cannot reach. And photorealistic covers depend on an image model: if that model fails, the post waits and retries rather than shipping with the wrong kind of image.
Blog automation: the rhythm is the hard part.
Most business blogs do not fail on quality. They fail on cadence — a burst in the first month, then silence until somebody remembers. Automating a blog is mostly about the calendar, and about the parts nobody enjoys doing twice.
- The month-three problem
- Six posts in the first month, then silence until somebody remembers. The failure is rarely the writing — it is a schedule nobody is holding, and a date that survives a busy week is most of the fix.
- The topic list that never becomes a schedule
- Ideas in a document are wishes. An entry here is committed to a publishing day before it is written, so the calendar fills itself and the writing follows the calendar rather than the other way round.
- The second market
- A second language normally means a second project, a second brief and a second budget conversation. Each market here is its own plan, its own dates and its own searches — one more line, not one more campaign.
- The blog nobody briefs
- A freelancer needs a brief, a chase and a review pass. The plan is the brief, so nothing is briefed twice and nothing is lost between months.
What stays human when a blog runs itself
- Whether the business is willing to make the claim the post makes
- The price, the availability and the policy a commercial post restates
- Which markets are worth serving this quarter
- Whether an entry belongs in the calendar at all
- The one answer only the operator knows
More on content
- SEO content audit
How to find what your site is missing, what to pull from Search Console, and which of the four findings matter — plus the part of the job worth automating.
- Publishing more posts
Four ways to raise output, and the failure mode each one has.
Questions about leaving it running.
Including what happens when a post does not pass its own checks.
What does “automation” actually cover here?
Planning, research, writing, images, metadata, internal links, publishing on the date, and the reporting loop. What it does not cover is the judgement: what a site should say, which entries survive review, and whether a piece sounds like the business. That part is a decision you make, on a plan, before anything is written.
Do posts go live without me seeing them?
They can, and that is the default for a site publishing on its own schedule. If you would rather approve each one, turn on Require approval before publishing: every new post then waits in the calendar as Needs approval until you approve it, and approving writes the article and publishes it exactly as it would have.
How much can it publish in a month?
As often as your publishing days allow and your budget covers. The planner holds the next three publishing days at a time and refills as posts go out, so a daily site and a weekly site both stay planned without generating a year of speculative posts nobody will read.
What happens when a post fails its checks?
It is repaired with the specific problem attached — an unsupported figure, a missing product mention, advice where material was needed — and re-checked. If it is still not good enough it is discarded and marked failed review in the calendar, visible to you, rather than published quietly.
Can it use my product's real pricing?
Yes, and it prefers to. The conversion page an article links is fetched and its text becomes part of the facts the writer works from, so a commercial post can state your own price instead of inventing market rates. Point the call to action at your pricing or sign-up page and that page becomes the anchor.
What does the pipeline need from my site?
A reachable site, because research and internal linking come from your real pages and sitemap. If the site is unreachable the article degrades — fewer links, less grounding — rather than publishing with invented specifics.
How is this different from an AI writing tool?
A writing tool starts with a prompt and ends with a draft you still have to plan, illustrate, link, publish and watch. This starts with a plan aimed at queries your site does not answer and ends with a published URL and the impressions it earned. Writing is the middle of the job, and it is the part a model is actually good at.
How much does it cost?
20€ per website, per language, per month, with nothing charged per article — a market that publishes one post a month costs the same as one that publishes twenty. There are no credit packs, no per-word meters and no extra charge for regenerating a post.
Can I publish into an existing CMS instead of the hosted blog?
Yes. Finished posts are available through a read API as markdown or rendered HTML, with the metadata, the images and the language attached, so an existing CMS, a static-site build or a warehouse can be the destination. Nothing in that API can write back to your site, so a bad integration can lose you a post but cannot publish a wrong one.
Does it guarantee rankings, traffic or revenue?
No, and no honest supplier can. What you get is a plan aimed at queries your site does not answer yet, dated publishing on your own domain, the on-page work done, and Search Console numbers per site to see what it earned. Ranking, traffic and revenue are not guaranteed, and the terms say so.
Where does the content live?
On a blog hosted on your own domain, with its own sitemap, canonical URLs and structured data — or, if you would rather publish through an existing CMS, the finished posts are available through a read API as markdown or rendered HTML.
Is this a bulk article generator?
No. There is no button that asks for fifty articles, and no way to publish a post with no plan entry behind it. Each entry is a search query the site does not answer yet, committed to a date and written to that brief — which is why the calendar can be reviewed before anything is written.
What happens to the blog if I miss a week?
Nothing breaks. The calendar is the source of truth, so a post publishes on its own date whether or not anyone looked that week — and with approval switched on it waits instead, in the order it was planned. That combination is the point: the schedule keeps its promises, and it stops when you tell it to.
See it run
Connect a site, watch one post go end to end.
A plan lands within minutes, dated on the days you publish. Discard the entries that do not fit, and see what the pipeline does with a real brief.
Approval mode includedPublish on your own domain3 days free, no card