A month of content on autopilot, done properly

"It writes your blog for you" is the promise, and it's half true in a way that matters. Rankbank will find the topics, schedule them across your publishing days, and write drafts from your own Search Console data and your own pages. What it will not do is decide what your company actually knows, or notice when a draft is competent and forgettable.

So the useful question isn't whether it can write a month of content. It's where you stay in the loop, and how little time that has to take. This walkthrough sets up a real month, and the honest answer is about twenty minutes a week after the first hour.

What you'll have by the end

• A month of publishing days planned from measured search demand, not guesses
• A plan you audited before anything was written, with the wrong entries removed and two steered
• One article drafted, edited where it was generic, and published to WordPress
• A weekly routine and a month-end check that tells you whether any of it earned clicks

Prerequisites, done right

Two things, and only one is required.

Your site must have been read at least once. The plan is built from the topics your site already covers, so the calendar can avoid repeating them and can link to them. Without a read there is nothing to plan from, and the third stage stays locked. If you haven't done that, your first site audit is half an hour.

Search Console is optional, and here's the honest difference it makes. Connected, your plan comes from searches Google already shows your site for, with real impressions and real positions attached, and those keywords outrank everything else in the queue. Not connected, the plan comes from your own pages plus AI-generated adjacent topics, which is a reasonable starting point and a weaker one. The onboarding says as much: without it the plan comes from your pages alone, with it, from what people actually search for.

Connect it. It takes two minutes and connecting Search Console and turning its data into clicks walks the whole flow.

The first build

Open Content calendar for your site. On a fresh calendar you get an onboarding panel headed with the number of days it's about to plan, and a promise underneath: four steps, in order, and only one is ever waiting on you.

The fourth stage is the one to read twice, because it's the product describing its own limits: each article is written on its scheduled day, you review it and send it to your CMS, and nothing publishes on its own.

Press Build my {N} days.

Four things happen, and a progress banner tells you which one it's on.

Keywords get found from three sources. They get scored. The highest scoring ones get assigned to your open publishing days, spread across the week according to your cadence rather than clumped. Then articles are written on their own days, not now.

When it settles you get a month grid and four tiles: Days planned, Ready to write, Drafts, and Published.

Each day cell carries a format badge like guide or list, the keyword or the article title, search volume and difficulty when they're known, and a button whose label depends on the state. Future days read Opens {date} and are deliberately disabled, because articles are written on their scheduled day so they reflect the site as it is then. Today's read Write it now. Drafted ones read Review & publish. Published ones read View published post.

That's the whole calendar. There's no list view, no filters, and no dragging entries between days. It's a month grid. Reading the calendar covers every cell state.

Audit the plan before writing anything

This is the step that separates a useful calendar from a content farm, and almost nobody does it. Spend twenty minutes here.

Open the Keywords tab. Every keyword behind your calendar is listed with exactly why it was chosen.

The line above the table tells you the split: how many of your keywords come from real Search Console measurements, and that the rest are suggestions to widen coverage, deliberately scored lower so they only get scheduled once the measured ones run out. That sentence is the whole scoring philosophy, stated by the product about itself.

The source badges make it concrete. Search Console means Google measured it. Your pages means it was mined from your titles, headings and descriptions during the read. AI suggestion means it was generated as an adjacent topic around a proven cluster, and it sits lower for that reason.

Read down the Why column. The reasons are specific: impressions in the period, the average position and whether it's in review range, whether the click-through rate is low for the demand, how many internal links the relevant page has. If a keyword's reasons don't convince you, it's not going to convince a reader either. Where the keywords come from explains the scoring.

Now go back to the calendar and remove the entries that don't fit. Hover a cell and three controls appear.

The trash icon removes the day, and the keyword goes back to the pool rather than being destroyed. Remove anything you'd be embarrassed to publish, anything that duplicates a page you already have, and anything about a product you don't sell. On a typical first plan I remove three or four out of twenty, and it takes five minutes.

Then steer two entries. This is the part that decides whether the drafts sound like you.

The lightbulb opens interview questions, under a heading reading "Make this article yours", with an instruction that's worth quoting: answer whichever are useful, and your answers become the parts of the article nobody else could write.

The questions are about the topic, and they're the ones a journalist would ask. What goes wrong most often. What you'd tell somebody starting today. What the number actually looks like in practice. Answer two or three in a couple of sentences each. That's it. Those answers are the only thing in the finished article that a competitor cannot also generate.

The question-mark icon opens freeform instructions instead, for when you know exactly what you want. Its example gives the register: lead with the pricing comparison, and mention that the free plan has no page limit. Use it for structural demands and specific facts you want included.

Instructions travel with the entry and show up in the article's evidence afterwards, so you can see whether they were used. Steering an article before it's written covers both dialogs.

Reading a draft with its evidence

Press Write it now on today's entry, or open the entry and press Write this article. It takes up to a minute.

Then you land in a two-column screen: the article on the left, publish controls and an evidence panel on the right.

Read the evidence panel first. It tells you what this draft was built from: the target query, the Search Console queries it answers with their impressions and average positions, whether your instructions were used, and notes about how it was assembled, including how many internal link targets from your own site were offered to the writer. A badge says whether it was AI written or is an evidence brief, which is what you get when article generation isn't configured on your deployment. An evidence brief is structured real data rather than invented prose, and it's clearly labelled as such.

Now read the article against that evidence, with three questions.

Does it actually answer the target query in the first two paragraphs? Drafts that warm up for four paragraphs lose the reader who searched for a specific thing.

Where is it generic? This is where your editing time goes. Any paragraph that could appear on a competitor's site unchanged is a paragraph worth replacing with something only you know. If you answered the interview questions, look for whether those answers made it in, and if they landed in the wrong section, move them up.

Are the claims safe? Generation is instructed not to fabricate statistics, and any phrases you banned in settings are enforced. That's a floor, not a guarantee of judgement. If a sentence states something about your product or your pricing, check it.

Edit anything you like. Title and meta description carry live character counts, warning you when a title runs past the point search results usually cut off, and when a description is likely to be clipped. Body is markdown. Press Save changes, and the entry gets marked as edited by hand, which is a useful signal later when you're wondering which pieces you actually touched.

If the whole thing is wrong rather than generic, regenerate it after adding instructions. A second pass with two sentences of steering is usually dramatically better than the first. Reviewing and editing a draft covers the editor in full.

Connecting WordPress

Open the Where it publishes tab. Its description sets the expectation: connect the blog you publish to, and finished articles go straight there, as a draft for you to approve or live, whichever you choose.

Press Add a connection. Pick WordPress from the platform list, which also includes Ghost, Webflow, Wix, Shopify, Strapi, Medium, and your own webhook.

For WordPress you need three things, and the hint on the form tells you where the third comes from: in WordPress, go to Users, then Profile, then Application Passwords, and create one for Rankbank.

Fill in your site URL, which is your public site rather than the admin URL, your username, and that application password. Credentials are stored encrypted, and secret fields stay write-only afterwards, showing a placeholder rather than the value.

Then set the defaults. Post articles as takes Draft or Published, and for the first month choose Draft so nothing reaches your readers before you look at it in your own CMS. Author name, default categories and default tags round it out.

Save with Add connection, then press Test on the row. A working WordPress connection replies "Connected to WordPress." A failure tells you what went wrong, which is nearly always a wrong URL or an application password that got a space pasted into it.

Connecting where it publishes covers the other platforms.

Publishing the first one

Back on the drafted entry, the publish panel on the right has two selects and a button. Send it to picks the connection. As lets you override the default for this one article, either using the connection default, sending it as a draft to review in your CMS first, or publishing immediately.

Press Publish. If you have unsaved edits the button reads Save and publish instead, which saves you the classic mistake of publishing the version before your changes.

The entry's publishing history records what happened. A success carries the remote URL, and the day cell on the grid turns green with View published post. A failure records the error message from your CMS, which is usually a permissions problem and is fixable without guessing. Publishing and what happens next covers the statuses.

Then go and look at the post on your own site. Check the formatting survived, check the internal links point where you expect, and add your image.

Letting the rhythm carry it

In Style & settings there's a toggle called Keep this calendar rolling. Left on, it adds a new day as each one comes into range, so you always have about a month planned ahead. Turned off, the calendar simply runs out.

Each night the calendar plans the newly-due day, mines more keywords if the pool has run dry, and writes a few of the articles that are due. Not all of them at once, and not at a time you should build a routine around. What you'll notice in the morning is that a couple of cells have moved from ready to drafted, and a notification is waiting to tell you a draft is ready.

Your weekly twenty minutes looks like this.

Open the calendar and review whatever drafts are waiting, using the three questions above. Publish the ones that are good, edit the ones that are close, regenerate the one that missed with instructions attached.

Glance at the coming week's cells and steer anything that looks thin, with interview answers rather than more instructions.

Check that last week's published pieces are actually live and look right on your site.

That's it. If a week gets away from you, nothing breaks. The drafts wait.

While you're in settings, it's worth setting the rest properly once: articles per week and your timezone under the publishing schedule, since an article unlocks at midnight in the timezone you set, not the server's. Tone, reading level, target length, audience and banned phrases under how the writing should sound. Internal and external links per article under what every article should include, where internal link targets are taken from pages found in your site audit, so the links are real. Style and settings covers every field.

Month end: did it work?

Four to six weeks in, go to Search traffic and press Sync 28 days.

New articles need time. A month is early, and impressions arrive well before clicks do, so the first sign of life is usually a page appearing with impressions and a position in the twenties rather than a page earning traffic.

Look at three things.

Pages with impressions across the six stat cards. If your published pieces are showing up at all, Google has found and indexed them, which is the first hurdle.

Pages that gained, for anything you published this month.

And the growth plan, which will start pricing the new pages once they have demand. A new article that lands at position fourteen becomes a page-two entry with a clear next step, and that's a much easier piece of work than writing it was.

The pieces that don't move within two months are the ones to learn from. Look them up in the keyword pool and check the source badge. In my experience the misses are disproportionately AI suggestions rather than measured queries, which is exactly why they're scored lower, and it's a good argument for connecting Search Console if you haven't.

Then adjust: fewer articles a week at higher quality, more interview answers, tighter topics. The calendar will keep planning. Your judgement is the input that makes it worth reading.

Where to go from here

• Reviewing and editing a draft: the editor and the evidence panel in detail
• Where the keywords come from: the three sources, the scoring, and the honest ranking between them
• Style and settings: cadence, voice, structure, and what applies when