Method
Research first. Then words.
Most AI writing tools take a keyword and start generating. That is why the output reads like it could belong to anyone — because it was written from nothing. A page has to be built on something, and the only material worth building on is what is already working for that search.
Here is exactly what happens between you naming a keyword and a page appearing.
We pull the top five results for your search
Before a single word is written we ask Google what it currently rewards for that exact search, in your country and language. Not a guess about what might work — the five pages that are actually there today.
This matters because a search result is a verdict. Google has already tested thousands of pages against millions of people for that query. The pages sitting at the top are the ones that survived. Ignoring them and writing from imagination is how most content fails.
We read those pages and take them apart
Each of the five is downloaded and broken down: how they structured the topic, which sections they all agreed were necessary, how long they are, which terms and tools and organisations they name, which numbers they quote and where those numbers came from, and which questions they answer directly.
Out of that comes a picture of the subject as the market currently understands it — the things a reader will expect to see, and the things they will notice are missing.
The opportunity is what all five of them skip
Five pages covering the same ground always leave the same gaps. A question everyone raises and nobody answers. A step described in theory but never in practice. A claim repeated across all five with no source behind any of them.
That gap is the reason your page has a right to exist. Publishing a sixth version of the same article gives a reader no reason to stay and gives Google no reason to move you up.
We do not reinvent the format
If the winning pages open with a plain definition, ours opens with a plain definition. If they all use a comparison table, we use one. That is not laziness — it is the format the audience and the algorithm have already agreed on, and breaking it costs you readers for no gain.
Originality belongs in what the page says, not in surprising the reader with an unfamiliar structure. Same shape, better contents, plus the part nobody else covered.
Readable first, optimised second
The answer to the search appears in the first two sentences. No warm-up paragraph, no “in today’s fast-moving digital landscape”. Short sentences. Sections a reader can skim and still understand. A table when a table is genuinely clearer than a paragraph.
The target phrase appears where it belongs and nowhere else. Keyword stuffing has not worked for a decade and it makes a page unpleasant to read, which costs you the thing you were optimising for.
Structured so machines can quote it
Answer engines assemble replies from fragments. A definition given as one self-contained sentence under a heading that matches the question is far more likely to be quoted than the same fact buried mid-paragraph.
So every page gets a clear definition, labelled sections, direct answers to direct questions, and structured data that matches what is actually visible on the page. This is what makes a page eligible to be cited by ChatGPT, Perplexity and Google’s AI answers rather than paraphrased without credit.
Two checks, and both can stop it
A draft is not a page. Everything produced has to survive a mechanical pass and then a second model whose only instruction is to find a reason not to publish.
First: the mechanical pass — no AI involved
Plain rules, applied identically every time. If one fails, the draft goes back to be rewritten and never reaches the second stage.
- Length
- Long enough to cover the subject as the winning pages do — not padded to hit a number.
- One H1
- A page with three main headings is telling search engines it is three pages.
- Sourced numbers
- Every statistic must name where it came from. Unsourced figures are rejected automatically — this is the check that fails most often.
- Keyword sanity
- The phrase must appear, and must not be repeated past the point of comfort.
- Internal links
- A page nothing links to is a page you are asking Google to find by accident.
- No filler
- A blocklist of the phrases that signal generated text. If they appear, the draft goes back.
Second: Substance
- Would someone who knows this industry spot an error?
- Is anything asserted as fact with nothing behind it?
- Does it cover what the winning pages agreed was necessary?
- Does it actually fill the gap we identified, or just restate the topic?
Second: Craft
- Does the opening make you want the next sentence?
- Is it written by someone with a point of view, or assembled?
- Could this paragraph appear on a competitor’s site unchanged?
- Would you send this to a colleague?
Both are scored separately and both have to pass. A factually solid page that reads like a manual fails. A beautifully written page with an invented statistic fails harder. The model that wrote the draft never scores it — a writer marking their own work is not a check.
When a draft fails, the exact reason is fed back and it is rewritten. If it fails twice, it goes to a person — not to your website.
Why this is slower on purpose
A tool that publishes daily can do it because nothing stands between the model and your website. We put two things in the way, so we publish less. That is the product.
A page that adds nothing is not neutral — Google assesses your site as a whole, so cheap volume drags down the pages that were working. The fastest way to damage a site is to fill it.
See what your search results look like first.
Free scan: we crawl your pages, pull the real search volumes for the terms you name, and show you where you stand. No signup, no card.
Scan my siteRelated: the full system · getting cited by AI