You asked for the full picture. This is the machine: every stage, every gate, every check that fires between "pick a topic" and "published and maintained". AI writing tools do stage 05 and stop. We run all of it, every time.
15+ live results read per article50+ banned AI patternsevery source fetched & verified0 made-up numbers
Pipeline one
Publishing a new article
Eight stages. If the winnability gate says no at stage 02, everything stops, and no credit is spent on a fight you can't win.
01
Topic selection, from evidence
Pulls every keyword your site already ranks for, with the exact page that carries it
Scans for pages competing with each other for one query, in live rankings AND your real Search Console data
Finds near-wins: real queries sitting at positions 11-25, one push from page one
Scores every candidate topic on demand, winnability, and your business goal
Stay-in-lane filter: a topic must match YOUR business vocabulary. Competitor-only topics are rejected, and the number filtered out is disclosed
Mines real community questions (the demand keyword tools can't see)
Anti-cannibalization: never queues two variants of the same query
02
The winnability gate
Reads 15+ live results for the target search before a word is written
Classifies every result: tool, listicle, guide, product page, forum, video
REFUSES to write when the search can't be won by an article (like a results page owned by software tools)
Picks the article's format from what actually ranks, not from a template
03
The research brief
Fetches and reads the full text of the top-ranking pages
Maps what every competitor covers (table stakes you must match)
Maps what they ALL miss (your opening to win)
Adds real questions people ask on Reddit about the topic
Sets length from evidence: the median depth of what ranks, plus margin. 800 words or 5,000, whatever fully answers the intent
04
Source verification (only when the topic needs it)
A classifier decides IF the topic needs citations at all: health, money, legal, and statistics do; practical how-tos don't get links stuffed in for show
Collects the references the ranking pages themselves cite, preferring official and research sources
Fetches every candidate source LIVE, seconds before writing. Dead pages and soft-404s never enter
The writer may only cite from that verified list. Any other link is stripped by code before publishing. A made-up study physically can't ship
05
Writing, on frontier models
Runs on the newest frontier AI models, not the cheap ones
Writes in your measured voice: sentence length, formality, humor, point of view, reading level, heading style, all learned from your real posts
Grounded in your business facts, products, and never-say rules
50+ banned AI words, phrases, constructions, and sentence openers, enforced
Rhythm rules against pattern-writing (the tell that screams AI)
Answer-first structure with key facts up top, built for AI answer engines to quote
Word count measured against the evidence target; short drafts are automatically sent back and rewritten deeper
06
Deterministic enforcement (code, not trust)
Citation whitelist gate: every external link checked against the verified pool
Your casing rules and punctuation rules applied mechanically (no em dashes, ever, if that's your style)
Your real product photos placed where their descriptions match the section, capped per article, never at random
Title and heading case enforced to your brand's convention
07
Delivery
SEO title, meta description, and structured data rendered in the page head, never pasted into the body
Internal links with destination-first anchors: 2-6 words, must carry the destination's differentiator, never a term two pages share, one anchor phrase per destination site-wide
A branded cover image generated per article: your colors, your logo, a topic-matched icon, geometry that never repeats
Draft-for-review or straight to published, your choice
Every action written to your activity log in plain words, with one-click undo
08
Evaluate, forever
Tracks every article's clicks, impressions, and position from your real Search Console
Watches for decay: a page that starts slipping gets a fix scheduled before the traffic is gone
Checks index status through Google's own inspection system, with the cause and the fix when a page is stuck
What worked feeds the next plan. The loop never stops
Pipeline two
Updating a live page
Different job, different pipeline. The page keeps its URL and every bit of authority it earned; only what's missing gets added.
01
Detection, from real signals
Decay watch: position and click slides on pages that used to win
Near-miss radar: queries at positions 11-25 with real impressions, where one push beats a new article
Content audit: every page measured against what currently ranks (depth, structure, internal links)
Broken links and redirect chains from real HTTP checks
Index diagnosis: why Google is skipping a page, with the specific cause
02
Diagnosis, measured not guessed
Word depth compared against the ranking median for the page's real queries
Queries Google already tests on the page (your Search Console data) checked against the page's actual text: answered, half-answered, or missing
Every verdict carries the evidence, never a hunch
03
The targeted refresh
Updates the existing URL, so the page keeps every bit of authority it earned
Adds the missing sections and answers the tested-but-unanswered queries
Runs the same full enforcement pipeline as a new article (voice, citations, casing)
Review mode: nothing changes on your site without your approval
04
Re-measure
Before and after in your activity log
The refreshed page goes back under the decay watch
One update = one credit, honestly counted. The first 3 revisions of any draft are free
The honest part
What we're not showing you
The exact scoring formulas, thresholds, and prompts stay ours. But every claim on this page is a real stage in the shipped product, and everything it does on your site is logged in plain words with an undo button. Watch it work in the live demo.
Now watch it run on your site.
$7 for 7 days, 2 full pipeline runs. Cancel anytime.