UNCHARTEDSEOHow it works: seven steps, and you hold the last one.
Uncharted SEO automates SEO content in seven steps: onboarding facts, keyword research and clustering, competitor scoping, drafting, gates, your approval, and publishing with daily measurement. Nothing reaches your site without the approval step. Before a draft ever gets to you, the gates check it against the facts you approved, grade it on Yoast on-page checks, hold it to a readability check, and test it for cannibalization against the pages you already have. This page walks each step: what happens, what you see, and what can stop it.
How does SEO automation with human approval work at Uncharted SEO?
SEO automation with human approval at Uncharted SEO runs in seven steps: you approve a fact registry, the system researches live search results, scopes the competing pages, drafts, runs the gates, waits for your approval, then publishes to WordPress after a snapshot and measures every keyword daily. Nothing goes live until a person on your side approves it.
One pipeline, run every day, with a person at the gate.
Most SEO automation tools describe themselves as a button: press it, articles appear. This one is better described as a queue with checkpoints. Work moves from research to draft to gates to your inbox, and only the last checkpoint publishes. The steps below are in the order a piece of content travels. Steps one through five are software. Step six is you. Step seven is software again, watched daily.
Know before you write
Your approved facts, the keyword map, and a measured read of the pages already ranking.
Draft and test
The article is written against the registry, then checked by every gate before a person sees it.
You approve
The reviewer reads the draft, the gate report and the diff. Nothing publishes unapproved.
Publish and measure
Snapshot, publish to WordPress, then daily rankings and Google AI Overview citation state.
What happens, what you see, and what can stop it.
Onboarding: the fact registry
Everything starts with a list of what your business is allowed to say. At onboarding you approve the claims, prices, dates and credentials that may appear in your content: the services you actually sell, the prices you actually charge, the year you were founded, the licenses and certifications you hold, the results you can document, and the phrases you never want associated with your name. This list is the fact registry, and it is the reference point for every later step. A draft may state what is in the registry. It may not state anything else about your business.
What happens
We collect the facts from you in plain language, confirm each one back to you, and load the approved set into the registry. Banned phrases go in beside the approved ones, so the gate can refuse a sentence as well as a number. When your prices or credentials change, the registry changes with them, and every future draft follows the new version.
What you see
The registry itself, as a readable document: every approved claim with the date you approved it, every banned phrase, and every open question we could not answer without you. You can add to it at any time. You will also see the same registry named on every gate report, so you can trace any sentence back to the fact that permitted it.
Research: keywords, clusters and the cannibalization check
The second step decides what to write. Live search research finds the queries your buyers actually type, groups them into clusters by what the results pages have in common, and maps each cluster to one page. Then, before any new page is planned, the cannibalization check compares the proposed target against every page you already have. Two pages chasing the same query split the signal between them and both lose. The check exists because that mistake is easy to make at scale and expensive to undo.
What happens
Research runs against the live results for your market, not a stale database. Clusters are formed from what ranks together, so the page plan reflects how search engines already group the intent. Each proposed page gets a target keyword, a cluster of related queries, and a note on which of your existing pages sits closest to it.
What you see
A page plan: the target keyword, the intent behind it, the cluster it serves, and the cannibalization verdict for each planned page. Where an existing page already owns the intent, the plan says so and proposes strengthening that page instead of adding a new one.
Scoping: what the pages beating you actually have
Before a word is drafted, the pipeline reads the pages that currently rank for the target and measures them. Word count, heading structure, the questions they answer, the schema they carry, the tables and dates they include, and how they link internally. This is competitor scoping, and its purpose is to set a bar the draft must clear, measured rather than guessed. An article that is shorter, thinner and less structured than the pages already ranking is not going to displace them, however well it reads.
What happens
The top results are fetched and torn down into a structural profile. The profile becomes the brief for the draft: the sections it must include, the questions it must answer, the depth it needs, and the markup that the ranking pages share.
What you see
The brief, with the measured bar written into it. You can read exactly why the draft has the length and structure it has, because the reason is a number taken from the page it needs to beat.
Draft: written against the registry
The draft is produced from three inputs and nothing else: the brief from scoping, the cluster from research, and your fact registry. Where the article needs a fact about your business, it takes the fact from the registry. Where the registry has no such fact, the draft does not supply one. This is the difference between drafting against a registry and drafting against the open web. The open web will tell a writing model anything it wants to hear about your company. The registry tells it only what you approved.
What happens
The article is written to the brief: the headings, the answer to the query in the first hundred words, the sections the ranking pages share, the FAQ, and the internal links to your own pages. Prices and dates are inserted from the registry by reference, so a price change in the registry changes the draft before it is ever reviewed.
What you see
Nothing yet. A draft that has not passed the gates is not shown to a reviewer, because a reviewer’s time is the scarcest thing in this pipeline and the gates exist to protect it.
Gates: fact lock, on-page grade, readability, cannibalization
Every draft passes a battery of checks before a person sees it. The first and most important is the fact lock: every claim, price, date and credential in the draft is compared with the registry. A violation is repaired to match the registry when the correction is unambiguous, and refused when it is not. The other gates check what search engines and readers will judge: the Yoast on-page grade for the target keyword, a readability check so the article reads like a person wrote it, and a second cannibalization pass against your live pages now that the actual text exists.
What happens
The gates run in sequence and each produces a verdict with evidence: the sentence, the rule it broke, and what was done about it. A draft that fails a gate goes back for repair. A draft that cannot be repaired is refused and the reviewer is told why. The table below lists the gates in the order they run.
| Gate | What it checks | On failure |
|---|---|---|
| Fact lock | Every claim, price, date and credential against your approved registry; every banned phrase | Repaired to the registry value when unambiguous, otherwise refused with the sentence quoted |
| Yoast on-page grade | Target keyword in title, heading, intro and body; meta description; internal links; image text | Draft revised until the on-page checks read green |
| Readability floor | Sentence length, paragraph length, transition density, passive voice, reading grade | Draft rewritten for the failing passages only; good copy is not churned to chase a score |
| Cannibalization pass | Overlap between the finished draft and your live pages, by intent and by text | Draft retargeted or the plan folded into the existing page |
| Structure check | The brief’s measured bar: sections present, questions answered, schema and tables included | Missing sections added; the bar is not lowered |
What you see
The gate report, attached to every draft that reaches you: each gate, its verdict, and the evidence behind any repair. If the fact lock changed a price to match the registry, you see the original sentence and the corrected one side by side.
Your approval: the feature, not the formality
Nothing publishes unapproved. This is the step other tools let you switch off, and the step we treat as the product. The reason is simple: the fact lock can only refuse what it can recognize, and a person reading the draft with the registry beside it can recognize more. Approval is where a claim that is technically permitted but commercially wrong gets caught, where tone is judged, and where the decision to publish is made by someone accountable for it. For a law firm, a clinic or a financial advisor, that accountability is not optional, and we do not pretend otherwise.
What happens
The approved-pending draft lands in your review queue with everything the reviewer needs on one screen: the full article as it will appear, the gate report, the diff against the registry showing every fact the draft relied on, the target keyword and the pages it links to. The reviewer approves, requests changes, or rejects. Requested changes go back through the gates, because an edit can introduce a violation as easily as a draft can.
What you see
The approval screen. Article, gate report, registry references, links, and a single approve action that records who approved and when. That record travels with the page: if a question is ever asked about why a sentence was published, the answer is a name and a date, not a shrug.
Publish and measure: snapshot, rollback, daily tracking
When a draft is approved, the pipeline takes a snapshot of the page as it stands, then publishes the approved version to your WordPress site. The snapshot means rollback is one click, and the snapshot is kept, so a restore is possible days later as well as seconds later. From that moment the page joins the measurement layer: daily rank tracking on every keyword in your plan, and for each keyword a record of whether Google shows an AI Overview and who it cites. Monthly, the technical audit re-reads the whole site and the backlink opportunity report lists the domains linking to two or more of your competitors and not to you.
What happens
Snapshot, publish, verify. The published page is read back and compared with the approved draft so that what went live is what you approved. Then the daily series begins for its keywords, and the page appears in the next technical audit with its index status read from Search Console.
What you see
Daily positions for every tracked keyword with full history. For each keyword, the Google AI Overview citation state: whether an overview appears, who it quotes, and when that becomes you. The monthly technical audit covering broken links, redirect chains, schema state, internal link structure and index status. The monthly backlink opportunity report with outreach targets exported. And the rollback control for every page we have ever touched.
What stops a publish, and who acts.
A pipeline is defined as much by what it refuses as by what it produces. This table is the list of conditions under which nothing goes live, in the order they occur.
| Condition | Result | Who acts |
|---|---|---|
| A fact the content needs is not in the registry | The claim is left out or the draft waits | You, by answering or approving the fact |
| A proposed page overlaps a live page | The new page is not drafted | Software retargets; you approve the retarget |
| The ranking pages win on a claim you cannot support | The section is cut, not fabricated | Software, recorded in the brief |
| The fact lock finds a claim outside the registry | Repaired to the registry value or refused | Software; the reviewer sees the evidence |
| Any other gate fails | The draft goes back for repair | Software, until every gate passes |
| The reviewer requests changes or rejects | Nothing publishes | You, or your named reviewer |
| The reviewer has not acted | Nothing publishes; there is no timer | You, whenever you are ready |
| The published page does not match the approved draft | Snapshot restored immediately | Software; the reviewer is told first |
What the gates have actually refused.
A gate that has never caught anything is a decoration. These are the catches on record, stated the way the record states them, with the honest caveat that most of them were legacy copy, not fresh drafts.
The check has caught real errors on live sites: around 1,000 unapproved claims on one client’s site, nine fabricated customer reviews on our own site, and six invented dates caught in testing.
The approval step stays for a simple reason: a gate that has not failed yet is not a gate that cannot fail.
Figures are from our own system records dated 2026-08-26 to 2026-08-31 and are published here only after our founder signs them. Full detail on the proof page.What we do not do, and do not claim to.
The same discipline that governs your content governs this page. These are the things the pipeline does not do today. Some are policy and will stay that way. Some are roadmap and will be sold when they ship, not before.
- No link exchange and no placed links. Policy. The monthly backlink opportunity report gives you targets; it does not build links for you.
- No unattended publishing. Policy today. Every publish is approved by a person.
- No tracking of ChatGPT, Perplexity or Gemini answers. We track Google AI Overview citations on every tracked keyword. Say exactly that, because that is exactly what it is.
- No publishing to Shopify or any platform other than WordPress. Roadmap, not sold.
- No Reddit or Quora agents, and no multi-language content. Roadmap, not sold.
How the pipeline works: straight answers.
Can I turn the approval step off?
No. Every publish is approved by a person today. Approval is not a setting we hide behind a toggle; it is the control that lets the fact lock mean something. The reviewer can be you, someone on your team, or a strategist on the Scale plan working from your registry. If unattended publishing ever ships, it will ship as a feature with its own proof, not as a default.
Do you track ChatGPT or Perplexity?
Not today. We track Google rankings daily on every keyword in your plan, and for each of those keywords we record whether Google shows an AI Overview and who it cites. That is Google AI Overview citation tracking, refreshed continuously, and it is the only AI search measurement we sell.
Do you build backlinks?
No. We deliver a monthly backlink opportunity report: domains that link to two or more of your competitors and not to you, exported as outreach targets. We do not run a link exchange and we do not place links. Whether to pursue the targets is your decision.
How long until I see results?
Most sites see movement in three to six months. New pages are indexed faster than that, and rank tracking shows the early positions from the first days, but durable positions and AI Overview citations build over months of consistent publishing. Anyone who promises faster is describing something other than search.
What happens if a published page goes wrong?
Every publish is preceded by a snapshot of the page as it was, and rollback is one click. The snapshot is kept, so a page can be restored days later, not only in the first minute. The measurement layer will also show the damage: a ranking drop or lost citation appears in the daily series.
What is in the fact registry?
The claims, prices, dates and credentials you approve at onboarding: what you sell, what it costs, when you were founded, which licenses or certifications you hold, which results you can prove, and the phrases you never want used. Drafts may state only what is in that registry. Anything outside it is repaired to match or refused.
Which platforms do you publish to?
WordPress. The publish step writes the approved draft to your WordPress site over its REST API, with the snapshot taken first. Other platforms are on the roadmap and are not sold until they ship.
Who does the research and drafting, people or software?
Software does the research, the drafting and the gate checks; people set the registry, review the draft and approve the publish. That division is deliberate. Software is good at reading a thousand search results and bad at knowing what your business is allowed to say. The registry and the approval step put that knowledge where it belongs.
Ready to see the plans? Two plans, published before you talk to anyone. Serving a regulated industry? Read how the pipeline is built for legal, financial, medical and home services businesses.
See what the two plans include and what they cost on the pricing page.
Page reviewed 2026-09-03. Every figure above is re-verified on that date or removed.
Updated September 6, 2026