UNCHARTEDSEOSEO automation for medical practices that cannot invent a credential.
A medical practice can use automated content on one condition: the system must be unable to publish a credential, a certification, an outcome or a price your practice has not approved. Here, every one of those comes from a registry you sign off first, and a draft containing anything outside it is repaired or refused before a person ever sees it. Then someone at your practice approves each article before it goes live; nothing publishes unattended. We do not write clinical claims, and we do not make claims about patient outcomes. Launch is $100 per month, or $1,000 per year. Scale is $250 per month, or $2,500 per year.
What is SEO automation for a medical practice, exactly?
It is the production side of search work, run by a machine, with the parts that require a person’s judgment left with the person. Concretely, on the two plans sold here: search engine and answer engine article production for one practice website, drafted against a fact registry your practice approved before any drafting began; automatic publishing to WordPress with a snapshot taken before every change and one action to roll it back; daily rank tracking on your tracked keywords together with Google AI Overview citation tracking; a full technical site audit; keyword research with clustering built from live search results; and a monthly backlink opportunity report. Underneath all of it sits the fact lock, which is the part that matters in health content.
The word automation gets used loosely in this category, so it is worth being blunt about the boundary. Automating the writing of a paragraph is easy and every tool in the market does it. Automating the question of whether that paragraph is true of your practice is the hard problem, and most tools do not attempt it. They generate, they present a draft, and they leave the checking to whoever opens the editor at four in the afternoon. That arrangement works acceptably for a blog about project management software. It does not survive contact with a page that names a physician, a certification and a procedure in the same sentence.
Your approved facts are a boundary, not a briefing document
Your approved facts are not context handed to a writer as helpful background. They are the boundary of what any draft is permitted to state. A draft that steps outside the registry is repaired to match it or it is refused. The system cannot generate a claim outside the registry it was given.
The check runs before a person reads the draft
Approval is a second line, not the first one. By the time a draft reaches your reviewer it has already been through the fact lock, the on-page checks, the readability floor and the duplication ceiling. Your reviewer is reading a draft that has already been made to conform, not auditing raw output.
What it is not
It is not a link building program, it is not paid media, and it is not website design. It does not manage your Google Business Profile, it does not touch your reviews, and it does not publish anywhere except WordPress. Those boundaries are stated in full further down rather than left for you to discover.
Why is generic AI content a specific risk for a medical practice?
Because health content is nothing but specifics, and a general purpose writing tool produces specifics fluently whether or not they are true of you. The work itself is not exotic: condition and treatment content in the language patients use, a service line structure that concentrates authority instead of scattering it, provider and location pages that stay consistent everywhere they appear, and the answer formatting that lets a search engine quote you. Every vendor in this category can describe that much. What separates them is what happens when the drafting system needs a specific.
The specifics are the whole job
A page about a procedure has to name the procedure, the providers who perform it, the conditions it addresses, the locations it is offered at, and usually what a patient should expect on the day. Each of those is a checkable fact about your practice. A general purpose writing tool produces all of them in the right register, with the right formatting, at the right length, whether or not any of them describes you.
The failure mode is not gibberish, which is what makes it dangerous. It is a paragraph that reads exactly like your marketing and contains a fellowship nobody in the practice completed, a service you stopped offering two years ago, or an outcome figure that came from nowhere. That paragraph is not a typo. It is a public statement your practice made about patient care, and the practice owns it from the moment it publishes.
Why a general web fact check does not solve it
Several content tools check statements against the open web. That catches a wrong date on a historical event. It does not catch a claim about you, because the open web holds no authoritative record of which certifications your physicians hold, which procedures your practice performs this year, which insurance you accept at which office, or which of your providers moved to a different location in March. Only your practice holds that record.
So the check has to run against a record you supply and approve, not against the internet. That record is the registry, and enforcing it on every draft before a human ever reads the draft is the reason this pipeline exists. The design argument in full is on the fact-checked AI SEO page, and the reason regulated categories need it most is on the YMYL SEO automation hub.
Scale is what turns a small error rate into a problem
One wrong sentence in one article is a correction. The same wrong sentence pattern repeated across a service line, a set of provider bios and eleven location pages is a position your practice appears to hold. Volume is exactly what automation is for, which means volume is also what makes an unchecked pipeline expensive. Either the check scales with the writing or the writing should not scale.
This is the practical reason the lock runs on every draft rather than on a sample. A sampled check tells you about the sample. It tells you nothing useful about the ninety articles it did not read, and by the time a quarterly review finds the pattern, the pages have been indexed, quoted and linked.
The reviewer fatigue problem is real and is designed around
A person asked to verify thirty articles a month, each containing perhaps forty checkable assertions about the practice, will be excellent at it in week one and unreliable by month four. That is not a character flaw, it is what sustained attention does. Every content operation that depends on a human catching every error eventually ships an error.
So the human step here is deliberately not the checking step. The machine does the exhaustive, boring, identical comparison against the registry, a thousand times, without the thousandth being worse than the first. The person does the thing a person is good at: deciding whether this page should exist and whether it says what the practice wants said.
What does your practice’s approved registry actually hold?
Onboarding is not a questionnaire about your goals. It is the assembly of the list of things your pages are permitted to state. Your practice writes it, your reviewer approves every entry, and nothing enters it on the system’s authority. For scale, one client’s registry currently holds 24 facts, 10 numbers, 6 category rules, 37 banned terms, as recorded 2026-09-02, in a single client registry. A medical registry is usually larger, because a practice has more named people and more named services than most businesses do. These are the categories it holds.
Providers, names and titles
Every clinician who may be named on the site, with the exact name, degree and title the practice uses for them, and the locations each one practices at. If a physician is not in the registry, no page can name them. If a title is written one way in the registry, that is the way it is written everywhere, which also closes the consistency problem that follows most multi provider practices around: three spellings of one name across a bio, a location page and a schema block, each of which a search engine has to decide whether to treat as the same person.
Board certifications and credentials
Which board, which certifying body, held by which named provider, and the wording you approve for it. Fellowship training, residencies, hospital affiliations, teaching appointments and society memberships live here too. This is the category a language model reaches for by reflex, because every practice website it has ever read is full of these phrases, and they are graded by fluency rather than by truth. Nothing here can be assembled from that reflex. It is stated as your practice supplied it or it is not stated at all.
Specialties and services actually offered
The conditions your practice treats and the procedures it performs, as of the date you approve the list, with discontinued services marked so that no page revives them. A service that is not on the list cannot become a page, cannot appear in a list of services, and cannot be implied in a sentence about what patients come to you for. This is also where the difference between a service you offer and a service you refer out gets written down, because that distinction disappears in generated copy almost every time.
Locations, hours and coverage
Every office, the address as you want it published, which providers see patients there, which services are available there, and the hours. Location detail is where invented specifics are hardest to spot, because one wrong suite number, or one office listed as offering a procedure it does not offer, reads perfectly normally to everyone except the patient who drives there. Consistent address and phone formatting across the site is a registry output rather than a cleanup project.
Numbers, prices and payment
Any figure a page may state: fees you publish, financing you offer, the payers and plans you accept, years in practice, patient volumes if you choose to publish them. Each entry carries the source and the date it was verified. A number that is not in the registry does not ship, however reasonable it looks in the sentence around it, and that includes anything resembling a result. Insurance acceptance in particular is a registry entry per location, not a site-wide sentence, because it is rarely true site-wide.
Banned phrases
The banned list runs alongside the approved one. Superlatives, comparative claims about other practices, outcome language your reviewer has struck before, marketing words your specialty avoids, and any phrase your counsel has told you not to publish. A draft that uses a banned phrase is refused rather than softened, because softening is exactly how a struck phrase comes back wearing different clothes. The list is yours to grow, and most practices grow it fastest in the first two months.
Category rules, for the cases a list cannot enumerate
Some constraints are shapes rather than strings. No page may state a recovery time. No page may compare a provider to another provider. No page may state what a reader should do about a symptom. Category rules cover those, so the lock is not limited to matching phrases it has seen before. They are the reason a novel sentence that breaks an old rule still gets caught.
The registry is yours. You can add to it, correct it or remove from it at any time, and the change applies to every draft written after it. Nothing in it is inferred, and nothing in it arrives from a competitor’s website, a directory listing or a model’s general impression of what a practice like yours probably offers.
What does the fact lock refuse on a medical page?
These are the categories where a fluent sentence becomes a problem, and what the pipeline does with each one. The examples below are generic on purpose. None of them is drawn from any client’s pages.
Outcomes, success rates and recovery timelines
“Ninety percent of patients return to work within a week.” “Most patients see complete resolution.” “Typical recovery is ten days.” Figures like these are the most common invented specific in health drafting, because the sentence structure demands a number and a language model supplies one that reads plausibly. Any outcome figure has to be an approved registry entry with a source and a date on it, or the draft does not ship carrying it. If your practice has no approved figure, the page is written without one, which is a perfectly normal way for a page to read.
Credential and certification words
“Board-certified”, “fellowship-trained”, “double board-certified”, “nationally recognized”. Each of these attaches to a named person and each is checkable by anyone who cares to check. The registry carries the credential, the body that issued it and the provider who holds it, and a draft may state exactly that pairing. It cannot promote a physician’s training a level, extend one provider’s certification to a colleague, or attach a credential to the practice as a whole because the sentence flowed better that way.
Clinical assertions and anything shaped like advice
Statements about what a condition is, what causes it, what a treatment does, when a patient should seek care, and what happens if they do not. The pipeline does not compose clinical content from a model’s general knowledge and publish it under your practice’s name. Clinical statements come from what your clinicians have approved, and anything that would read as advice to an individual reader is refused rather than written and caught later. This is a category rule, not a phrase list, so a new sentence that has never been seen before is still refused if it has that shape.
Drug, device and brand claims
Named medications, named implants, named systems, the properties attributed to them, and whether your practice uses them at all. Equipment lists age quickly, and every one of these names is simultaneously a factual claim about your practice and a factual claim about a manufacturer’s product. If the device is not in the registry, no page says you use it, and no page says what it does.
Patient stories, reviews and ratings
No draft writes a patient quote, invents a case, composes a before and after narrative, or attributes a sentiment to “our patients”. Review and rating structured data is not generated for you under any circumstances. If your practice publishes patient content, it comes from the practice with whatever consent your own process requires, and it enters the registry the same way every other approved fact does.
Coverage, cost and access promises
“We accept most insurance.” “Same-day appointments available.” “Financing available for every procedure.” Each of these is a promise made to a patient at the exact moment they are deciding whether to call. Insurance networks, appointment availability, wait times and payment options are registry entries per location or they do not appear, on any page, in any sentence.
Dates, tenure and volume
Founding years, years in practice, patients treated, procedures performed, staff counts. These are the quietest fabrications because nobody reads them closely and they are almost never wrong by much. They are still numbers the practice is publishing about itself, so they are registry entries with a source and a verification date, or they are absent.
The lock is a refusal mechanism. It is not a claim that a model never errs, and it is not a claim of compliance with any regulation. It stops a defined set of failures, and the approval step exists precisely because the set is defined rather than universal. Where the lock cannot see a problem, and there are places it cannot, that is documented on the fact-checked AI SEO page rather than hidden.
SEO automation for medical practices: what is automated and what is not?
The honest version of this answer is a table rather than a slogan. Research, scoping, drafting against your registry, the gates, publishing with a snapshot, and daily measurement are automated. The content of the registry, every judgment about care, and the signature on every publish stay with your practice. There is no plan and no setting that moves an item from the right column to the left.
| Step | Automated | Stays with your practice |
|---|---|---|
| Fact registry | Enforced on every draft: providers, credentials, services, locations, numbers, banned phrases, category rules. | You write it and approve every entry. Nothing enters it without a person putting it there. |
| Keyword research | Live search results research, clustering by patient intent, and a cannibalization check before any new page is proposed. | Approving the topics. A condition your clinicians will not write about never becomes a page. |
| Competitor scoping | Measures what the pages currently ranking for a term contain: length, heading structure, structured data, where the answer sits on the page. | Deciding what your practice should say about it, and whether it should say anything at all. |
| Drafting | Written against your registry, with on-page grading and a readability floor, in the language patients use. | Clinical judgment. The pipeline writes about what you have approved saying. It does not decide what care a reader should seek. |
| Gates | Fact lock, on-page checks, readability, duplication ceiling. Violations are repaired to match the registry or the draft is refused. | Deciding to change a rule. The gates never relax themselves and no plan turns them off. |
| Review | The approval screen assembles the draft with every registry-governed claim marked and traced to the entry it matches. | The signature. Your reviewer approves, requests changes, or refuses, on every publish. |
| Publishing | Snapshot of the live page, publish to WordPress, one-action rollback. | Choosing to roll back, and choosing when a page comes down. |
| Measurement | Daily rank tracking on your tracked keywords, Google AI Overview citation state on the daily citation set, technical site audit, monthly backlink opportunity report. | Deciding what to do about what it shows. |
Read the table as a boundary rather than as a division of labor. The left column is work that can be done identically a thousand times without a person losing attention on the nine hundredth. The right column is work where a person’s judgment is the product. A vendor that automates something from the right column is not faster than this; it is publishing without the thing that makes the publishing defensible, and in health content that difference does not surface until it surfaces on a page a patient is reading. The seven step version of the same pipeline, gate by gate, is on the how it works page.
Who approves each article, and what do they actually see?
Someone at your practice, on every publish, with no setting that removes the step. That much is easy to say and every tool in this category says something like it, which is exactly why it is worth being precise about what the approver is looking at, because that is where the difference actually lives.
What is in front of the approver
The full draft as it will render. Every registry-governed claim marked in place and traced back to the specific entry it matched. The target keyword. The internal links the page will carry, in and out. The on-page and readability readings. The cannibalization check against pages you already have.
Three, and only three
Approve, request changes, or refuse. A refused draft does not go into a queue to be nudged through later. A change request goes back through the same gates, so a revision cannot introduce a claim the first version was not allowed to make.
Approval is not the differentiator
Every serious tool in this market ships a draft mode and lets you approve before publishing. The thing that is different here is not that you get to approve. It is that we cannot publish a claim about your practice that you have not approved, so the sentence you would have had to catch never gets written.
That distinction is worth holding onto when you are comparing options, because “you approve everything” is available everywhere and tells you nothing about what you are approving. If the draft in front of you was written with no constraint on what it could assert, your approval is the only thing standing between a plausible sentence and a published one, and you are back to the fatigue problem. If the draft was written under a constraint you set, your approval is a second line rather than the only line. Both arrangements produce a screen with an approve button on it. They are not the same product.
In practice the approver at a medical practice is usually not the marketing coordinator. It is a clinician, a practice manager with clinical background, or in larger groups a marketing lead who routes anything clinical to a physician before signing. That person is expensive and hard to convene, which is the real reason the pre-approval gates matter: their time is the scarce resource, and the pipeline is built to spend as little of it as possible on catching things a machine can catch.
What does Google’s own guidance say about health content?
It says health topics get more weight placed on trust signals, and it names the category. This is the one quotation on this page, taken from one page, read on one date, so you can go and check it yourself rather than take a marketing paraphrase of it.
“For example, our systems give even more weight to content that aligns with strong E-E-A-T for topics that could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society. We call these ‘Your Money or Your Life’ topics, or YMYL for short.”
GOOGLE SEARCH CENTRAL, “CREATING HELPFUL, RELIABLE, PEOPLE-FIRST CONTENT” · READ 2026-09-04 · developers.google.com/search/docs/fundamentals/creating-helpful-content
Three things follow from that sentence, and it is worth separating what it says from what it does not.
It says health content is inside the category Google weights most heavily for trust. It does not publish a score, a threshold, a percentage or a formula, and this page will not invent one. It also does not say that automated content is disallowed; the same document discusses how content was produced as something worth telling readers about, and it explicitly declines to endorse writing to a word count. What the guidance does establish is that the burden on a page in this category is a burden of trust, and trust is made of specifics that hold up when someone checks them. That is a description of a registry problem, which is why the mechanism on this page is shaped the way it is. The broader treatment of what that framework asks for across every regulated vertical is on the YMYL SEO automation hub.
Sitting on top of Google’s guidance are the bodies your practice already answers to: your state medical board, your malpractice carrier, whoever reviews your advertising, and the privacy obligations that govern anything touching a real patient. This page does not tell you what any of them require, and no page here will. Your practice knows that already. The registry is where those decisions stop being a memo in a shared drive and start being enforced on every draft.
What does this never automate for a practice?
Four things, permanently, on both plans. These are not upgrade paths and there is no configuration that changes them.
- Clinical judgment. The pipeline writes about what your clinicians have approved saying. It does not decide what a patient should do, it does not compose medical guidance, and it does not reason from a model’s general knowledge to a statement published under your practice’s name.
- Unattended publishing. Every publish stops at your reviewer. There is no plan on which that step is removed and no setting that skips it. If nobody at your practice approves anything for three weeks, nothing publishes for three weeks.
- Patient content. No draft writes a testimonial, a case study, a patient quote or a review node. Patient material comes from the practice, with your consent process, or it does not exist on the site.
- Link exchanges. We do not run a backlink network and we do not trade links between clients. You receive a monthly report of opportunity domains and the outreach targets behind them. Those are opportunities found, never links delivered.
What is not built yet, and what is deliberately not sold?
A vendor page that only lists what a product does is telling you half of what you need. Here is the other half, so that you find it now rather than in month two.
Does not exist today
- Publishing to any platform other than WordPress.
- Prompt tracking inside ChatGPT, Perplexity or Gemini. Google AI Overview citation tracking is what exists.
- Reddit or Quora participation of any kind.
- Content in any language other than English.
Each of these is listed because it does not exist, not because it is arriving. If one ships, it gets added here afterward rather than promised here beforehand. Early access to new capabilities on the Scale plan means a new capability reaches Scale accounts first once it exists; it is not a commitment that anything on this list is being built.
Exists elsewhere, but is not part of either plan
- Google Business Profile management, posts and the local pack work that sits around it.
- Review generation, review response, and reputation management of any kind.
- Paid search and paid social.
- Website design and development. Web design is never part of Launch or Scale.
These matter for a practice and several of them matter more than content in a tight local market. They are simply not what these two plans are.
If the map pack is where your patients actually come from and your website is a formality, the honest answer is that this product is not the first thing you should buy. That is worth knowing before a card is charged rather than after.
How do condition, procedure, provider and location pages differ?
A practice site is mostly four shapes repeated. Each shape has its own way of going wrong, so each carries its own registry constraints and its own review emphasis.
Condition pages
The page a patient reaches before they know what they need, usually from a symptom query rather than a procedure name. It exists to answer the question in the language they typed and to make clear that your practice treats this. What it may state about the condition comes from your clinicians. What it states about you comes from the registry: which providers treat it, at which locations, through which of your approved services. It carries no prognosis, no statistics you have not approved, and no instruction to the reader about their own care.
Procedure and service pages
The page that has to convert. Named service, approved description, the providers who perform it, the locations that offer it, and the questions patients ask before booking. This is the page that most often accumulates quiet errors over years, because a service moves offices, a provider leaves, a device is replaced, and the page keeps saying what it said three years ago. Registry changes apply to every draft written afterward, and the technical audit surfaces the already-published pages that still need a human to edit them.
Provider pages
The highest risk page on a practice site, and also the one that does the most to help a search engine work out who stands behind everything else on the domain. Name, degree, title, credentials, certifications, affiliations, languages and locations, each an approved entry, each written the same way everywhere it appears. A provider page here is assembled from the registry rather than composed around it, which is precisely why it cannot quietly gain a fellowship between drafts.
Location pages
Address as published, the providers who see patients there, the services available there, the hours, and the surrounding area terms patients actually search. This is where multi location practices do themselves the most damage, because five near-identical pages compete with each other and none of them ranks. The cannibalization check runs before a new location or service page is proposed, so the build does not manufacture the problem it exists to solve.
How does a multi-location practice stop its own pages competing?
By checking before building rather than auditing after. Keyword clustering is built from live search results, not from a keyword tool’s grouping, and a proposed page is compared against what the site already ranks for before it is written.
The mechanism is worth describing because it is the part practices most often get wrong on their own. A practice with four offices and eleven services is tempted to build forty-four pages. Search engines will treat most of those as near-duplicates of each other, pick one more or less at random, and leave the rest to accumulate crawl budget and confusion. The result is not four times the traffic. It is a site where the strongest page for a term keeps changing and none of them earns authority.
So before a page is proposed, three things are read: what the site already has published, what those existing pages already rank for, and what the pages currently ranking for the target term actually contain. If an existing page already answers the query, the recommendation is to strengthen it rather than to build a sibling. A duplication ceiling then runs on the draft itself, so a page that survives the planning check but comes out too similar to an existing one is refused at the gate.
This is also the reason article volume is stated as a ceiling rather than a quota. The plans allow up to 30 articles per month. If the right answer for your site this month is eleven articles and two rewrites, that is what should happen, and a vendor that ships thirty regardless is optimizing for the invoice rather than for the site. Volume is capacity, not a target.
What gets measured, and how soon can you see it?
Measurement starts in week one, well before rankings move, because the point of it is to make the question of whether anything is happening a matter of record rather than a matter of opinion.
Rank tracking
Daily positions on your tracked keywords, with history, so a condition term moving from page three to page one is visible while it is happening rather than in a quarterly summary. Launch covers up to 25 tracked keywords. Scale covers up to 75.
Google AI Overview citation tracking
Whether Google’s AI Overview appears for a query and whether it cites your practice. A rank read is cheap and a citation read is a heavier request, so the daily citation set is deliberately the smaller of the two, and the split is stated below. Google AI Overviews specifically, never a sample of prompts inside another assistant, and nothing else is claimed.
Technical site audit
Broken links, redirect chains, structured data state, internal link structure and index status. On Scale, Search Console integration adds index verification and query mining from your own traffic rather than from a third party estimate.
For scale, one client account currently sits at 105 keywords rank-tracked, 33 under daily SERP and citation research, as recorded 2026-09-02. The two numbers differ on purpose: a rank read is cheap and broad, a citation read is heavier, so the daily citation set is the smaller one and we say which is which. The monthly backlink opportunity report on that same set of client accounts found 231, 99 and 288 gap domains respectively, read 2026-09-02: those are domains linking to two or more of a client’s competitors and not to that client. They are opportunities found. They are not links delivered, and nothing on either plan delivers a link.
No timeline is promised on this page. Search results depend on your market, the practices competing with you, and the state of your site on the day you start, and any vendor quoting you a month number for a medical vertical is quoting you a feeling. What is promised is that the measurement is visible to you from the first week, so the absence of movement is as observable as its presence.
What happens when a published page turns out to be wrong?
It gets rolled back in one action, because a snapshot of the live page was taken before the change that published it. That is worth stating plainly, because “we take backups” and “you can undo this specific page in one step” are very different promises.
There have been over 300 page deployments across four client sites since August 2026, every one preceded by a snapshot. The snapshot is not an optional flag someone remembers to set. It is the step before the write, which means the recovery path exists whether or not anyone anticipated needing it.
The second half of the answer is what happens to the cause rather than to the page. A wrong claim that reached publication means either the registry entry was wrong, or the claim was outside the registry and the lock did not classify it correctly. Both of those are fixable in a way that applies to every future draft: correct the entry, or add the category rule. A correction that only fixes the one page leaves the pipeline free to make the same mistake next month, which is the difference between fixing an article and fixing a system.
What does the first month actually look like?
Mostly registry work, and the practices that get the most out of this are the ones that treat the registry as the deliverable rather than as paperwork standing between them and articles.
Registry assembly
Your providers, their credentials and the exact wording for each. Your services as they stand today, with discontinued ones marked. Your locations, hours and per-location service availability. Your approved numbers with their sources. Your banned phrases and your category rules. This is the part that needs a clinician’s attention, and it is the only part of the engagement that does.
Technical site audit and keyword research
The site is audited: broken links, redirect chains, structured data, internal links, index status. In parallel, keyword research and clustering are built from live search results, and a cannibalization check is run against what you already have published, so the first article proposals are not competing with pages you already rank for.
The first drafts, and the first refusals
Expect refusals early. A first-month registry is always incomplete, and every refusal is telling you about a gap in it. Practices that read the refusals and add entries end the first month with a registry that holds, and a pipeline that stops arguing with them. Practices that ask for the gate to be loosened get a system that no longer protects them, which is why it does not loosen.
Approval rhythm
Whoever signs needs a repeating slot in the week. The single biggest cause of a slow first month is not drafting speed; it is drafts sitting in a queue because the only person allowed to approve them is in clinic. On Scale, the monthly strategy session is 60 minutes and is delivered personally; support responses come within one business day, which is explicitly not same-day.
What does this cost?
Two plans, both published in full, no call required to see them. Annual is two months free.
or $1,000 per year, equivalent to $83 a month.
- Up to 30 articles per month
- Up to 25 tracked keywords
- Fact registry and fact lock on every draft
- Human approval before every publish
- Auto-publish to WordPress, snapshot and one-action rollback
- Daily rank tracking with Google AI Overview citation tracking
- Full technical site audit
- Monthly backlink opportunity report
or $2,500 per year, equivalent to $208 a month.
- Everything in Launch
- Up to 75 tracked keywords
- A 60-minute monthly strategy session, delivered personally
- Support response within one business day, which is not same-day
- Competitor teardown reports
- Search Console integration
- Early access to new capabilities as they ship
- For businesses and agencies running three or more sites
Cancel anytime. Your plan runs to the end of the period you have paid for. Annual billing is available for both plans at $1,000 and $2,500 per year on the pricing page, where all four options are wired to checkout. Nothing on either plan removes the approval step, and web design is not part of either plan.
How is this different from a monthly agency retainer?
In cost, in throughput, and in where the accuracy risk sits. Those are three separate differences and only the first one is usually discussed.
A conventional medical marketing retainer buys you a person’s attention for some number of hours a month, and the output is limited by those hours. That has a real advantage: the person accumulates context about your practice, and after a year they know which of your surgeons hates the phrase “minimally invasive”. What it cannot do is exceed the hours you bought, which is why output and price move together under that model.
What a retainer does not usually give you is a mechanism. The accuracy of the work depends on the individual doing it and on how carefully they were briefed, which means it degrades when they are busy, and it resets when they leave. Nothing about the arrangement makes it structurally impossible for a wrong credential to reach a published page; it just makes it less likely while the right person is paying attention.
The trade here runs the other way. You get throughput and you get a constraint that does not get tired, and in exchange you take on the registry work yourself, once, up front. If your practice is unwilling or unable to assemble and approve that registry, this product will underperform a good retainer and you should buy the retainer. That is not a sales concession; it is the actual condition under which the mechanism works.
What has the gate actually caught, and what has this ranked?
Every reading below carries the date it was read and the base it was read from. The first item is a set of incidents rather than a reading taken on one day, so it carries a base and no date. None of it is from a medical practice, because no medical practice is a client of this product yet, and inventing one would be the exact failure this page is about.
What the gate refused
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. Read those for what they are. Two of the three are legacy copy that predates the gate, on a client site and on our own site, and the third was caught in testing. That is a gate doing its job on real pages, not a claim that a model fabricates constantly.
Position 1 holdings, read 2026-09-02
A national records platform holds position 1 for “nationwide property search”. An elevator safety company holds position 1 for “elevator smoke testing Portland”. Both read 2026-09-02. Both are non-brand commercial terms in verticals with their own accuracy constraints, which is the reason they are cited here rather than a medical result that does not exist.
Position 2 and 3 holdings, read 2026-09-02
Four keywords at position 3 on a national asset records platform: “nationwide asset search”, “tampa asset search”, “denver asset search”, “minneapolis asset search”. Four at position 2 across two other clients: “title search online” and “preliminary title report online”; “elevator modernization bid review” and “elevator safety inspection”. All read 2026-09-02.
Deployment and rollback
Over 300 page deployments across four client sites since August 2026, every one preceded by a snapshot. A failed deploy is not counted as a deployment. The snapshot count is the point rather than the deployment count: it is the reason a wrong page is a one-action problem.
What is not claimed here, deliberately: no fleet-wide citation ratio, because most tracked rows have never had a citation check run against them and a ratio computed over rows that were never checked is not a ratio. No claim that the gate finds fabrications continuously, because the measurements we have do not support one. No rating, no review count, no testimonial, and no aggregate score of any kind on this page or in its structured data.
Where does the same pipeline run under other rulebooks?
The mechanism is one mechanism. What changes between verticals is the content of the registry and which category rules are switched on.
YMYL SEO automation
The hub: why regulated categories are held to a different standard, what the fact lock refuses across all four verticals, and where the line between automated and human sits.
Fact-checked AI SEO
How the fact lock works in detail, what it does with a violation, and the honest account of what it cannot see.
SEO automation with editorial control
The seven things to demand before you let a machine publish, and the test for each.
Law firms
Bar advertising rules, practice area pages that cite only approved authority, and attorney bios built from the admissions in the registry.
Financial advisors
Performance figures, registrations and disclosures, where an unchecked number is a regulatory problem rather than a marketing one.
Licensed home services
License numbers, service areas, warranty terms and prices that are true on the day they publish, across a lot of service and city pages.
The pipeline, step by step
Every gate in order, what each one checks, and what happens to a draft that fails one.
Medical SEO services: the questions practices ask
Can SEO be automated safely for a medical practice?
Parts of it can. Research, competitor scoping, drafting against an approved registry, on-page and readability checks, publishing with a snapshot, and daily measurement are automated. Clinical judgment, the content of the registry, and the approval of every publish stay with your practice. The fact lock enforces that boundary on every draft rather than leaving it to a reviewer to catch. Nothing here is a claim of compliance with any regulation.
How do you stop an AI article from inventing a credential or an outcome?
Credentials, certifications, providers, services, locations and numbers exist in a registry your practice approves. Every draft is checked against it, and a claim that is not in the registry is repaired to match it or the draft is refused. There is no path on which an unapproved credential or outcome figure reaches a published page, and your reviewer still signs before anything goes live.
Who approves the content before it publishes?
Your reviewer, on every publish, with no option to switch it off. The approval screen shows the full draft as it will render, every registry-governed claim marked and traced to the entry it matches, the target keyword, the internal links, the on-page and readability readings, and the cannibalization check. They approve, request changes, or refuse.
Will the pipeline write clinical content or medical advice?
No. It writes about what your clinicians have approved saying. Statements about conditions, treatments and what a reader should do come from your practice and enter the registry the same way every other approved fact does. Anything that would read as advice to an individual patient is refused rather than drafted and caught afterward.
Do you write patient testimonials or add review markup?
No. No draft composes a patient quote, a case narrative or a sentiment attributed to your patients, and no review or rating structured data is generated for you. If your practice publishes patient content, it comes from the practice with whatever consent your process requires.
Do you track ChatGPT or Perplexity for health queries?
No. Google AI Overview citation tracking is what exists today: on the keywords in the daily citation set, whether Google’s AI Overview appears and whether it cites your practice. That set is deliberately narrower than the rank-tracked set, and on one client account the split is 105 keywords rank-tracked, 33 under daily SERP and citation research, as recorded 2026-09-02. Tracking of other AI answer engines is not offered and is not claimed anywhere on this site.
How many articles and keywords does each plan cover?
Both plans cover up to 30 articles per month. Launch covers up to 25 tracked keywords and Scale covers up to 75. Those are ceilings rather than quotas: if the right answer for your site in a given month is fewer articles, that is what should happen.
What does it cost, and can we stop?
Launch is $100 per month, or $1,000 per year. Scale is $250 per month, or $2,500 per year. Annual is two months free. Cancel anytime. Your plan runs to the end of the period you have paid for.
Do you manage our Google Business Profile or our reviews?
No. Neither plan includes Google Business Profile management, review generation, review response or reputation management, and neither includes paid media or website design. If the map pack is where your patients actually come from, this product is not the first thing you should buy.
What happens if a published page turns out to be wrong?
A snapshot of the live page is taken before every change, so the page is rolled back in one action. The second step is to fix the cause rather than the page: either the registry entry was wrong and gets corrected, or the claim was outside the registry and a category rule gets added, so the same mistake cannot recur next month.
Updated September 6, 2026