Process
Diagnose, prescribe, build, monitor.
The same four phases every time, in the same order, because the order is what stops the work from being guesswork.
The short answer
Every project runs through four phases. Diagnose takes about a week, Prescribe one to two, and Build two to three, so a Rebuild runs four to six weeks from start to launch, and longer if content approvals are slow. Diagnose is intake plus a measured audit. Prescribe is the plan, agreed in writing before any building starts. Build is the brand and the site, with speed, structured data and accessibility designed in. Monitor is the monthly measurement afterwards, including a fixed list of patient questions re-run across the AI assistants so visibility is tracked rather than assumed.
The phases are how a project runs. The five steps on the services page are what you buy, and they map onto these: Diagnose is the Credibility Audit, Prescribe is Foundations, Build is the Rebuild and the automation behind it, and Monitor is the monthly partnership.
Phase 1 · 1 to 2 weeks
Diagnose
Nobody prescribes before they diagnose.
It starts with intake, not with your website. What the practice actually does, who actually pays, which services carry the margin, who you lose work to, and who you will not serve. Then the measurement, across everything a patient can already find: a crawl of every reachable page, the Google Business Profile and the directories, the social accounts and what they actually publish, the reviews and how they are answered, structured data checked against what is visible, mobile performance at the 75th percentile, a word-by-word claim scan, and a fixed set of patient questions run live across ChatGPT, Perplexity, Gemini and Claude.
| What is needed from you | Completed intake form, read access to Search Console and Google Business Profile if they exist, the handles of the channels you actually use, and the name of the person who approves public claims. |
|---|---|
| What you receive | The Credibility Audit: recorded walkthrough, five-axis scorecard, three prioritised fixes, compliance flag list, your channels read the way a patient reads them, and the AI answers recorded verbatim. |
| How long | 1 to 2 weeks |
Phase 2 · 2 to 3 weeks
Prescribe
The plan is agreed in writing before anything is built.
Positioning and message, then the site plan: one search intent per page, the sitemap, what each page has to answer, which schema each page carries, and where the evidence for every claim comes from. Claims without evidence get weaker copy rather than a better adjective. This is also where the honest arguments happen, while changing your mind still costs nothing.
| What is needed from you | Two working sessions, your credential and licence details, testimonial permissions in writing, and decisions on anything you want said that the evidence does not yet support. |
|---|---|
| What you receive | A positioning and messaging document, a sitemap with the intent for every page, a content plan, a schema plan, and a claim register showing the source behind each claim. |
| How long | 2 to 3 weeks |
Phase 3 · 4 to 5 weeks
Build
Speed, structure and accessibility are built in, not added at the end.
Brand system and site, together. Every colour pair is contrast-checked and the results are recorded. Images are sized and served in modern formats. Fonts are self-hosted. Structured data is generated from one source of truth so no two pages can contradict each other. Everything is reviewed against the launch checklist before it goes live, and the claim pass is done by a human, not by the machine that wrote the draft.
| What is needed from you | Content approvals in batches, photography or a decision to art-direct a local shoot, and a final sign-off on every health claim and testimonial. |
|---|---|
| What you receive | The live site, the brand usage guide, the schema, the launch QA record, and the measured before-and-after on speed, contrast and structured data. |
| How long | 4 to 5 weeks |
Phase 4 · Ongoing
Monitor
Visibility is measured monthly, not assumed.
Search Console and Bing Webmaster Tools both matter, because ChatGPT retrieval leans on Bing. Core Web Vitals and broken links are checked monthly. The same fixed list of 20 to 30 patient questions is re-run across the assistants every month and the answers are diffed, so drift shows up as a trend rather than as a surprise. Judgement is made over three to six months, not on one bad week.
| What is needed from you | An hour a month, and a reply when something needs a human decision. |
|---|---|
| What you receive | A monthly report in plain language: what changed, what it means, what is being done, and what needs you. |
| How long | Ongoing |
Who decides what, and why it is written down.
Software researches, drafts, formats, prepares structured data and runs the monthly measurement. A human approves every credential, every health claim, every testimonial, every patient outcome and every line of public copy. That boundary is not a preference. For a regulated practice it is the difference between a marketing mistake and a compliance problem, and it is the reason the drafting tools here have no ability to publish anything.