Bob Michaels/ai
All projectsCase study 09 of 10

A rebuild a regulator has to read

Wealth-management site rebuild, North Texas

A 110-page WordPress site nobody could find, rebuilt as a new site built to be found and checked.

I rebuilt the whole site, wrote every page against search and AI research, and checked each one against the firm's compliance rules. The client chose the design direction. I built it.

When
Aug to Nov 2026
Status
Built; in client and compliance review
old web addresses checked after the rebuild, none failed
654
home page grade from the AI grader, up from 56.5 on the same grader
84.2
old pages showed almost nothing to AI crawlers
109
chance the old site was named in local AI answers, 17th of 104 firms
5%

What this does not claim

Not live: the site is in client and compliance review, and the old site still serves the public. No traffic, ranking or lead results exist yet. The numbers are grades from an AI grader and redirect checks, not business results.

Before and after, page by page

Same pages, old site on the left, rebuild on the right. These are the real pages, unedited.

HomeThe old home page was a photo and a two-word headline. The new one opens on a video hero, a service grid and a numbered list of the five situations the firm serves.

Before: the existing site

The old home page of the advisory firm: a photo, a two-word headline and grey boxes.

After: the rebuild

The rebuilt home page: a hero, a quote band, a six-card service grid and a numbered list.
RetirementThe old page was a headline and a short list of questions. The new one opens with the answer, then question headings, short lists and numbered steps.

Before: the existing site

The old retirement page: a headline and a short list.

After: the rebuild

The rebuilt retirement page for pre-retirees and retirees.
Investment managementA bare list with no hero and no call to action became a page that answers first and ends on an appointment button.

Before: the existing site

The old investment page: a bare list.

After: the rebuild

The rebuilt investment management page.
Office pageFourteen city pages were one page with the town swapped. The new office page carries the address, phone and hours.

Before: the existing site

The old city page for the Plano office.

After: the rebuild

The rebuilt Plano office page with an address and hours card.
Home on a phoneFirst screen at phone width: the new one leads with the headline, a summary and the appointment button.

Before: the existing site

The old home page on a phone.

After: the rebuild

The rebuilt home page on a phone.

Select any screenshot to open it full size.

01

The problem

The old site was a WordPress site with 652 addresses serving 110 real pages. It earned 156 organic visits a month and ranked for August birthdays, not for retirement planning. 66 pages had no h1. Six pages were unedited chatbot output, and 14 city pages were one page with the town swapped. Several pages described the firm's fee model in ways its own other pages contradicted, which is a compliance problem, not only a content one.

02

What I built

A custom site in one design system, 48 pages in the sitemap. I researched what people ask AI assistants, wrote each page to answer first, and ran every page through rule packs for an adviser that is also a broker-dealer representative. Unsupported sentences were deleted, not reworded. AI agents did the drafting and checking under my direction. I looked at every page at full size and fixed what a test cannot see: dead half-screens, broken tables, lone words.

03

What it produced

Rewritten pages graded 64 to 90 against an old-site median of 36. Every one of 654 legacy addresses was checked and none failed. The site is built and in client and compliance review. It is not live.

Measured from the repository

564 commits, my largest workspace. 654 old addresses checked, 0 failed.

The stack

  • Next.js
  • TypeScript
  • Tailwind
  • Vercel
  • AI page grader
  • Compliance rule packs