We run every client build the same way. Research first. Check every claim. Label what's verified and what's someone's memory. Then build.
This month I finally ran that process on us. It was humbling, and I learned things about my own story that I didn't know.
Aha 1: The menu was trying to sell everything
Our old menu had a little of everything: a $95 LinkedIn fix, a working session, a rebrand sprint, and a box marked "something else." Every one of those was real work. Together they looked scattered.
When we stepped back, the thread was obvious. I'm a coach. So is Nate. We help one person, then their team, then their business. That became the menu: for you, for your team, for your business, built on three practices (Teach, Brand and Systems), with one front door, the working session.
Aha 2: I'm not a web designer
The site called me a self-taught web designer. I'm not. I learned HTML because I needed a job and grad school wasn't going to pay for itself. In college I worked three jobs: Rite Aid in Perrysburg, webmaster for the BGSU Graduate College, and general manager of WFAL, the campus radio station.
When I introduce myself to sales teams, I say I'm a sales person first, or a self-hating marketer. They love it. It's true. And it's a better description of what I bring than anything the site said.
Aha 3: The research corrected my own story
We ran a public-records check on the whole team using our TEACH method. Every finding got a label: Verified, Probable or Unverified.
The site said I learned HTML at twelve. I was nineteen. It said I built my first site on a school computer with a beige tower. It was an orange iMac, freshman year. I was one of the few Mac kids on campus. It said I spent ten years at my last job. It was about one.
None of those came from me. They came from an AI draft, written months ago, that filled the gaps with details that sounded true. Then they got copied from page to page until nobody remembered where they started. My favorite: a guestbook signed by "my mom, my best friend, an aunt, and one stranger who told me my marquee was too fast." It never existed.
Aha 4: The man who taught me HTML was a Bowling Green radio guy
I learned HTML from HTML Goodies. I still go back to it when I can't get a bulleted list to behave.
What I didn't know until this week: Joe Burns started it as a handful of tutorials for a class he taught at Bowling Green. Before that he spent twelve years as a morning DJ and radio programmer, and his dissertation was on radio airplay. He'd left Bowling Green before I got there. I ran the campus station a few years later, learning from what he left behind.
Somebody shared it with me first. That's the whole point of the Lab.
Aha 5: I found the real thing
After all that invented history, I went looking and found the actual site: our 2002–2003 MOD Cohort page, on my old BGSU account. I was the Pagemaster. It has an APA cheat sheet, gold drop caps and a lot of clip art. I also found a community survey I built for a housing nonprofit and a mental health agency exploring an alliance. That's the same kind of work I do for healthcare clients now.
The guestbook was fake. This one is real.
Aha 6: Our stylesheets were fighting
The site felt inconsistent page to page, and it was. Every page loaded two stylesheets, an old dark theme and the current one, and they argued. We merged them into one, removed about 95 rules nothing used, and every page now pulls from the same system.
Aha 7: The research tool wandered until we used TEACH
Our first research run kept drifting. It wrote career stories and next steps we didn't ask for, then asked for more credits. So we rewrote the prompt in TEACH order and told it to research, return one table and stop. Then we ran all six people in a single pass, with the AI's research mode turned on. The result was accurate enough to change our bios, and honest enough to tell us which claims didn't hold up.
Aha 8: Verified and firsthand are different claims
Some of the best parts of our story can't be found online. That's normal. Most real work isn't indexed. The fix isn't to drop it. It's to say whose account it is. Our bios now use what's verified, and what's firsthand is written as ours.
The prompt we used
It works in any AI model that can search the web. Turn on web search or research mode, fill in the details in brackets, and run it on yourself before you rewrite your About page.
T · Summon
You are a research librarian and professional-records researcher. Verify the public career record of the person below, and show what can and can't be proven.
E · Divine
In 3 bullets or fewer, name the best sources for this person's field. Don't wait for my reply.
A · Cast
Person: [name, other names, city]
Verify each claim: [roles, employers, years, schools, awards]
Public professional sources only. A search snippet is not proof. Open the page.
C · Ward
Label every finding Verified, Probable or Unverified. Remove namesakes and say who you excluded.
H · Inscribe
One table: claim | status | source | year | note. Then three lines: strongest proof, what you couldn't verify, namesakes excluded. Then stop.Run it, and then do the part no tool can do for you: ask where every fact came from.
— L.
