Why "I built it with AI" is three different builds (pure generated code, CMS with an AI layer, headless + AI front end) with three different risk profiles
Path one's trap: no content layer, so editing isn't editing anymore, it's regenerating the whole site from scratch
The security roll of the dice: Veracode found roughly half of AI-generated code came back secure (55% one year, 56% the next) on blind prompts with no security instructions
Your website is a regulated surface: Ontario's AODA and WCAG 2 AA, the US ADA, and why "I built it in a weekend" is not a compliance strategy
Path two's real risk isn't what the AI writes, it's handing a model write access to your live production site instead of staging with a human approving
Why the difference was never the tool, it's the operator: knowing which defaults to override, how to isolate a change so one button doesn't rewrite the whole site
"Make me a website" versus a real pipeline, and why garbage in still means garbage out
The one test to run this week: can anyone in the loop actually validate what came back?
📖 Chapters:
0:00 — We lost the project: eight months, no website
0:59 — The question: what do you actually walk away with?
1:34 — Three paths, and they're not equally dangerous
1:37 — Path one: pure generated code, security and legal exposure
4:01 — Path two: a real CMS with an AI layer
5:07 — Path three: headless CMS and AI front end
6:26 — The difference was never the tool, it's the operator
8:52 — "Make me a website": what that actually means
10:52 — What to do this week: can anyone validate it?
11:31 — Where AI actually belongs
🌐 Tennis web design & development services: https://www.designtennis.com/services/website-design-and-development
Podden och tillhörande omslagsbild på den här sidan tillhör
tennis. Innehållet i podden är skapat av tennis och inte av,
eller tillsammans med, Poddtoppen.
