Ask AI about your site once the project is "ready". That's where the part nobody
shows on YouTube begins.
The domain and the email that never arrives
You connected the domain, uploaded the files, and the page loads. Nice. Now send a test
message from the contact form to a Gmail address. Did it arrive? Or did it land in spam
because nobody configured SPF, DKIM and DMARC? I've seen businesses lose customers for
months over this and only find out when someone said "I emailed you and you never replied".
Payments don't forgive
Taking money is easy until this happens: the customer pays, the webhook fails, the order
never gets marked as paid, and you have the money with zero records. Or worse: the user
double-clicks "pay" and you charge them twice. Or they ask for a refund and there's no
button for that. A payment gateway isn't a pretty form — it's a system with states,
retries, idempotency and reconciliation. AI gets you 60%. The other 40% you have to know.
The security nobody tests
A form that writes straight into the database without validation is an open door. A file
with the server keys pushed to a public repo is an open door with a sign that says
"please come in". AI generates code that works; whether it's also secure
is a different conversation, and it doesn't always make it into the answer.
Technical SEO, which is boring and decides everything
A site looking good doesn't mean Google understands it. Canonical tags, hreflang, sitemaps,
301 redirects, structured data, real load time on a mid-range phone with bad signal. This
very site you're reading needed a bilingual architecture with separate URLs — otherwise
Google received two different contents at the same address and ranked neither. That's not
a prompt.
The 90% rule. AI takes you from 0% to 90% in no time.
That last 10% is 90% of the work: it's where a project stops being a demo
and becomes something that survives real people using it.