Our Website Is AI-Generated. So What?

Yes, a large portion of this site was scaffolded with generative AI.

It does resemble a bunch of other AI-born sites. Familiar hero, feature grid, pricing table, testimonial-ish shapes.

So what?

There are deliberate reasons behind that choice. Not a mistake, not laziness, not surrendering craft.

It’s a case study in pragmatic AI adoption inside an early-stage company.

Our initial thought: skip it-just a tiny static HTML page with contact info. Regulatory and credibility realities nudged us toward a “real” site: something that states what we do and who we are.

A "standard" marketing site runs roughly €3k–€15k (≈$3.2k–$16k) depending on scope and polish. For a company that started with €10k in seed capital, burning a chunky slice of that on brochure aesthetics felt... questionable.

So we built it ourselves, with AI accelerating the boring bits-to compress cost and cycle time.

And where to start if not with Lovable?

If you haven’t tried it: Lovable takes a prompt and spits out a full web page-HTML, CSS, JS (they’re leaning into backend now too). Super fast path to “something you can click.”

The first output? Good enough for our humble bar. So we pushed further. Then we hit the wall.

Tiny tweaks became weirdly expensive. Generated code was tangled. And: React (not our current stack preference).

Add $20/mo just to attach a custom domain.

We walked away. Rebuilt from scratch-still AI-assisted-without vendor gravity.

We already maintain templates that spin up Single Page Apps: hydrate once, route client-side, minimal reload friction.

We fed our AI copilots the same prompts we had used inside Lovable. Predictably: similar visible structure.

But this time the invisible part was ours to refactor: simpler, cleaner, maintainable. No credit meter ticking.

Using our base deployment template, shipping to the web was trivial. Hosting cost: effectively zero (our own server).

True cost so far? Our time-maybe half a day. Internally costed: call it €500 of opportunity cost (arguably we’d have burned at least that coordinating an external freelancer anyway).

Classic SPA drawback: SEO discoverability. Bots can index more JS today, sure, but guaranteed, structured, prerendered HTML still wins for reliability and speed-to-index.

Solution: Nuxt (Vue-based), generating either statically or via SSR. Content lands in HTML at response time. Crawlers smile.

Yes: higher initial complexity than a vanilla SPA. Luckily one of us had prior Nuxt mileage, so ramp cost stayed low.

AI copilots kept bridging repetitive glue: same prompts we used for Lovable, now producing cleaner, framework-aligned code.

End result: SEO-friendly, fast, portable, not locked to a paid walled garden.

Nuxt’s plugin ecosystem gave us quick wins in security hardening and content features-like the blog you’re reading.

We spent:

• €0 on external consultants
• €2 on paid dev tooling (GitHub Copilot)
• ~€3000 internal time (aggregate, including later polish)
• Fractions of a euro on hosting (existing infra)

We got:

• A complete, performant, SEO-conscious marketing site
• A codebase we can evolve without vendor friction
• Zero dependence on subscription website builders
• A replicable template for future launches

Original? Maybe not. Effective? Yes.

Predictable patterns reduce cognitive friction: hero → value props → proof → CTA. Users already “speak” this structure.

Commodity layout means we iterate on copy & offers without re-inventing shells. With the template, we can spin up a fresh vertical site in hours.

More time to hunt the perfect GIF for a closing beat.

Every euro NOT burned on ornamental originality goes into:

• Infrastructre stability
• Real product capability
• Acquisition experiments
• Support & retention loops

Originality has a cost. If originality isn’t the product, it’s often vanity. We don’t fund vanity; we fund traction.

Looking like others isn’t a flaw if those others already converged on functional patterns. Reinventing obvious sections is wheel re-invention: slower, worse, pricier.

We used AI for scaffolding & acceleration:

• Draft copy passes (human-trimmed where nuance matters)
• Layout skeletons
• Repetitive integration glue
• Initial content taxonomy

Humans owned the substance: prioritization, value prop, pricing stance, message tone. AI did non-differentiating labor.

Exactly its lane.

The site being AI-generated is a feature of operational discipline: remove waste, shorten loops, focus on leverage.

We’ll earn the right to be “fancy” later. For now: clear, fast, lean, intentional.

If this resonates, welcome-you’re among friends.