RESEARCH · FRAMEWORKS

Next.js vs Astro vs Hono: which one are AI builders actually generating in 2026?

We sampled 200 AI-generated projects across the major builders. The framework distribution is not what marketing pages suggest. Here's the data.

8 min readApril 2026Article
RC
Riley Chen
Researcher · Side-by-sides & benchmarks
TL;DR

We sampled 200 projects scaffolded by 6 popular AI builders. Next.js still leads (54%), Astro is rising fast for content/landing (18%), Hono is taking the API/edge slice (14%). SvelteKit, Remix and "other" make up the rest. The interesting question isn't who's #1 — it's why the AI defaults are starting to disagree with what humans pick.

The sample

200 projects, generated April 2026, across:

Each prompt was one of: "build a SaaS", "build a landing page", "build an API", "build an internal dashboard". Default outputs only — we didn't push the tool toward a specific stack.

The headline distribution

Framework% of 200 projects
Next.js (App Router)54%
Astro18%
Hono (Workers / Bun)14%
SvelteKit6%
Remix4%
Vanilla / other4%

By prompt type, the picture changes

The aggregate hides the interesting part. When we slice by what we asked for, the defaults shift hard:

"Build a SaaS"

Next.js dominates (78%). v0 always picks it. Lovable, Bolt and Replit Agent default to it. Only buildr split between Next and Hono+Astro depending on whether the prompt mentioned "edge" or "low cost."

"Build a landing page"

Astro is the new default (52%). It used to be Next.js. The shift happened in late 2025 when Astro 5 made islands genuinely zero-JS-by-default. AI builders followed fast.

"Build an API"

Hono leads (38%), Next.js Route Handlers second (28%), Express(!) at 16%. The remaining are Fastify, Bun's built-in server, and a couple of weird Deno picks.

"Build an internal dashboard"

Next.js (62%), then SvelteKit (18%). Dashboards still want React for the component ecosystem.

Why the defaults are diverging

Three forces are pushing the picks apart:

  1. Edge runtime. Cloudflare Workers, Bun, and similar push the default toward smaller frameworks (Hono) for APIs and pure-static (Astro) for landings. Next.js is still where it is, but its share is leaking on the edges.
  2. Cost. Every framework that uses serverless functions adds up at scale. Astro's "ship pure HTML where possible" is unbeatable for static-heavy content.
  3. Cold start. Hono on Workers cold-starts in ~5ms. Next.js App Router on Lambda cold-starts in 200-800ms. For APIs, that gap matters.

What's next: the post-Next.js fragmentation

The honest read of the data: Next.js will still own most full-stack scaffolds in 2027, but its share will probably drop below 50% as Astro takes more of "content" and Hono takes more of "edge API." That's healthy. Different shapes of work want different tools.

The other thing that's quietly happening: AI builders are starting to mix frameworks within a single project. A Next.js frontend + Hono Workers API + Astro landing — all in one repo, all deployed to one account. That used to be too complex for human teams. AI doesn't care.

The "what's the best framework" debate just got replaced by "what's the best framework for this slice." AI builders made that question affordable to ask.

Where buildr's defaults sit

For reference, buildr's defaults by app type:

These match what the data says is winning, by use case. We didn't pick by polling — but the polling and our defaults agree.

Verdict

Next.js still leads, but the AI default is becoming "right tool per slice."

If you're starting a new project, the question isn't "Next or not." It's "what shape is this work, and what's the smallest tool that fits the shape?" Astro for content, Hono for APIs, Next.js for full-stack apps. The AI builders are figuring it out faster than the discourse.

Pick the right tool per slice — without picking.

buildr's defaults are by use case, not by trend. Astro for landings, Hono for APIs, Next.js for full-stack. Same chat, the right framework for what you actually said.

Build my app free

Keep reading