AI Website Agencies: What Should You Ask Before Signing?

Plenty of businesses sign with an "AI website agency" because it sounds fast and cheap, then discover months later that they can't change a single line of copy without emailing the vendor. What actually separates good agencies from bad ones isn't which AI they use — it's who owns the output, how easily it can be edited, and whether it was built to be found.

How is an AI agency different from a traditional one?

The visible difference is speed: teams using AI pipelines tend to show you a working screen before the spec document is finished, then iterate from there. That alone tells you very little, because almost everyone can produce a fast first draft now.

The real difference shows up later. If a human cleaned up and structured what the AI generated, you get a site you can maintain. If not, you get something that looks fine and is impossible to touch. So the question to ask isn't "do you use AI?" — it's "what can we do with this after handoff?"

What should you verify before accepting a quote?

Scope of delivery matters more than the price line. Each of the items below should appear explicitly in the proposal or contract, and vague answers are themselves a signal.

1. Source code handover — do you receive the full GitHub repository, or does the site only exist inside the vendor's account?
2. Domain and hosting ownership — are they registered under your company name, or held on your behalf?
3. Edit rights — can your team swap text and images directly, or is every change billed per request?
4. What happens after the contract ends — does the site keep running if you stop the maintenance retainer?
5. Search and AI visibility — are structured data (JSON-LD), metadata, and a sitemap included? Increasingly it also matters whether assistants like ChatGPT and Gemini can parse the page.
6. Performance and accessibility — mobile load time, image optimization, alt text.
7. Verifiable work — can they show live URLs, not just mockup images?

What are the most common mistakes?

Most regret comes not from paying too much, but from never checking what wasn't included. The recurring patterns:

• Comparing by page count alone — a ten-page quote and a five-page quote say nothing about whether the contact form or booking integration you actually need is in scope.
• Deciding from design mockups only — if the copy is baked into images, neither search engines nor AI assistants can read it.
• Postponing "we'll add it later" — structured data and multilingual support cost more to retrofit than to build in from the start.
• Tracking the launch date but not the handover docs — without deployment steps, environment variables, and admin credentials written down, the next team starts from zero.

How does SAENRU approach this?

SAENRU runs an open launch stage for AI builders, and also builds directly. SAENRU Studio uses an AI pipeline to deliver websites from ₩2.7M (about 2 weeks) and web app MVPs from ₩8.7M (4–6 weeks), with a first prototype within 48 hours. Everything ships as a full GitHub repository transfer, so you can hand it to another developer or maintain it yourself. All prices exclude VAT.

If AI visibility is the goal, we run that separately as AEO/GEO consulting: a measured diagnostic report on how ChatGPT, Gemini, and Claude actually answer about your brand for ₩190,000 (5 business days), implementation packages from ₩3.6M, and monthly operations from ₩1.47M per month. The diagnostic fee is fully credited if you proceed with implementation. Our own site is a worked example — raw HTML with embedded JSON-LD and llms.txt, viewable in the page source.

If you've already built something with AI, consider publishing it first via project launch to gather reactions before hiring anyone. Listing is free, projects collect votes and feedback, and the weekly vote winner receives a week of free main-banner promotion plus an interview feature. Code ownership always stays with the builder.

In short: a good agency doesn't brag about which model it uses. It tells you clearly what you'll be left holding — the source, the domain, the documentation, and the ability to change things yourself. Asking about those four narrows the shortlist quickly.