How AI changes a website build in 2027, and what it doesn't

I use AI every day on our website projects. It writes a lot of the code now. It drafts tests, checks accessibility, and catches mistakes I’d have missed at 11pm. So I get asked a fair question: if AI does all that, why does a website still cost real money?

Here’s an honest breakdown of what’s changed in a build heading into 2027, and what hasn’t.

What AI speeds up

• Scaffolding. The first working version of a page, a form, a content structure. What used to take a day can take an hour.
• Tests. Writing the checks that prove a form submits, a page loads and a link works. AI is good at the tedious ones, which means more of them actually get written.
• Accessibility checks. Running automated scans and suggesting fixes for contrast, missing labels and heading order. A person still checks the result, BUT the first pass is faster and more thorough.
• Repetitive edits. Updating the same component across thirty pages, renaming things, keeping structured data consistent.
• First drafts of supporting text. Alt text, meta descriptions, FAQ starting points. Starting points, not finished copy.

What still takes a person

• Figuring out what the site is for. Who it’s talking to, what they need to see, and what you want them to do. AI can’t interview your best customers.
• Your voice. AI-written copy tends to sound like everyone else’s. Your site should sound like you. A draft is only a draft until a person has edited it to sound like the owner.
• Real photos. Of your team, your work, your space. Stock and generated images don’t build trust the way the real thing does.
• Judgment on design. AI will happily add motion, gradients and sections. Knowing what to leave out is the job. More on that in Web design for 2027.
• Verification. Every claim, number and price on the site gets checked against the source. AI states wrong things confidently, and a wrong price on your website is your problem, not the AI’s.
• Launch and the year after. DNS, redirects, monitoring, and keeping things working when browsers, plugins or AI models change. The model retirement story in When an AI model gets retired is a good example.

How we keep AI-written code honest

AI writes code fast, and it writes mistakes fast too. Our guardrails:

• Tests come first. The code isn’t done until the tests pass.
• Automated checks run before anything goes live: type checks, tests, accessibility and performance.
• A separate review of every change before it’s merged.
• Nothing ships to production without a person approving it.

That process is what lets us use AI heavily without the site getting worse. I wrote more about it in Testing strategy for a boutique software studio.

What this means for your budget

I’d rather not promise a percentage, because it varies by project. What I see is that the time AI saves gets spent on things that used to get squeezed: more testing, better accessibility, more care on content. The site is better for the same money more often than it’s dramatically cheaper.

If someone quotes you a very low price “because AI does it,” ask what’s included after launch, and who checks the work.

Questions worth asking any web designer

• What do you use AI for on a project?
• Who reviews what it produces?
• Will the copy sound like my business, or like a template?
• How do you test the finished site, on a phone, on a slow connection, with a screen reader?
• What happens when something breaks six months from now?

Good answers to those tell you more than whether they use AI at all. If you’d like to hear ours, our process page walks through it, or send me a note. Please let me know if you have any questions, thanks!