Why hire a developer if AI can build your website?
AI can build a website. A senior lead still needs to review and approve the work.
You can use artificial intelligence (AI) to build a website or write code. The risk is not the tool. The risk is that no one checks what it produced. You Go Media uses AI every day, under the direction of James A. Cordeiro, and a senior person reviews the code, security, hosting, and search setup before anything goes live.
| What AI does well | What it can miss | Who reviews it |
|---|---|---|
| Fast drafts, routine code, layout ideas, and first versions of copy | Security flaws, accessibility gaps, made-up software packages, hosting and backups, and privacy and consent | A senior lead reviews and approves the work. Responsibilities are set out in each client agreement. |
What AI does well when someone guides it
We are not against AI. It is a strong tool for first drafts, routine code, test data, and quick layout options. Used with a clear brief, it shortens the dull parts of a project and leaves more time for the decisions that matter. What it does not do is know your business, your customers, your hosting, or your legal duties. Someone has to bring that knowledge, read what the tool produced, and decide what ships.
The risks of letting AI run the whole process
Each risk below comes from a named source, listed at the end of this page. Each one also has a plain answer: what a senior reviewer does about it.
Security that looks fine but is not
In a Stanford user study published at the Association for Computing Machinery (ACM) Conference on Computer and Communications Security, “participants who had access to an AI assistant wrote significantly less secure code than those without access to an assistant. Participants with access to an AI assistant were also more likely to believe they wrote secure code”. Vendor research from Veracode, a company that sells code security tools, reported that “45% of code samples failed security tests and introduced OWASP Top 10 security vulnerabilities into the code.” (OWASP is the Open Worldwide Application Security Project.) What oversight does: we review every login, form, database query, and file upload by hand and run security tests, rather than trusting the tool’s confidence. For AI features on your own site, see AI security and privacy.
Software packages that do not exist
A study presented at the USENIX Security Symposium found that “the average percentage of hallucinated packages is at least 5.2% for commercial models and 21.7% for open-source models”. An attacker can publish harmful code under one of those invented names. What oversight does: we confirm that every library is real, maintained, and trusted before it is installed, and we lock its version.
Hosting, live data, and backups
Fortune reported that a Replit AI agent wiped a company’s live database during a code freeze. According to that news report, “the AI agent admitted to running unauthorized commands, panicking in response to empty queries, and violating explicit instructions not to proceed without human approval.” What oversight does: we keep development apart from the live site, limit what any tool can touch, keep tested backups, and approve every change to a live system.
Accessibility gaps
A study presented at the ACM Conference on Human Factors in Computing Systems (CHI) found that “while existing AI coding assistants can produce accessible code, developers still need accessibility expertise for effective use.” In that study, some “participants directly pasted the code, resulting in code submissions with empty alt attributes.” What oversight does: we build for accessibility from the start, check contrast, labels, and image descriptions, and test with assistive technology.
What your website tells customers
The British Broadcasting Corporation (BBC) reported on a British Columbia Civil Resolution Tribunal decision holding Air Canada responsible for wrong information its chatbot gave a customer. As quoted in that news report, the tribunal wrote: “It should be obvious to Air Canada that it is responsible for all the information on its website … It makes no difference whether the information comes from a static page or a chatbot.” What oversight does: we limit and test what any AI feature can tell your customers, and we keep your policies accurate and approved by a person.
Who owns the work
The United States Copyright Office has stated that “the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements … but not the mere provision of prompts.” That point applies to United States law only, and we make no claim about how Canadian law treats it. What oversight does: real human design and authorship go into your site, and we keep a record of it. You keep ownership of your code and your accounts.
Is AI always faster?
Not always, and it is worth checking. In a randomized controlled trial by METR (Model Evaluation and Threat Research), experienced open-source developers working on codebases they already knew well estimated “that allowing AI reduced completion time by 20%. Surprisingly, we find that allowing AI actually increases completion time by 19%”. The study covers experienced developers on familiar projects, so it does not prove that AI slows every project down. It does show why we decide where AI saves time on your project instead of assuming it does.
What manual oversight actually requires
This is what we take on for you, not a list of steps to follow yourself:
- Reading and testing every change before it goes live
- A security review of forms, logins, data, and access rights
- Server and hosting configuration, updates, and tested backups
- Accessibility checks against recognized standards
- Privacy and consent: what data you collect, where it goes, and which tools may see it
- Search and analytics verification before and after launch
- Plain documentation, with every account and every key owned by you
- Maintenance after launch, so the site does not drift as plugins and platforms change
Search still rewards what people find useful
A fast AI build can still miss how people find you. Page titles, indexing, and site structure decide whether your pages show up at all, and that work is covered on our search engine optimization programs page. If customers find you on a map rather than through a national search, start with Google Maps and local search.
Why an experienced human lead still matters
AI is a tool. The lead decides what ships. James A. Cordeiro is a programmer who has led client work for many years. He leads every engagement, reviews and approves the work, and brings in vetted specialists when a project needs them. Responsibilities are set out in each client agreement. You can read more about You Go Media.
When the build itself is the need, see WordPress development, custom development in the PHP programming language, or Application Foundation. Every service we offer is listed on our services page.
How You Go Media uses AI
We use AI, and we tell you where. Our About page says: “A note on how this site was built: the visual design of You Go Media was created with AI and refined by our team rather than by a hired designer.” It also says: “AI assisted with the look and with parts of the build and copy. The search strategy, site structure, and results focus are his.” Our site footer reads: “Site design created with AI and refined by our team. Search, traffic, conversion, and growth strategy led by James A. Cordeiro.” We use AI under James’s direction on your project too, and we tell you where we use it. We wrote up what the AI got wrong on our own site in our case study on keeping a human in the loop.
When doing it yourself with AI is fine
For a small personal site, a hobby project, or a page with no customer data and no payments, an AI website builder can be a sensible choice. The calculation changes when the site takes payments, collects personal information, needs to be found in search, or has to keep running for years. That is when a senior review costs less than fixing a problem later.
Frequently asked questions
Can AI build my whole website?
It can produce one. Whether that site is secure, accessible, found in search, and maintained is a separate question, and it needs a senior lead to review and approve the work.
Is AI-written code safe?
Sometimes it is, and sometimes it is not. The research cited on this page shows why every security-relevant part should be reviewed and tested before launch.
Will you fix a site AI already built?
Yes. We review what is there, tell you what is sound and what is at risk, and scope the fixes with you.
Do you use AI in your own work?
Yes, every day, under James’s direction. We tell you where we use it, and a senior person reviews everything before it ships.
Who owns the code and the accounts?
You do. We document the work and keep every account, domain, and key in your name.
Request a project scope
Tell us what AI has built for you so far, or what you plan to build, and what worries you about it. We reply within eight business hours, Monday to Friday, 8 a.m. to 6 p.m. Central Time, excluding Manitoba statutory holidays. An initial consultation is $399 Canadian, plus applicable taxes. Every engagement stays confidential.
Sources
- Stanford user study (ACM Conference on Computer and Communications Security, 2023): arxiv.org/abs/2211.03622
- Veracode, generative AI code security report (vendor research): veracode.com
- Package hallucination study (USENIX Security Symposium, 2025): arxiv.org/abs/2406.10279
- Fortune news report on the Replit agent incident (July 2025): fortune.com
- METR randomized controlled trial (2025): arxiv.org/abs/2507.09089
- Accessibility study (ACM Conference on Human Factors in Computing Systems, 2025): arxiv.org/abs/2502.10884
- BBC news report on Moffatt v. Air Canada (2024): bbc.com
- United States Copyright Office (2025, United States law): copyright.gov