What you pay for when the agency uses AI too
An AI website builder may be enough if you can define the content, review the result, publish it and look after it. Hiring an agency makes sense when you need someone to take on that work against clear acceptance criteria. The tool used does not, by itself, demonstrate the quality of the delivery.
An agency may use AI to prepare drafts, explore layouts or assist with parts of implementation. This can reduce repetitive work. The value of the service depends on what happens around it: understanding the business, deciding what is worth building, reviewing proposals and checking that the website does the agreed job.
If a supplier says everything is built with AI, ask what they review and what they deliver. If they say everything is handcrafted, ask the same questions. Neither label replaces a demonstration. There is also no universal reduction from months to days: scope, content, integrations and outstanding decisions still shape the project.
A worked example: a furniture restoration workshop
Imagine a fictional furniture restoration workshop. It wants to explain the pieces it accepts, show its own photographs and receive enquiries describing the work. It does not need online payments or an automatic quotation. The owner has photographs but has not decided what information is needed before assessing each piece.
An initial draft could arrange the services, place the photographs and add a form. That creates something visible to discuss. A decision then emerges that changes the website: does the workshop accept every piece, or exclude certain materials? The owner must supply the answer; whoever leads the project should identify the gap and turn the answer into a clear explanation.
Suppose the draft promises collection within 48 hours without confirmation from the business. Publishing it would introduce an invented commitment. The useful correction is to ask how collections are arranged and describe the actual process. Replacing the sentence with more polished wording would not fix the underlying problem.
The customer journey also needs checking. Someone describes a chest of drawers, sends an enquiry and sees a confirmation. Before accepting the delivery, the workshop should be able to check that the email reaches the intended inbox and can be replied to. If the sending provider fails, the interface should explain the problem and offer a way to continue. That work still needs resolving even when the first design looks finished.
The evidence that turns a draft into a delivery
Ask to see the journey that matters to the business. For this workshop, that means finding the service conditions, viewing an example of the work and submitting an enquiry on a phone. You do not need to understand code to request concrete evidence for each step.
These initial checks are not a complete audit. MDN explains that automated auditing tools alone cannot provide a complete picture of accessibility. It also notes that checking form fields in the browser does not replace server-side checks. The proposal should identify what will be tested, what is outside scope and who maintains it.
- Content: the business confirms services, conditions, photographs and factual claims before publication.
- Use: reading, controls, keyboard access and error messages are checked on the agreed screens.
- Functionality: submission, arrival in the inbox and recovery from a failure are checked.
- Continuity: the agreed access is handed over, with an explanation of content changes, recovery and support.
Can an AI-generated website rank in Google?
A useful question is what the page offers the person who finds it. Google’s guidance on AI content calls for accuracy, quality and relevance, and warns against generating many pages without added value. Using AI does not make a text useful or automatically disqualify it.
In this example, explaining which furniture the workshop restores, which photographs help assess a piece and how an enquiry is handled provides real information. Repeating furniture restoration in every section does not supply that information. A service page or article should answer a specific question and connect to a relevant next step.
Ask the supplier to explain the SEO work included and how it will be assessed after launch. Delivering titles or a keyword list cannot guarantee visitors. Indexing, the queries that surface the page and actual enquiries need to be observed through data. A search position or sales total cannot be treated as achieved simply because the website has been published.
When to build it yourself and when to commission help
Choose according to the work you can take on and the importance of each function. A small business may need a complex integration; a large organisation may need a straightforward campaign page. Business size and AI use alone do not settle the decision.
You can also start with a tool and commission a specific review later. Before choosing it, check its editing, export and hosting terms. Builders provide different capabilities, and moving to another supplier should not be assumed to be immediate.
- Build it yourself: suitable when the tool covers the functions and you have time to prepare, check and maintain the website.
- Commission part of the work: useful when a site already exists and a defined design, content or functionality problem needs solving.
- Commission the project: appropriate when those decisions need coordinating and you want an agreed delivery with a named person responsible.
Questions to ask an agency about AI-assisted web design
You could ask: “Which parts will use AI? What information do you need about my business? Who reviews the result, and which working journeys will you demonstrate before handover? What will I be able to change afterwards, and what support is agreed?” Specific answers help you assess the service without guessing how much work sits behind a screenshot.
At VDNX, we see AI as a tool for moving the work forward, alongside considered design, solid development and attention to security. Scope and checks need to be defined for each project. If you are considering a website, share what it needs to do and what you can manage yourself. That context helps establish the support you actually need.
Sources and documentation
- Google Search Central: guidance on generative AI content ↗
- Google Search Central: creating helpful, reliable, people-first content ↗
- MDN: accessibility tooling and the limits of automated audits ↗
- MDN: client-side form validation and its limits ↗
Content created with AI assistance. Check the sources, date and scope of each guide.