Beyond AI Slop: A Strategic Workflow for Building Websites That Actually Perform

Every week a new YouTube video promises “build a website in an hour with AI prompts.” Name the pattern without naming platforms — templated output, stock imagery bolted on, no strategy behind it. Position this as the moment to explain what’s actually different about a workflow built by people who design effective websites for a living.

AI didn’t remove the need for strategy — it raised the bar for what “good” looks like, because now anyone can produce something. The differentiator is no longer “can you build a website,” it’s “does it convert, does it hold up, can you maintain it.”

I’ve been building websites for my clients and for other agencies for over twenty years and the workflow and expectations is what makes a project a success. Usual every day people are sold the lie in marketing and have been for years even with the DIY platforms that anyone can build a website. AI built websites is just an extension of this.

The Problem: What “AI Slop” Websites Actually Get Wrong

Break down the hype workflow into its real steps, then show the gaps:

• No brand strategy or positioning — the AI is guessing at tone and messaging
• No keyword or SEO foundation — content is generic, not built around what the client’s buyers actually search
• Images “injected” into a template rather than chosen to support a message
• A site that looks fine on day one but can’t be edited or extended without breaking

Land the point: this isn’t an argument against AI — it’s an argument against skipping strategy and calling it AI.

A Human-Led, AI-Accelerated Workflow

Here’s how every agency and web designer have been building websites for the past 20 years.

• Discovery Call/Meeting
• Strategy
• Content Development / Copy Writing
• Professional Photography
• Design (time consuming)
• Collaborate in Figma with client
• Designer Manually changes
• Handoff to Dev to Build / Develop
• Implement (time consuming)
• Rinse and repeat….

This process produces an on-brand, on-strategy website but at the cost of 50+ hours and resources.

Here’s what AI-led websites do via AI Chat which every Youtuber is cashing in on the hype.

• Discovery
• Basic questions, basic unstructured strategic approach
• Organise your own photography or use what you have
• Create a design using AI slop content and page suggestions
• Creates static templates with placeholders
• Re-prompt to refine content and insert real images (can take hours and hours and burns tokens)
• Deploy manually via FTP if Git is too hard on hosting environment
• Deploy automatically via Git on a Cloudflare workers setup

Realise you want to change something or it doesn’t work – either manually edit code or re-prompt and re-upload

Of course I’m going into the detail more on the AI side than on the human-led workflow but even the workflow I described for AI is not easy for an everyday person.

Here’s the new AI Web Design Workflow

• Discovery Call/Meeting
• Establish measurement with current analytics in Content Flow
• Define Configured Strategy includes positioning and SEO, brand vocab etc. Get Leads AU – Content Flow
• Collaborate and share page content development with client or Copy Writer and edit
• Preference is still Professional Photography
• Design via command line pulling content, brand, style and assets from Content Flow into Paper (design app)
• Share and collaborate with clients
• Iteration on design has two options: 1. Manual changes. 2. Prompt changes
• Developer uses MCP command line to import design system, assets into WP and create pages builds / develops lightning fast
• Implement
• Continue to measure and report
• Client has access to monthly reports from system and can prompt directly in their own LLM.

Stage One — Strategy First, With Content Flow AU

Nothing gets designed until the strategy exists.

• Brand positioning and messaging framework — who the site is for, what it promises, how it’s different
• Target keyword and SEO research baked in before a single page is planned
• Page-by-page content strategy — what each page needs to say and to whom
• Why this is human-led: strategists and writers, not a prompt, decide what the business needs to say

Stage Two — Collaborative, AI-Accelerated Design With Paper

AI speeds up design exploration without replacing judgment or the client relationship. This is where 99% of “build a website with AI” goes wrong or is no different to a templated approach. Freelancers making money from this area will quickly find that not all customers will accept their AI Slop all that easily.

• Paper pulls inspiration from multiple sources rather than one AI’s default aesthetic
• The client’s own media and content are used — not stock photography standing in for their brand from their existing website or from Content Flow
• Design happens collaboratively with the client in the room, so expectations are aligned before build, not after
• This is the point to contrast directly with “injecting your images into a website” hype workflows — same input (client photos), completely different process (co-designed vs. templated)

Engineering the Build: WordPress, Best-Practice SEO, and ACSS

The foundation has to be maintainable, not just good-looking on launch day.

I’m hearing a lot of “WordPress is dead” especially from the DIY crowd as has been the case for years but I would actually say it’s the other way around. DIY platforms like SquareSpace are absolutely cooked. They have add-on AI and market that to build and edit on their platform quickly but their appeal will continue to be for the designer with zero skills or the start-up mum and dad business.

• Why WordPress as the environment? — ownership, ecosystem, editability
• Best-practice SEO baked into the build? (structure, semantics, page speed), not bolted on after
• Automatic CSS (ACSS) and design systems/frameworks even Core framework that keep the site consistent and easy to extend

Outcomes:

A site a client — or their next developer — can actually edit in the future, unlike a black-box AI export.

Stage Four — AI Implementation That Doesn’t Look Like AI

AI is used at the implementation layer (via Novamira or other MCPs) to move fast from design to working site — but the output is judged against a design system and brand strategy already set by humans. It all starts with a design built in Paper or Figma.

Even this part of the process is not for the every day joe. We’re continuing to add more and more skillsets on top of AI design, AI development, AI implementation. For developers like myself we’re already familiar with command line

• What Novamira / MCP-based implementation actually accelerates (design-to-build handoff)
• Why the guardrails set in Stages 1–3 are what stop this from producing generic AI output
• The speed win: design-to-implementation in as little as an hour — because the strategy work removes the guesswork

Why This Produces Websites That Actually PerformTie it back to outcomes, using the required brand phrases naturally:

• Effective websites that stay effective, because the underlying system is editable long-term
• Websites that perform because they’re built on strategy and keyword research, not a prompt
• Websites that grow business because messaging and design are aligned to a conversion goal

Compare the processes

The traditional workflow and the new workflow reach the same on-brand, on-strategy outcome — the difference is everything it costs to get there. The hype workflow promises speed and instead adds hours of re-prompting to reach something that still can’t be maintained.

Stage	Traditional Agency Workflow (Last 20 Years)	AI Chat / DIY Hype Workflow	The New AI Web Design Workflow
Discovery	Discovery call / meeting	Discovery via chat prompts	Discovery call / meeting
Measurement	Rarely formalised upfront	Skipped	Baseline set from current analytics in Content Flow
Strategy	Strategy phase, but manual and time-consuming	Basic questions, unstructured, generic approach	Configured strategy — positioning, SEO, brand vocabulary — via Content Flow
Content / copy	Dedicated content development and copywriting	Folded into AI-generated page suggestions	Page content developed collaboratively with client or copywriter, then edited
Photography	Professional photography	Client sources their own or reuses what they have	Professional photography remains the preference
Design	Manual design, then collaborate in Figma with client	AI generates a design from slop content and page suggestions	Design pulls content, brand and style from Content Flow into Paper via command line
Client collaboration	In Figma, but adds another tool and review cycle	Minimal — mostly one person prompting alone	Share and collaborate with the client directly in Paper
Refinement	Designer manually implements every change	Re-prompt to refine content and insert real images — can take hours and burn tokens	Two options: manual changes or prompt-based changes
Handoff / build	Manual handoff from designer to developer	Deploy via FTP, or Git on Cloudflare Workers if the host allows it	Developer uses MCP command line to pull the design system and assets straight into WordPress
Implementation	Time-consuming manual build	Bundled into the deploy step above	Fast, structured implementation on a proven WordPress + design system foundation
Ongoing changes	Back to the top of the process — “rinse and repeat”	Manually edit code, or re-prompt and re-upload from scratch	Continue to measure and report; client can request changes or prompt directly in their own LLM
Outcome	On-brand, on-strategy — at the cost of 50+ hours and resources	Looks fast, but hours of re-prompting for a templated result most people can’t actually execute	On-brand, on-strategy, and built for ongoing editability — without the 50+ hour resource cost

Proven Case Study – System Specific Websites for Health Professionals

Last week I met with the largest Australian health online publication Pulse+IT who had some slots open to advertise and said they’d be happy to as long as we could market our brand different to “Get Leads”. I totally get it after a few years, customers thing its a bit “Salesy”. I personally hate sales and would much prefer to be helpful but what sells differs industry by industry and with Get Leads we were going with a strategic approach that was primarily around conversion.

I was quick to build a strategy and do some competitor analysis, keyword identification and SEO detailed research all in Content Flow ready to push out page structures to Paper. Because I’m also the client I didn’t need to collaborate quite as much but did setup our measurement from day 1. Within 2 days I had a design with 10 pages ready to ship to a WordPress site. By next week I expect to have a website live and be running ads with Pulse+IT.

Compare this to building Get Leads AU which took and is still taking years to complete.

Let’s also highlight some of the challenges here with a brand new business though. I have zero authority but have used a domain with some backlinks 20 years old. I don’t expect it to rank number 1 but based on Content Flow Authority analysis on existing competitors (which there are a lot) I should be able to crack that page 1 over the next 12 months.

Closing thoughts on the industry

This article was built with the help of a strategic framework in Content Flow. We connected via MCP and told AI our brief for the article and then asked it to cross referenced our solid SEO keywords, vocab, pillars and customer pain points to write the article. This really is the next iteration of consistent structured human-led AI content development. Exciting stuff.

Book a free growth strategy consultation experience the difference.