4 copywriting mistakes that are killing your AI website

AI can write clean copy, but it can't get under your customer's skin. Here are the four website copywriting mistakes I see most, and how to fix each one.

Copywriting Mistakes That Are Killing Your Website video thumbnail by Ali Amin
On this page
  1. Watch the video
  2. Mistake 1: You don't actually understand your customer
  3. Mistake 2: You're not researching your competitors
  4. Mistake 3: You're using vague prompts
  5. Mistake 4: You're not testing your copy
  6. Quick recap

Google doesn't penalize AI-written content. So why do so few AI-built websites rank, and even fewer turn visitors into customers?

In my experience, it's almost never the AI's fault. It's the copy. In this post I'll walk you through the four website copywriting mistakes I see people make again and again, and how to fix each one. You can apply this to any type of website.

Watch the video

Watch: Copywriting Mistakes That Are Killing Your Website

▶ Watch "Copywriting Mistakes That Are Killing Your Website" on YouTube

Prefer reading? Here's the full breakdown.

Mistake 1: You don't actually understand your customer

Before you write a single line of copy, ask yourself:

  • Who exactly is this website for?
  • What do they actually want?
  • What problem are they trying to solve?
  • What frustrates them?
  • What words do they use to describe it?

If you can't answer these, there's a good chance your AI can't either.

Here's a real example. On my own site I offer custom business dashboards. People who land on that page already have one intention: they want to control their whole business from one screen. So the headline says exactly that, in their language: run your business from one dashboard. Below it, I list who it's for (e-commerce brands, clinics, real estate agents) and the starting price. Within five to ten seconds, a visitor knows they're in the right place.

This idea isn't new. Russell Brunson's books make the point that you need to understand your customer's pain better than they can describe it themselves. When people feel understood, they buy. That works in 2026 and it will keep working long after.

What doesn't work is asking AI to guess your customer's pain points and then write copy around those guesses. You are the business owner. You should understand your customer better than any model does. Get under their skin first, then bring AI in to help.

Mistake 2: You're not researching your competitors

A lot of us skip competitor research because we assume our AI will do a better job than whatever is already out there. It won't, because there's a whole layer of information AI simply doesn't give you.

When I searched for hiking shoes in the UK, one retailer's product page showed me things no prompt would have: a filter by activity like hill walking, a gallery of product shots, an "at a glance" section built from what customers say (durable, waterproof, good value), the sole material, the weight, free-delivery thresholds, trust ratings in the header, and a clean mega menu. That company has spent hundreds of hours and a whole team on that page. You can learn from it in ten minutes.

Here's what to study on competitor sites yourself:

  • Structure: How are pages organized? What comes first, what comes next, and where is the call to action?
  • Persuasion: Are they using reviews, proof, comparisons, urgency, benefits, or objection handling?
  • Positioning: What are they promising, who are they speaking to, and what makes them look different?
  • Design: Color scheme, typography, font sizes, spacing, icons, layouts, screenshots, visual hierarchy.
  • Tone: Which words do they repeat? Is the voice expert, simple, premium, urgent, friendly, technical, emotional?
  • Gaps: What questions, objections, features or customer needs are they failing to address?

That last one is where your opportunity is. If something already works for the big players in your niche, borrow the principle. Then fill the gaps they left.

Mistake 3: You're using vague prompts

"Write copy for my e-commerce website" is not a prompt. Your AI doesn't know your audience, your product, the problems in your industry, your tone of voice, or your goal, so it fills all of that in with assumptions.

Give it context instead. Here's how I'd brief AI for a hiking and mountain gear store:

  • Task: Write conversion-focused website copy for an e-commerce brand that sells hiking and mountain gear.
  • Audience: People who hike, trek, camp and spend time in the mountains. (Go deeper if you can: what they do for work, how often they go out.)
  • Products: Hiking shoes, mountain jackets, backpacks, thermal clothing, gloves, outdoor gear.
  • Problems: Wet feet, cold weather, poor grip, uncomfortable clothing, heavy gear, unreliable products.
  • What matters to them: Durability, comfort, weather resistance, safety, grip, lightweight materials, confidence outdoors.
  • Tone: Clear, practical, confident, premium, with an adventurous edge.
  • Goal: Make visitors feel the products are built for real mountain conditions, and guide them to the right gear.

Notice how everything connects. Poor grip leads to the hiking shoes. Cold weather leads to the thermal clothing. Every line of the brief ties back to the customer's pain. And remember: analyze your audience first, then ask AI to help with the copy. Don't hand it the whole job.

Mistake 4: You're not testing your copy

Compare these two hero sections for the same home-cleaning business.

Version A: Welcome to Clearnest. Experience the Clearnest difference. We deliver exceptional solutions tailored to your unique needs. After three sentences, you still don't know what the business does.

Version B: Home cleaning, Bristol. A clean home, your weekend back. One-time and recurring cleaning, supplies included, checklist agreed before we visit. Instantly clear: what it is, where it is, and what pain it takes away. The customer gets their weekend back, which moves them toward their dream outcome of a clean home without doing the work.

This isn't just theory. One founder I came across had a fully AI-written page getting around 600 visitors. It read smoothly and had no grammar mistakes, but converted at roughly 1.6%. He realized he didn't understand his own value well enough to say it plainly, so he asked a handful of people who'd signed up what made them try. He rewrote the page in their words, blunt rather than polished, and conversion climbed to around 3.1%, roughly double, with more people reading past the first section.

The lesson: AI doesn't write bad copy. It just can't get under your customer's skin. That part is your job. So write a version, measure it, rewrite it in your customers' language, and measure again.

Quick recap

  1. Understand your customer. Know their pain better than they can describe it.
  2. Research competitors. Study structure, persuasion, design, tone, and the gaps they leave.
  3. Write context-rich prompts. Audience, product, problems, priorities, tone and goal.
  4. Test your copy. Compare before and after, and keep the version your customers respond to.

This takes more time than pasting a one-line prompt into ChatGPT. But it's the difference between a website that looks finished and one that actually sells.


Want to go further and build high-ticket client portals and AI web apps? Join me inside AI Engineer Academy, where I take you from beginner to pro step by step. And if you haven't yet, watch the full video on YouTube and subscribe to @AgenticAli.

Keep building. And keep writing.

Ali Amin

The community

Learn it with me, not alone.

AI Engineer Academy on Skool is where the courses, live calls and other builders are. Learn at your own pace, with people who are building too.