AI in marketing — news for you who still think "human-made" is a mark of quality

I've spent the last two years living on the AI train. Testing every new model, every new tool, every new solution in practice. I've rebuilt my entire workflow so AI does what a human used to do. And I am 100% certain about what I'm about to write.

Over and over I hear the same line: *"Oh, so you use AI? We prefer human-made content. Quality matters to us."*

OK.

But where does this attitude actually come from?

First — bad early AI attempts. At the end of 2022, ChatGPT came out, every other marketer posted "AI is the future" on LinkedIn and started pumping out mass-produced "I would be delighted to invite you to our beauty services today!" type drivel. It was genuinely bad. It gave people the impression that AI-written text is robotic, generic, and lifeless.

But that was three years ago. In that time, models have turned over about 15 times. Today's GPT-5, Claude Opus 4.7 or Gemini 3 Ultra are not the same tool as 2022 ChatGPT. Claiming the opposite is like evaluating today's tech industry based on Windows 98.

Second — people whose paycheck depends on fear of AI. When a marketing manager or agency owner tells you "AI is low quality," they're not saying it because the best outcome for you is at stake. They're saying it because their business model depends on you believing their handmade work is automatically better.

That's business-model protection, not an objective assessment.

Just like an SEO expert tells you SEO is complicated, and a lawn care agency tells you mowing requires a "baseline analysis" — an old-school marketing manager tells you AI isn't enough.

They have to say that. Otherwise they lose their job.

When I say LLMs are trained on hundreds of billions of euros, it's not a metaphor. It's close to reality.

What goes into that investment:

• Training compute. Training a single frontier model (GPT-4 class) costs 100M+ USD in compute alone. Across the whole generational shift (2020–2026), we're talking billions.
• Infrastructure and data centers. OpenAI, Anthropic, Google, and Meta have collectively invested hundreds of billions of USD into data centers specifically for AI training.
• Talent. A top researcher's annual package at an AI company is routinely $1–10M. In some cases tens of millions. Multiply that by thousands of people.
• Training data. Licenses to Reddit, the New York Times, academic databases — billions of dollars.

Combined, the global AI industry is investing sums into training and infrastructure that are measured in trillions of euros over the next decade.

And these models have read:

• Nearly every marketing book written in English — Ogilvy, Hopkins, Kennedy, Halbert, Sugarman, Cialdini, Kahneman, Godin, Sinek, Drucker. No living marketer has read all of them.
• Every official Google SEO guide, every study Ahrefs/Moz/Semrush has ever published, every major A/B test result that made it to the public internet.
• Millions of high-converting sales pages, emails, ad copy, landing pages — along with how they were optimized.
• Academic research on persuasion psychology, consumer behavior, purchase-decision processes, cognitive biases.

No human can hold all of this in their head. Not one. The best living marketer has read maybe 100 books and analyzed maybe 500 campaigns. That's 0.01% of what an AI has access to on every single prompt.

When you say "human-made is better," you're comparing a 100-book human to a model that has digested 10 million texts. And you're seriously claiming the 100-book one is smarter.

Let's get concrete. You need website copy that ranks on Google. Who do you give the job to?

Option A: LLM

You sit down at your computer. You give the model the full context: your product or service, your target audience, positioning, competitors, pricing, the questions your customers keep asking. You give it access to your last 6 months of data, customer feedback, Google Search Console, and the 5 best competitor pages.

You ask: *"Write me a page that hits top 3 on Google for [keyword]. First analyze the 5 best competitor pages, find what they're missing, and do better."*

What happens in the next 3 minutes:

• The model reads all 5 competitor pages. Fully. Every word. Every H2, every FAQ, every CTA.
• Compares them against each other. Identifies which topics everyone covers, which only one covers, which are missing everywhere.
• Brings in your unique positioning, pricing, and your customers' favorite questions.
• Combines this with marketing psychology principles (AIDA, PAS, FAB, buyer stages, language contexts).
• Writes you a 2000-word page that beats all 5 competitors.

Time: 3 minutes. Cost: a few cents.

Option B: a local "marketing specialist"

You hire a local marketing agency or freelance marketing specialist. You pay them $100–200 an hour. They promise to do the job in 15–25 hours (total: $1500–5000).

What happens in the next two weeks:

• Monday morning they get started on your project. They're tired because the kid's birthday was last night and the children wouldn't go to sleep before 11 PM.
• They look at your competitors' pages. Not all five — two or three at best. They don't read them fully, they skim.
• They Google three generic articles on the topic and skim those diagonally too. You're paying hourly for this "research."
• They don't get access to your Search Console data because "it's too technical" or "you don't need it." The real reason: they can't read it.
• The next day they have a mandatory meeting with their direct manager where they have to explain why projects are running over the hours they sold.
• In week two they write you the copy. The copy is based on their personal experience, which in reality is 8–10 local small businesses, none of which are in your sector.
• The text is 900 words. Generic. Broad, not deep. It's not going to move your Google position.

Time: 2 weeks. Cost: $2500. Result: mediocre.

Compare:

| | LLM | Local specialist | |---|---|---| | Competitors analyzed | 5/5 fully | 2/5 skimmed | | Training data access | hundreds of billions of texts | 8–10 personal projects | | Uses marketing psychology | always | sometimes, from memory | | Time taken | 3 minutes | 2 weeks | | Cost | a few cents | $2500 | | Tired, burned out, personal problems | no | yes | | Can read Search Console data | yes, every file | usually not | | Produces 5 variants for A/B testing | yes, in 5 minutes | no, no time |

This isn't a biased comparison. It's an honest comparison with today's reality.

You probably have an example where AI produced something bad. Of course you do. We all do. I've personally gotten hundreds of bad outputs from AI.

But the problem isn't the AI. It's you. And me.

If you went up to a random person — or even a random marketing manager at some conference — and asked "give me 10 ideas for how to market my company," you'd get a bad result too. Worse than AI's.

The problem is context. The problem is:

• What prompt you give. "Write me a page for SEO" = bad result. "Write a page for keyword X, my target audience is Y, my competitors are these five, my unique positioning is Z, my customers' recurring questions are these, my tone is direct and opinionated" = excellent result.
• What info you give the AI. AI is not a mind reader. If you don't give it context, it gives you average output. Give it your brand guide, your last 5 campaigns, customer feedback, a competitor analysis — and you'll get back things your in-house marketer would never reach.
• What tools you use. The ChatGPT web app is the most basic. You get a much better result using chained workflows (Claude Projects, custom GPTs, MCP tools, agent structures). We use a purpose-built toolchain for every project, developed through years of practice.
• What model you use. GPT-3.5 was bad. Claude 3 was okay. Today's top tier (GPT-5, Claude Opus 4.7, Gemini 3 Ultra) is on a completely different level. If you use a 2023 model, you get 2023-era output. Logical.

It's a skill. Like skilled camera use is a skill. Give the best camera in the world to someone who's never held a camera — and you'll get a bad photo. Give the same camera to a photographer with 10 years of practice — you'll get a masterpiece.

The reason for 90% of bad AI results: the user doesn't know how to use AI.

I'll be honest. There is one thing AI really doesn't have. It's taste.

Taste is the ability to know what works and what doesn't — what touches people, what makes them laugh, what makes them think, what gets them to pull out their wallet. Taste is what separates "technically correct" from "brilliant." It's what makes the difference between a good ad and a bad ad when both follow the same rules.

AI can write you a technically perfect, grammatically flawless, structurally sound sales page — that leaves you cold. Because it doesn't land. Because it's not interesting.

Taste comes from years of accumulated attention. Seeing lots of bad examples. Making your own mistakes. Watching real customer reactions.

AI doesn't have this.

But — let's be honest — 90% of marketers don't have it either. Most marketers don't have taste. They've read one Seth Godin book, they follow Neil Patel's blog, they use Canva templates and make reel videos with "trending audio." They don't do anything bad. But they don't do anything memorable either.

Real taste is rare. AI doesn't have it. But neither do most humans.

Let's combine the two points:

• AI is smarter, faster, more thorough than 99% of marketers on the technical side (content, structure, SEO, data analysis, variation generation).
• AI doesn't have taste. Neither do most marketers.
• But a few marketers do have taste.

Result: who wins in marketing today?

Not "the human" and not "the AI." It's the person with taste who uses AI at every step.

That person isn't competing against AI. They're competing alongside AI against another person who doesn't use AI. And they win — every single time. Because:

• They generate 50 ad variations while the competitor writes one.
• They analyze all competitor pages in 30 minutes while the competitor analyzes two in a day.
• They send 500 personalized emails in 20 minutes while the competitor sends 20 by hand.
• Their evenings and weekends are free because AI does the routine work while they sleep.

This isn't a theoretical scenario. It's today's reality. We do exactly this at Dopamin.

The winning equation is simple:

A human with taste × AI capability = the marketing model of the future.

Everyone who doesn't do this becomes marginal.

If you're a business owner, I have good news: you don't need an in-house marketer anymore.

An in-house marketing manager typically costs (salary + benefits + workspace + software):

• Small-medium business: $4,000–6,000/month + taxes = ~$5,500–8,000 per month
• Medium-large business: $6,000–10,000/month + taxes = ~$8,000–13,000 per month

But what do you get for that money?

• One person. One pair of eyes. One background.
• 40 hours a week, when they're not sick or on vacation.
• Their skills are limited to what they've personally learned. If you need something they don't know — hire another person alongside them.
• They'll likely change jobs in 2 years, and the next person has to be trained from scratch.

Dopamin does it for you:

• For a third of what your in-house marketing manager costs.
• 10× more work. Because we have the whole AI toolchain, which means a task that would take your in-house marketer 40 hours takes us 4.
• 20× faster. Because AI workflows run 24/7. Our projects hit results in weeks, not months.
• Project-based. No monthly retainer. No long contract. You pay for a specific outcome.
• With a guarantee. We don't hit the result — you get your money back, no questions.

This isn't just swapping out your marketer for a cheaper one. It's a model shift: from a machine that runs at human speed and produces human output → to a machine that runs at AI speed and produces AI output.

→ [Request a proposal](/en/contact)

If you're a marketer, I'll be honest with you: you have very little time.

The people who started seriously using AI in 2023 are today doing alone what used to take a 5-person team. That's not theory. That's actually happening in the market today. We know who those people are. They earn 2–5× more than the average marketing manager.

People who haven't moved to AI by 2026 will be replaced by 2028. They'll be replaced by AI-fluent colleagues, by AI-first agencies (like Dopamin), or just by AI itself in the hands of a younger manager.

This isn't a punishment. It's just math.

But — you can learn. Today. We teach.

Book a free 60-minute call. We'll go through:

• Which tools to use (models, frameworks, workflows)
• Which prompting techniques actually work (not the YouTube "10 prompt tricks")
• How to build an AI workflow for your daily tasks
• Which part of your job gets replaced first, and how to position yourself before it happens

At the end of the call, you'll have a concrete plan. If you decide to take it on yourself — great. If you decide to hire us — also great. No pressure.

→ [Book a free AI call](/en/contact)

A) We do your marketing for you — AI-first

Tell us your biggest marketing problem: more customers, more sales, more visibility, a specific campaign.

In 1–2 business days we analyze the situation with AI (and human taste). We send you a concrete proposal: what result we'll deliver, by when, for how much.

If it works for you — we start. In weeks, not months. No monthly retainer, no long contracts. If we don't deliver — full refund.

→ [Request a proposal](/en/contact)

B) We teach you how to use AI

If you're a marketer (or a business owner who wants to understand it yourself), book a free AI call.

We'll walk through your current situation, your most painful recurring tasks, and what AI workflow would free you from them.

The choice is yours: do it yourself (we give you the tools and plan) or hire us (we do it for you).

→ [Book a free AI call](/en/contact)

Can AI really write in my brand voice?

Yes, if you give it your brand guide, examples of your past content, and clear instructions. We do this every day. At Dopamin, every client has their own tone profile that AI uses to write content that sounds exactly like your company — not generic-robotic.

What if AI makes a mistake?

AI makes mistakes. Like any tool. The difference is that mistakes are controllable: a human reviews, corrects, iterates. Every AI-generated text goes through a double check: a second AI model reviews the first one's output, then a human makes the final call. Our error rate is lower than any purely human-run agency.

How do you make sure the content isn't generic?

Context, context, context. If you send AI "write a blog post about my company," you get a generic result. When we send AI your positioning, your customers' exact pain points, your competitor analysis, and the results of your past best campaigns — we get a hyper-specific result. Our edge is in how we build the prompt, not in the fact that we "use AI."

Will AI replace my marketing manager entirely?

If your marketing manager does 80% manual work (writing content, posting to social media, pulling data) — yes, most of that work can be done faster and cheaper with Dopamin's AI workflows. If your marketing manager is a strategist with taste and judgment — they're still valuable. But strategists are rare. 90% of marketing managers do manual work.

Does AI also make visuals and videos?

Yes. Today's AI tools (Midjourney, Flux, Ideogram, Google Veo 3, Runway, Sora) produce high-quality visuals and videos in seconds. We use them every day. But here "taste" matters even more — AI generates the image, a human picks the right frame.

How fast will I see results?

Depends on the project. SEO — a week to a month. Ad optimization — same day. Content engine up and running — 1–2 weeks. Brand repositioning — 2–4 weeks. Compared to a typical agency that promises "6–12 months" — we're talking about 10× the speed.

How are you different from other agencies that "use AI"?

Most agencies today are "AI-using" only on the surface — they try the ChatGPT web version, get a bad result, say "AI doesn't work," and go back to manual labor. We've been AI-first from day one: we have the full toolchain (MCP tools, agents, custom GPTs, Claude Projects, data pipelines) that makes AI use 10× more powerful than a casual "ChatGPT user." The gap is the same as between computer use and computer programming.

• "Human-made" is no longer an automatic mark of quality in 2026. Most of the time it means "slow, limited, expensive."
• Today's LLMs are trained on more data than any living human will read in a lifetime. That gives them a technical edge on 99% of marketing tasks.
• Bad AI results come from bad use — prompt, context, tools. Not from the AI itself.
• The one thing AI lacks is taste — and most marketers don't have it either.
• The winning equation: human with taste × AI = the marketer of the future. Everyone else becomes marginal.
• For business owners: you don't need an in-house marketer anymore. Dopamin does it for a third of the cost, 10× more work, 20× faster.
• For marketers: time is short. Learn AI now, or lose your job. We teach, if you want.

Want us to do it? [Request a proposal.](/en/contact) Want to learn yourself? [Book a free AI call.](/en/contact)