AI and Agents

The AI Copy Tell: Why Buyers Can Smell Generic Marketing

As AI-written posts show up in every feed, sameness has become a trust problem for experts who sell high-ticket. Use the model for research, structure and first drafts, and keep the claims, stories and voice unmistakably yours.

Sound like you
On this pageJump to a section

"In the ever-evolving landscape of coaching, success requires dedication, strategy and a growth mindset."

Read that sentence again. It is grammatical, positive and completely empty. It could open a post for a fitness coach, a tax advisor or a dentist, and nobody would notice the swap. That is the AI copy tell, and your buyers have started to feel it even when they cannot name it.

I read a lot of marketing, my own included. The pattern I keep noticing is not that the writing is bad. It is that it is interchangeable. For anyone selling a high-ticket service on trust, interchangeable is the worst thing you can be.

What is the AI copy tell?

It is the feeling that you could swap the author and nothing would change.

The usual signs

  • Openings that announce themselves. Sweeping statements about how business is changing, followed by a promise to explain.
  • Tidy triplets. Everything comes in threes with balanced rhythm, and nothing is allowed to be lopsided or true in a messy way.
  • Hedging. "It's important to note" and "may help" in sentences where an expert would just say what they think.
  • Vague claims. Words like "proven" and "powerful" with no number, name, date or constraint attached.
  • No stakes. Nothing is ever risked. No one loses money, gets it wrong or changes their mind.

Why humans spot it

You do not have to be a detective. You only need to have read the same shape of paragraph in ten different places. Pattern recognition is what readers do all day. When the shape is familiar and the substance is missing, they scroll.

Why does sameness cost an expert so much trust?

Because the price of your offer is mostly a bet on your judgment.

High-ticket buyers are buying a point of view

Someone paying thousands of dollars for coaching, a clinic program or a done-for-you service is not buying words. They are deciding whether you see their situation better than they do. Copy that could have been written about anyone suggests you see nobody in particular.

What buyers actually do

They skim for evidence that a real person with a real opinion is on the other side. A named example, a specific number with context, a mistake you admit, an opinion you hold even though it costs you some readers: those are the signals. Smooth, balanced, agreeable prose is the absence of all of them.

Marketing is still early

A McKinsey survey cited by Harvard Business Review in December 2023 found only 10% to 14% of companies consistently used generative AI in marketing and sales.3 That will change. When it does, the average level of polished-but-empty copy goes up, and the value of a real specific voice goes up with it.

10-14%

Companies consistently using generative AI in marketing and sales, per a McKinsey survey cited by HBR, December 2023Source 3

What does Google say about AI-generated content?

That quality and purpose matter, not the tool.

As I laid out in my AI Overviews piece, Google wrote in February 2023 that its ranking systems aim to reward original, high-quality content that demonstrates E-E-A-T, meaning experience, expertise, authoritativeness and trustworthiness, whatever way it is produced.1 It also said that using automation, including AI, to generate content with the primary purpose of manipulating search rankings is a violation of its spam policies, while acknowledging that not all use of automation, including AI generation, is spam.1

The takeaway is the same one your buyers reach by instinct. If a page was made to help someone, the tool matters little. If it was made to fill space, the tool does not save it.

Where should the model actually help?

In the parts of the work that are about language and structure, not about truth.

Anthropic describes Claude 3.5 Sonnet as exceptional at writing high-quality content with a natural, relatable tone, and as showing marked improvement in grasping nuance and humor.2 I have no reason to doubt that it writes cleanly. The problem is that clean is not the same as yours.

Strong uses

  • Research and summarizing. Pull together what is known about a topic, then verify the important parts yourself.
  • Outlines and structure. Ask for three ways to organize an argument and pick one.
  • First drafts. Get something on the page, then rewrite it in your voice.
  • Objection lists. Ask what a skeptical buyer would say, and answer every point honestly.
  • Editing. Tighten, cut repetition and flag vague claims.

The new reasoning models

OpenAI released its o1-preview reasoning model on September 12. It is designed to spend more time thinking before it responds, which can help with analysis and planning, but OpenAI itself notes that as an early model it lacks features such as browsing and file uploads, and that for many common cases GPT-4o is more capable in the near term.5 It does not change the central point. A model can give you a stronger argument. It cannot give you your own experience.

Anthropic put several of its researchers and prompt engineers on camera for a roundtable posted on September 5, 2024. One of them made a point that applies to every draft you generate:

If you aren't reading the model outputs, you might not even notice that it's making that mistake.

Anthropic researchers, AnthropicSource 6Watch at 9:45

Reading the output closely is the edit pass. The copy tell survives when nobody reads what the model wrote the way a buyer will.

What must stay yours?

Four things, and they are the whole difference between an expert and an echo.

  1. Claims. Every number, result and promise is yours to verify. A model can state a statistic confidently and be wrong.
  2. Stories. Real things that happened to you or to people who gave you permission to tell them. Never invent one.
  3. Proof. Names, dates, sources and context a stranger can check.
  4. Voice. How you actually talk: the words you use, the jokes you make, what makes you angry, what you refuse to say.

What is a 20-minute edit pass that keeps your voice?

  1. Read it aloud. Mark every sentence you would never say to a client's face.
  2. Cut the preamble. Delete the first paragraph. Most AI drafts start one paragraph too early.
  3. Add one specific. Give a real example, number or name to every major claim, or remove the claim.
  4. Add one stake. Say what you got wrong, what it cost or what you changed your mind about.
  5. Break the rhythm. Replace a triplet with two items or four. Make one sentence short and blunt.
  6. Run a swap test. Could another expert publish this unchanged? If yes, it is not finished.

If your copy could be sold to your competitor with the logo swapped, it is not copy yet. It is a draft.

Ekai Stone

AI is a good assistant and a poor substitute for experience. Use it to get to a draft faster, the way AI can draft your SOPs from material you already have, then spend the saved time on the part only you can do. If you want your message shaped so it sounds like you again, that is where I can help.

Keep building.Ekai

Questions people ask

Can buyers tell when marketing copy is written by AI?

Often they notice the effect even if they cannot name the cause: generic phrasing, no specifics, no stakes and no point of view. The problem is sameness, not the tool.

Does Google penalize AI-generated content?

Google says it rewards original, high-quality, people-first content however it is produced. Using automation or AI mainly to manipulate rankings violates its spam policies, but not all use of AI is spam.

What should I use AI for in marketing?

Research, outlining, first drafts, tightening and generating objection lists are strong uses. Claims, stories, proof and final voice should stay with you.

Can I use AI to write testimonials?

No. A testimonial from a person who does not exist or did not have the experience is a fake testimonial, and the FTC's August 2024 final rule bans fake reviews and testimonials, including AI-generated ones.

Sources

  1. 1
  2. 2
    Introducing Claude 3.5 Sonnet

    Anthropic, Jun 21, 2024

  3. 3
    A Practical Guide for Marketers Who Want to Use GenAI

    Harvard Business Review, Dec 11, 202312

  4. 4
  5. 5
    Introducing OpenAI o1-preview

    OpenAI, Sep 12, 2024

  6. 6
    Video: AI prompt engineering: A deep dive

    AnthropicYouTube, Sep 5, 2024

  • ai generated marketing copy
  • ai writing
  • brand voice
  • high-ticket marketing
  • claude 3.5 sonnet
Ekai Stone

Ekai Stone

17 years in business for myself. 12 years running Meta ads, since 2014. Ad director for three 8 figure brands and one 9 figure brand. I build revenue systems that owners keep and run.

ekaistone.comResultsThe Seven Star Build

Read next, from the archive

Related reading

More in AI and Agents

Disclaimer. Results shared on this site are from real clients and reflect years of experience, proven systems, high ad spend, and efficient work. They are not guarantees of income or performance. Every business is different and your results will depend on your offer, effort, decisions, and commitment. Marketing involves risk. If you're looking for shortcuts, this isn’t for you.

Not Facebook™. This site is not a part of the Facebook™ website or Facebook Inc. Additionally, This site is NOT endorsed by Facebook™ in any way. FACEBOOK™ is a trademark of FACEBOOK™, Inc.