Products SEO & GEO Agent Paid Ads Agent Website Agent Conversion Agent
How it works Pricing FAQ Blog Book a call

AI copywriting: combine human and machine for better copy

Updated · 5 min read

Short answer Let AI write the words; you supply what it can't know. In controlled tests, AI copy persuades about as well as human copy. What it lacks is your buyer's real worry, your offer, your proof and the promises you can keep. Give it those, check every claim, and never let it invent reviews.

Want help applying this to your business? Talk it through on a free 30-minute call.

Book a call

What can AI do in copywriting, and what can't it?

AI is good with words and bad with facts it was never given. Copywriting is writing that asks someone to act: an ad, a service page headline, a follow-up email. AI can draft all of it in seconds. It cannot know your prices, your guarantee, or why your last ten customers called you.

Copy jobAI aloneWhat a person must add
15 headline options for an adStrongPick the one that matches what callers actually say
Rewriting a long page shorter or plainerStrongCheck nothing true was lost
Versions for different buyers (landlord, homeowner)StrongTell it who those buyers are
The offer: price, terms, guaranteeCan't know itYou decide it; AI only words it
Proof: reviews, photos, numbersWill invent it if askedReal proof only
The customer's real worryGuesses a generic oneYou hear it on calls
Final check before it runsCan't be responsibleA person signs off

Google's guidance on AI content explains why the last row matters: generative models "predict a likely sequence of words based on their training data." Likely is not the same as true. Google's advice for anything AI wrote:

"It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." (Google Search Central: Guidance on using generative AI content)

Does AI copy persuade as well as a human's?

In tests so far, about as well. Three studies show the pattern:

  • No overall gap. A 2025 meta-analysis in Scientific Reports pooled 7 studies with 17,422 people. It found "no significant overall difference in persuasive performance between LLMs and humans." Results varied a lot by setting, so the authors say success depends on how the AI is used.
  • AI-written ads rated higher, until labeled. An MIT study (Zhang and Gosline, 2023) had 1,203 people rate short ads for products like an air fryer and an e-bike. Ads written by GPT-4 alone were rated higher than ads by professional content creators at a consulting firm when people didn't know who wrote them. Telling people the source narrowed the gap but did not reverse it. Ads labeled as human-written got a boost; the AI label did not lower scores.
  • Tailored beats generic. Matz and colleagues (2024, four studies, 1,788 people) found messages ChatGPT tailored to a reader's personality or values "exhibit significantly more influence than non-personalized messages."

So the claim that AI can't write emotional, convincing copy doesn't hold up in these tests. They measured short messages rated by survey panels, not booked jobs. Your own numbers decide the rest.

Where does human emotion still win?

Not in the wording. In knowing which feeling to write to, and in proving it. AI can write "worried about a flooded basement?" well. It can't know that your callers are mostly landlords afraid of a tenant's complaint, not homeowners afraid of mold. That changes the headline, the offer and the proof.

What only you bring:

  1. The buyer's real fear and words. Read your last 20 calls, emails or reviews. Copy their phrases into the prompt.
  2. The offer. Same-day service, a fixed price, a warranty. AI can phrase it; only you can stand behind it.
  3. Proof. Real reviews, job photos, years in business, license numbers.
  4. The promise you can keep. AI writes "guaranteed results" happily. You pay for it if it isn't true.
  5. Your voice. Paste two emails you wrote to customers so the draft sounds like you.

How do you combine AI and human copywriting?

Split the work in five steps:

  1. Brief (you, 10 minutes): service, town, buyer, their top worry in their words, offer, proof, words you never use.
  2. Draft (AI): ask for 10 versions, each built on a different angle: speed, price, trust, fear of damage.
  3. Choose (you): pick the two that sound like your best calls.
  4. Check (you): every price, time, claim and name. Delete anything you can't prove.
  5. Test (both): run both versions in an ad or email. Keep the one that books more, then ask AI for new rivals to it.

What can AI copy get you in trouble for?

The same things human copy can. The FTC's small business advertising guide lists three rules: ads "must be truthful and non-deceptive," advertisers "must have evidence to back up their claims," and ads "cannot be unfair." It adds:

"The law requires that advertisers have proof before the ad runs." (FTC: Advertising FAQ's, a guide for small business)

Fake reviews are a separate, sharper risk. The FTC's 2024 rule banning fake reviews and testimonials names reviews from people who don't exist, "such as AI-generated fake reviews."

In Canada, the Competition Bureau says:

"Anyone who writes or permits writing reviews that give a false or misleading impression to consumers could be liable under the Competition Act." (Competition Bureau Canada, January 18, 2024)

Example

A plumber's ad headlines

The owner asks AI for 15 headlines. Most say "fast, reliable plumbing." He adds the line his callers use most, "water coming through the ceiling," and the offer, "a plumber there in 2 hours or the call-out fee is waived." The new headlines name the problem and the offer. AI wrote the words; he supplied the facts.

A med spa's email

AI drafts a promo email that says results are "guaranteed" and quotes a "client" who doesn't exist. The owner deletes both, adds a real review she has permission to use, and states the price, $300, as it appears on her booking page.

A bookkeeper's service page

She has AI write two versions: one for a contractor who hates paperwork, one for a shop owner worried about an audit. She runs both on her website for a month and keeps the one that brings more calls.

FAQ

Will AI replace copywriters?

It can write the first draft. The person who knows the buyer, the offer and the rules still decides what runs and answers for it.

Should I tell customers AI wrote my ad?

For ordinary ad copy it's usually your call. In the MIT study, an AI label didn't lower ratings; a human label raised them.

Which AI tool is best for copywriting?

The tool matters less than the brief. Any major chatbot drafts well when you give it your buyer's words, your offer and your proof.

Can AI write my Google reviews or testimonials?

No. A review must come from a real customer. The FTC's rule bans AI-generated fake reviews, and Canada's Competition Bureau treats false reviews as misleading.

Different for you?

Keep learning

NextStorytelling in marketing: why "once upon a time" worksWhy stories lower a buyer's guard, what the research shows, a simple story structure for a small business, where to use it, and the rules for customer stories.

Ask us one question

One line is enough. We reply by email.

Sources

  1. Hölbling, Maier, Feuerriegel: A meta-analysis of the persuasive power of large language models (Scientific Reports, 2025; arXiv preprint)
  2. Zhang, Gosline: Human favoritism, not AI aversion (Judgment and Decision Making, 2023)
  3. Matz et al.: The potential of generative AI for personalized persuasion at scale (Scientific Reports, 2024)
  4. Google Search Central: Guidance on using generative AI content
  5. FTC: Advertising FAQ's, a guide for small business
  6. FTC: Final rule banning fake reviews and testimonials (August 14, 2024)
  7. Competition Bureau Canada: Online reviews posted by employees (January 18, 2024)