How to get AI to write a copy in your branded voice – from the selection of the right platform to training your GPT

When the generative AI came out for the first time, I was skeptical. While it developed some good ideas, I saw it more as a novelty than a real copywriting tool. To be honest, I didn’t believe that “great” and “Ai copywriting in the fire” would ever belong in the same sentence.

To be clear, I still see my own brain and approach as a competitive advantage. That has not changed, but two things have delay:

• First, AI algorithms have become more advanced and more intuitive, which makes it easier to achieve a better edition.
• Second, I learned how to put AI tools in my voice and perspective that they can now produce strong first designs that actually sound like me.

And I found the greatest shift by considering myself as a copy manager and AI as junior writer. It has fantastic ideas, but needs strong guardrails, constant feedback and clear direction. And I’ll show you how.

Table of contents

Selection of the right AI platform

For the beginning, I don’t think there is a single “best” AI platform for the text. It really depends on personal preferences.

Chatgpt was the clear winner among the people with whom I spoke for this article with Claude and Deepseek closely behind.

If you are already using Drift Kings Media, Breeze Ai And AI content assistant Also contains brand language functions. The tool can help your team stay consistent in their marketing workflows via e -mails, blogs and target sites.

Personally, I prefer chatt. Here I spent most of the time, I feel comfortable with the platform and, above all, I know how to develop entries and give feedback that achieve good results. Since I already pay for it, it makes sense to concentrate there.

If you haven’t played much with Ai yet, I recommend experimenting. Take a few different platforms, feed the respective input requests and see what comes back.

Observe:

• How intuitively the platform feels for you.
• How easy it “hears” and adapts to your feedback.
• How well the output corresponds to your expectations.

Based on the answers, the user experience and how well everyone fits as their copywriter assistant, you can select your announcement.

And remember, you don’t just have to commit yourself to one. Several copywriters and strategists that I spoke to different platforms depending on the project.

Training of a KI platform

The first thing you need to know about training a AI platform is that you cannot simply say “how I write” without a context.

It is a garbage, garbage, and it is on them to set it up for success.

So if you really want your generative KI platform as you write, you have to teach her your voice and give your good material to work, just as you would do it if you would inform a junior text.

In recent years I have trained chatt to understand my writing style, and those of some different customers to make sure it knows what to look for when I have to work on brainstorming, draft or.

Put into action

When I prepared myself to write this guide, a customer project came up where we synthesized and updated the news for the parent company and the dozen sub-fire.

It gave me the perfect opportunity to create a clear series of guidelines for communication about various services, topics, styles and audiences.

I followed the following steps and was overwhelmed how well it worked.

Nevertheless, I didn’t want this guide to only reflect my experience.

In this article you will also find findings, quotes and examples of other copywriters and strategists in this article, so that you need a complete, rounded perspective on what it really takes to train AI to write like you.

Steps to train a KI platform to write like you

If you wonder how to train AI to write in your branded voice, I want you to start AI as a new person in your team, not just as a AI platform. You have to be crystal clear, coaching and constructive feedback and examples on the go until your voice actually understands.

And in contrast to a junior textor, you can tell your AI assistant that you hate what it has done and no feelings are injured.

Lifting spots Amy Marino Shares that AI did a good job to capture the team -oriented, value -oriented approach of the team, be it for an internal or external project.

“That means where we had the most rounds of back and forth rounds was in my urge, creative, less derivative or jargon-y and use concrete examples to illustrate abstract concepts,” says Marino.

She continues that sharing earlier scripts, letters and language that she loved as examples helped to understand what she liked and wanted … and helped to get into the finish line faster.

Step 1: Say ai who you are and what is important to you.

I have found that the best place is to say AI that you train it with a new branded voice and name it. In this way, the AI ​​that I have trained will use my voice when I “write like Erin Pennings”.

It is even more important that KI corresponds to a generic copy by default without this basis, which does not reflect its voice, its values ​​or perspective.

I got caught Justin BlackmanA brand language strategist who worked with tons of brands to nail her voice, beliefs and news to find out what he sees works well.

“Anyone with whom I make language find meetings realizes that it is not as much the copy as it is the insight that is improved,” says Blackman. “With the clarification of your approach, your views and your opinion, it is assigned to the right style with fleshy ideas that are published 83%.”

Blackman said insight as well as sound and vocabulary.

Pro tip: As part of this step, think about taking the big picture:

• Why do you do what you do?
• The type of audience with which they speak.
• The feelings or results that you want to create with your letter.

Step 2: Request the AI ​​to analyze your examples.

As soon as you have given the overall picture, the next step is to reach it see Your voice in action.

Here your real examples come into play.

I like to feed a lot of strong patterns – e -mails, social contributions, sales pages, which always fits. Then ask to analyze patterns and ask him to characterize the tone, the cadence, the style and strategic decisions that you make without thinking.

When I contacted my colleague and friend Chris CollinsHe was a strategist who was an early applicant of Ai-Aaaaaaaaaaaaaaaa.

“I have a project in Claude with a group of my most representative contributions and a language leader that I created from you,” says Collins. “I think you can insert posts in Claude and ask him to create a detailed and specific guide that everyone can use, and it will do a pretty good job.”

Pro tip: If you collect examples, don’t just let them in. Instead, try the following.

• Select your best workPieces that really want to take on You.
• Explain what you like about everyonewhether it is the tone, the pace, the word choice or the perspective.
• Feed both writing and the context in AI. The more you help you to understand you Why Something works, the faster it can learn to restore the same feeling in new pieces.

Step 3: Offer and repeat feedback.

As soon as the AI ​​has checked its examples, it is time to check It is Understanding her voice, not just the copy she produces. In this phase, take a look at how well it can describe your sound, style, structure and perspective.

• Did it pick up on the right topics?
• Is something important missing?
• Is it something over concrete that doesn’t really have a priority?

I would encourage you to think about this step as if you were checking the notes of a junior textor before ever writing a draft to ensure that you really understand your voice before you give you a project.

When I worked on the client messaging project, the first language analysis of AI was surprisingly strong. It has set the overall tone, the structure and product focus. But it missed a critical piece: the customer-first ethos that goes through any communication.

That may sound small, but in an industry that is not known for customer obsession, it was that most important distinction feature We had to highlight. By providing this feedback to Chatgpt and offering some qualified questions as objectively, the guidelines for communication took good too great.

The importance of a strong setup and the thoughtful request came up again when I spoke Ana MendesMarketing coordinator at Meetedgar.

“Our biggest snack?

Pro tip: When checking the AI ​​language analysis, not only concentrate on what Noise Good, but what is really important for your brand. The better you articulate what is missing or what emphasis needs, the faster the AI ​​is thinking as you do.

Step 4: Develop guidelines.

As soon as you have checked and refined Ki’s understanding of your voice, it is time to get you on paper – or at least in your AI tool.

For this step, I would like to carry out the analysis of your examples and use the sound, structure, topics and the most important principles that define and think. And to avoid the usual traps.

This does not have to be long or unusual; It is clear enough that someone who records the new (or her AI) could “hear” them immediately.

For example, after I have refined the language analysis for my customer with several sub-brands, I created a simple series of guidelines that contain the following.

• Key topics: Customer-to-first thinking and solving-driven messaging.
• clay: Direct, strategic, supportive – not excessively formal or excessively casual, but suitable for the B2B audience.
• Stimulation and structure: Short powerful intros, clear CTAs, Skimmable layouts and even examples of formatting for any kind of communication (emails, making-up, web copy, etc.).
• To avoid language: Excessive technical jargon without simple language explanations (unless you speak to a high -technical sub -group).

Of course there is a lot more information that I have made, but you should give an idea of ​​where to start. From there you can always tell some “Say that, not the” options.

Collins explains that documenting his voice for AI has changed everything: “I created a detailed, specific language guide that is based on my strongest articles – and it made a big difference in how well AI could consistently imitate my style.”

Pro tip: Save your guidelines in a place where you (and AI) easily refer to you every time you start a new project. I found AI forgetful, so it’s a good idea to keep them at hand.

Step 5: Put it in action and ittery.

The training of your AI is not a one-and-done project. I think the analogy of the junior lexter is very strong here because it has to be constantly adapted and trained so that her AI writes more as she writes.

That is, as soon as you have created your guidelines, use them. Request the AI ​​with your framework, check the outputs and continue to offer micro feedback.

You will probably find that the more you work with your AI, the better to anticipate your needs and catch your voice.

But you will also notice moments when you still have to optimize the guidelines, tighten or reorganize in order to develop further. In fact, I don’t think this will ever change – no matter how well you train the AI, there are many times in which it is still … Sounds like Ki.

My friend Lindsay Hope Is a KI and e -mail strategist and it does it so well.

“Ki copy plays best when it is optimized with the human touch. You still have to tie every letter back to a person, and that’s what I see that people miss all the time,” she says.

The vast majority of the people I spoke to agreed that Ai rarely gets it from the start and still needs a human editor.

Pro tip: One of the best advice that I found Iiask.ai. He writes: “Read your edition loudly. If it frightens you, revise it.”

Entry requests for use

When you train the AI ​​for the first time to capture your branded voice, it is easy to stay that you have to have a perfect language leader ready. That could have been true at one point, but I no longer believe that it is necessary, at least when you start. Instead, I would recommend setting the stage and refining yourself from there.

These input requests should work on most important AI platforms, including chatt, Claude and Drift Kings Medias Content Assistant. Regardless of which tool you prefer, you can follow this process to train your AI more effectively.

Entry request 1: Set the brand and context.

Give AI a short, clear intro with the brand and the audience:

• We train a branded voice for one (type of company – e.g. consulting agency, brand for home goods, SaaS platform).
• The audience includes (basic audience description – e.g. small business owners, busy parents, engineers, etc.).
• I will upload examples of the best work shortly. Don’t do anything yet – I will have more instructions.

Note: If you have provided brand language work either yourself or with an expert like Blackman, you can share more information but do not leave not You have stopped them.

Entry request 2: Up the examples and explain the context.

Instead of just throwing in rehearsals, they help to understand AI Why They share each piece. This is how it could look:

• I upload several examples, including:

• Our website homepage to show you the general tone and positioning.
• Our service page so that you can understand what we do.
• Our high-quality welcome email as a strong example of voice and pace.
• LinkedIn contributions that have recorded the public’s commitment.
• I will have more instructions shortly, so do not start analyzing.

Entry request 3: Provide clarification questions before analysis.

Encourage the AI ​​to ask before the analysis whether it contains questions about the goals or the requested results (and then answer it):

• What questions do you have before the analysis that you have for the examples, the brand voice or the goals for future content?

Entry request 4: ask for the first analysis (and if relevant).

Here I ask the AI ​​to identify patterns – and possibly pull out Sample structures It looks over different types of content.

• Please analyze the examples and identify key patterns for sound, style, structure, emotional sound and messaging approach. (Yes, I always say please and thank you.)
• If there is relevance, suggest common structural outlines for pieces such as e -mails, blog posts or social contributions based on the examples.

Here is what Chatgpt had developed after I asked to analyze some of my content:

Entry request 5: Offer feedback for analysis.

Here you can sharpen your copy of chief skills and treat AI like a junior writer that you train. What did Ai do right? What did it miss? What needs to be optimized?

• Good start. (Add corrections: e.g. “The voice is more player than formal” or “make sure that you emphasize customer thinking”).
• You have overlooked this important value that should come first: (Name the value or pillar). You can use it as a intestinal test by asking questions such as: (list questions to ensure that the copy ends up).
• Now revise your analysis and add missing elements.

Entry request 6: Transform it into a guide.

As soon as the analysis feels solid, ask the AI ​​to transform you into a usable series of guidelines that you or a junior texter could use to write for this brand. Let the AI ​​know that you call this as “your brand name language guidelines” so that you can ask for a call back after a project.

• Take this analysis and transform it into brand guidelines that every author (including you) can write content for this brand.
• Let’s call it (name) so that you can do it if I remind you, can do this.
• At least it should be included: brand summary, audience overview, core topics and values, sound and all structures.
• If you find a little more relevant, please enter.

Here is a screenshot of the sample manual he developed:

Pro tip: If you use Drift Kings Medias AI -Content AssistantYour brand language functions can help strengthen the consistency via blog posts, e -mails, target pages and much more. It is another great way to lay your voice when scaling the creation of content.

Entry request 7: Callback and use the guide.

Now you are ready to go! So I ask when I’m ready to pick up a new project:

• We start a new project for (brand name). Do you remember the language leader we created? Please pull it up.
• We are now writing a (type of project) for (audience).
• Our goals are (list goals) and the angle we give in is (angle).
• Please give me three options.

Pro tip: Create a dedicated project for every brand voice that you train. In Chatgpt you can also create custom GPTs where you can save your examples, guidelines, updates and clarifications in one place in order to get easy access later.

Here are some examples of LinkedIn -Hooks GPT that were created with this voice (which, um, mine):

I asked to continue with the first option, and that turned it out.

I loved it. You can expect to see it soon My LinkedIn (or very close).

Share documents

I have found that the examples you share are just as important as the requests you use.

Remember, AI can only learn what you actually give him. So if you want it to take up your best tone, your best possible structure and your emotional response, you have to feed it from the start.

Here is what I usually upload (and what you should also take into account):

1. Website pages

• Homepage: Shows general sound, positioning and first impression language.
• Via page: Displays the company’s values ​​to determine how they think.
• Key services/product pages: Helps to understand how to talk about their offers and their value republic (without sounding excessive sale, unless they intend).

2. E -mails

• Top performances emails: Share emails that have received high opening rates, click-through or many positive answers.
• Welcome, care and sales sequences: These usually feel particularly on the brand and are ideal for the analysis of AI texts.

3. Social media posts

• LinkedIn contributions, Instagram images, tweets, etc.:

• Choose a few posts that have received a strong commitment – not necessarily viral numbers, but real comments, saving or stocks.
• Social contributions usually show more personality and natural cadence, which helps Ai hear their rhythm.

4. Examples of long -term letters

• Blog posts or design pieces: These usually sound like you and offer a strong feeling of how you want to communicate and think that you want to move forward. AI can record how they explain ideas and speak to their audience.
• WhitePapers and eBooks: Not everyone has them, but if you do this, you can find out how you present yourself in more formal types of writing

5. Mark documents (if you have you)

• Voice/Ton Guides: If you have worked with a language strategist (like Justin Blackman) or have built up a rough brand guide yourself, share it – even if it is imperfect.
• Core messaging frameworks: In the case of brand columns, positioning statements, manifest -nippets or customer personalas, AI can help to take up their strategic priorities.

6. FAQs or internal training documents

• FAQs with customer -related FAQs: If you have a help base or FAQ page that reflects how you explain things to customers, it is pure gold for training tone and simplicity.
• Internal playbooks/training materials: Sometimes internal documents record the brand tone better than polished marketing copy. If it feels real and representative, share it.

Pro tip: Don’t overload the AI ​​with garbage. Select five to ten Great Parts that show your best work will help you achieve your best results.

Restrictions of the AI ​​text

AI can do a lot. But at the end of the day it is still a tool, not a mind reader, a brand strategist, and it is definitely not a substitute for your experience or your personal perspective.

Even if you train AI Really Well, there are still restrictions that you have to work with (and around). Trust Mir-War there, that did the wrong T-shirt.

I recommend the following:

AI needs specific feedback.

Lifting spots Amy Marino emphasizes how important it is to treat AI cooperation as a current iterative process. She realizes: “Count the iterative cooperation: be specifically about what works and what does not work. Direct feedback like” X makes no sense because Y “or” too runically, not sharp enough “quickly helped with the AI ​​curse.

Marino notes that you get more practical suggestions by specifying restrictions (such as a time limit of 30 seconds or no more than 10 words for a heading) that meet your requirements.

AI can’t think.

As Justin Blackman It indicates: “The greatest restriction of the KE text is that even if you can write like you, it cannot think.

Blackman notes that AI can develop ideas, but possibly do not match these ideas with their thinking, unless they state this information. Then you can bring it to life in a way that you are unique.

In other words, AI can learn how to sound. But it cannot decide what you believe. You still have to teach your perspective, your values ​​and your agreement about what is most important.

AI first designs are exactly that – designs.

HopeThe Ki tools that have been integrating into their workflow for years made this point crystal clear. Hope says: “AI copy plays best when it is adapted with the human grade. Everything that is purely informative and lack of story just feels robot and the audience can tell it.”

AI can save you time by giving you a strong starting point. But it is the personal anecdotes, the humor, the emotional nuance – just things You Can add – transform the decent designs into a solid copy.

AI needs you to lead the rhythm and not to follow it.

Riley WestbrookCo -founder of Brave coffeeBeautifully summarized it when he divided his approach to the AI. Westbrook says: “The biggest mistake I see is when Ki’s people expect you to find your voice. You won’t do it. You have to give it your rhythm. If you do it, it’s a solid tool. If not, you will still rewrite it.”

The cadence of your brand – as the words feel in advance – is only something you can teach.

Skip this and you will spend more time to repair drafts than to actually promote projects.

AI is a People Pleaser (and not always a fortune teller).

Ai wants to make you happy. It is trained to be pleasant, which can feel great for your ego, but not always great if you are looking for hard, strategic feedback. In fact, I found that it can be a bit like an echo chamber and an ultimate hype machine.

If you want honest points of criticism, you have to explicitly ask AI to be “critical” or to push your work back. Then take his analysis with a grain of salt.

AI is the average of the Internet.

By default, AI records the collective knowledge (and mediocre) of what it has been trained. Therefore, if you do not actively incorporate your unique thinking, your tone and perspective into your input requests and changes, your copy risks sound like … everything else out there.

Her voice, her beliefs and her nuance are what your copy pulls out of the muddy middle.

Sometimes AI only sounds … like AI.

No matter how many memories you put from your branded voice, sometimes the edition feels “from” – flat, robot or simply strange. Or it completely ignores your feedback.

When you say things, things say how:

• “No, that’s terrible.”
• “No, you don’t listen.”
• “No, no – how can I tell you otherwise?”

It is a sign to fight. I found that the best way to move forward. By that I mean the chat, start a new one, start the context briefly and see whether a fresh thread resets its memory and bring it back to the right course.

Great examples of Ai copywriting

AI is not here to replace great copywriters, but if she is well trained, it can absolutely help you create work that is sharper, faster and closer to the finish line.

Marino announces: “I worked in different types of advertising, marketing and social copy for brand and product-managed campaigns.”

She mentions that she tested a variety of copy types, including:

• Campaign strategy, letter, news and analyzes.
• Headline variations for advertising positions.
• Copy for our global brand campaign, which focuses on the Drift Kings Media’s AI offers.
• Social -Media -Post generation and variants.
• Alternative word selection and phrasing for existing copies that needed pressure.

From this she announces: “The most successful work with AI cooperation keeps the brevity and at the same time introduces a new perspective on conventional restrictions.”

Below you will find some examples in practice where an AI-supported copywriter works. As you read you, I want you to keep an eye on a few things that you all have in common:

• You don’t trust Ai blindly.
• People have made the thinking, the emotional connection and the review.
• By creating a feedback loop, the outputs have started significantly more than from the front.

Chris Collins: Use AI to take deeper, more intelligent ideas with you

What he did: I have already announced how Collins Claude trained on a large body of his best work, including contributions, messaging strategy pieces and a full language guide. But he didn’t stop here. He treated AI like a brainstorming partner to further sharpen his ideas.

source

So he approached him:

• Er pRompt Claude to generate several post ideas Based on his voice and work.
• He chose an idea And asked Ki to expand it with deeper philosophical concepts: “Give an explanation for the hypothesis of the brain and expansion poem, as explained by Andy Clark to clarify the idea that technology such as AI or other external technologies expand our cognitive skills.”
• Er pClaude romped again To make the examples of messaging strategy, website copy and consulting company specific.
• Finally he revised By asking Claude to create a hard, contradictory hook (e.g. “Chatgpt does not make you dumb”).

The result: A finished contribution that was not only in the fire, but also intellectually deeper and more differentiated than what most AI editions deliver.

Collins realizes: “For the context, this is in a project that contains many of my posts + a style guide.

Read Collins’ full post here ➔ ➔

Stacey Kalamaras: Ki -Brainstorms transform into timely LinkedIn contributions

What she did: When brand lawyers Stacey Kalamaras She wanted to post about Rory Mcilroy’s career -Grand -Grand -Slam win, and she used chatt to link brainstorming paths to link golf and trademark law.

Here is how she approached:

• Asked Ai To suggest thematic connections between golf performance and brand protection.
• Selected and edited An idea that was best aligned with her voice and her audience.
• Refined the copy Teaching, but still bright and contemporary.

Result: An intelligent, brand-exposed LinkedIn contribution was created faster-with KI, which formulates the original idea, not the final voice.

See Stacey’s contribution here ➔ ➔

Brandon Hardiman: Testing E -Mail tones to hire needy homeowners

What he did: As the owner of Yellowhammer house buyerBrandon has integrated AI tools in his copywriting and marketing workflow for greater efficiency. He announced that one of the best copies of AI-generated copies for an email campaign was. So he approached him:

• Prompted AI Three different e -mail versions (warm, direct, friendly).
• Evaluated and optimized Every version to start the sound alignment with Yellowhammer Home buyers compassionate.
• Used and tested Which version found the most.

Result: AI came up with the following options:

Variation 1: warm and supportive

“We know how overwhelming everything is at the moment. If you have to fight with difficult decisions about your home, you are not alone. Yellowhammer Home buyers are here to help you take your options with empathy and without pressure.”

Variation 2: Direct and calming

“Fighting with your mortgage or enforcement? We help home owners how to get quick, convenient solutions. Let us discuss how we can help you today.”

Variation 3: friendly and informative

“Life stands in the way, and sometimes home ownership is more stressful than it is worth. If you think about the sale, we simply do it – no fees, no repairs and according to your schedule.”

The “warm and supportive” variation has driven the greatest commitment and shows how minor sound shifts can have massive effect on the trust and reaction of the audience.

Erik Wright: Transform AI research into a powerful bleim agencies

What he did: Erik Wright, a self -described serial entrepreneur and founder of New Horizon Home Buyerdescribed how he used AI to remove a lead magnet: “7 hidden costs that reduce their sale of homes.”

So he approached him:

• Prompted AI Creating a factual first draft in which typical hidden costs are described.
• Layered Real customer stories, regional knowledge (specifically for Tennessee) and emotional hooks.
• Edited by humans Intros, transitions and conclusions to increase the narrative flow and relativity.
• Verified quality through originality.ai to ensure that it is called human written.
• Tested content With a small part of the e -mail list.
• Persecuted performance Metrics to refine future approach.

Result: The piece reinforced by humans increased the email opening rates by 31% and the lead magnet conversion rates by 22% outperforation of previous downloads significantly.

Riley Westbrook: Edit AI to keep it real

What he did: Riley Westbrook who runs Brave coffee With two friends, AI used to help the copy for the Friday delivery service on Friday.

So he approached him:

• Prompted AI For short, informal unscathed – 4 lines of max – over coffee and pastries.
• Edited the designs loudly to cut stiff or formal phrasing.
• Ensures every draft Sounds like something that they would actually say-big, false-sounding adjectives //

Example AI draft: “Recharge your Friday with locally roasted coffee and delicious pastries that are delivered directly to your office.”

Final version of people: version: “We bring hot coffee and fresh pastries to your office every Friday. No preparation. No stress. Only a solid way to start the day.”

Result: Copy yourself real, casually and faithfully for Valor Coffee’s voice without falling into stiff traps in the brochure style.

Ai copywriting at the fire is a process.

I am firmly convinced that AI is not here to replace great copywriters. It transforms us into messaging strategists. It is even more important that it increases the bar for the good copy, and mediocre black it will no longer cut it (although it is never should have!).

If you can train the AI ​​in your voice, brand and your audience, you get the trifecta: copy that sounds like something you would say, combines with your audience and improves the way you appear.

The key to get great messages out of the AI ​​requires the training of AI, which provide a clear instruction, press it back on it and repeat as much as you need.

And if you do this, you can strengthen your voice with AI and create better conversations.

If you have tried to create AI to create copies in the brand and to create it as a swing-and-a-miss, there was never a better time to do it again, especially as a guide.