---
title: "How to Train ChatGPT to Write Like Me (and Keep DM Replies Accurate)"
slug: how-to-train-ai-to-reply-in-your-voice
author: "Leonardo Maldonado"
category: "AI Setter Operations"
articleType: how-to-guide
tags: ["train AI to write like you","AI writing style"]
publishedAt: 2026-09-11T09:00:00.000Z
updatedAt: 2026-10-26T12:00:00.000Z
canonical: https://setluca.com/blog/how-to-train-ai-to-reply-in-your-voice
---# How to Train ChatGPT to Write Like Me (and Keep DM Replies Accurate)

> If you're asking how to train ChatGPT to write like me, the method is the same in any AI tool. Give it 25 to 30 of your real messages across eight message types, a short style spec, and eight to ten worked examples. Correct every draft, and turn each repeated edit into a spec line. Keep prices, links and availability coming from you, so the replies stay accurate.

## Key takeaways

- Training AI to write like you starts with 25 to 30 real messages chosen for coverage, not a generic prompt or a personality slider.
- The style spec is a written artifact: length, openers, punctuation, emoji, how you say no, how you handle price, and a never-say list.
- Eight to ten worked exchanges beat another paragraph of rules. Google DeepMind's 2024 many-shot research found gains kept coming as examples were added.
- In ChatGPT, a project holds the spec as instructions and your samples as files. Your corrections are where most of the learning happens.
- Accuracy is the other half. Prices, links and availability come from you, and grief, illness or money trouble goes to you.

---

The answer to "how to train ChatGPT to write like me" works in any AI tool. Start with 25 to 30 real messages chosen for coverage, plus a short style spec and eight to ten worked examples. Correct every draft and fold repeated edits into the spec. Matching how you write is half the job. Accurate facts are the rest.

## What does it mean for AI to write like you?

It means the drafts read like something you'd type, so a lead would swear you wrote them. Same rhythm and length, the same openers and punctuation, the same way you say no or name a price. A generic reply reads as a generic offer, and people notice: 88% of consumers say they're more likely to buy when engagement is personalized in real time, while only 44% of brands say they deliver it (Twilio, 2025).

Training gives you three things you can actually hold: a set of chosen messages, a short spec, and a bank of examples. Counting your own patterns is a separate job, and the guide to [measure how you write](/blog/make-ai-dms-sound-like-you) covers it.

## Step 1: Which past DMs should you train on?

Pick 25 to 30 of your own sent replies from the last 90 days, chosen for coverage rather than quality. Nearly half of sales pros already use generative AI to help write sales content (HubSpot, 2024), so having a model isn't the edge. What you feed it is. Coverage means one real example of every situation the AI will meet. Most coaches paste in thirty first replies, then wonder why the price drafts are bad.

| Message type | How many | Why it earns a slot |
| --- | --- | --- |
| First reply to a new lead | 5 to 6 | Your most-used draft |
| A qualifying question | 4 | Probing before pitching |
| Naming your price | 3 | Highest-stakes draft you have |
| Handling an objection | 3 | "too expensive," "let me think" |
| Follow-up to a quiet lead | 3 | A lighter register |
| Booking or rescheduling | 2 to 3 | Short, logistical, easy to fumble |
| Turning someone down | 2 | Almost nobody has these |
| An ordinary busy-day reply | 3 | Your baseline, not your best |

Take your own messages only, unedited, one thread per lead. Then take five things out, however often they show up in your real DMs:

- **Any promise.** One sample saying "you'll be down two stone by June" teaches the model to say it to everyone.
- **Invented urgency.** "Two spots left" when there weren't.
- **Anything private.** Client names, health details, anything said to you in confidence.
- **Prices you no longer charge.** A stale number surfaces at the worst moment.
- **Real misspellings.** Loose grammar reads as human, but "recieve" just looks careless.

## Step 2: What goes into your style spec?

The spec is a short plain-text file the model reads before every draft. Ten things carry almost everything that makes a DM read as yours: message length and how many you send at once, greeting habit, capitalization, punctuation, emoji, whether you end on a question, how you say no, how you name a price, your stock phrases, and a never-say list.

Answer one line for each, in words you'd actually use. From memory is fine, since a wrong answer you fix in week one costs you nothing. Keep it under 40 lines, because rules that argue with each other produce stiffer drafts.

```text
STYLE SPEC: [your name]        Updated: [date]

I sell [one 12-week 1:1 program at GBP 900] to [lifters who've stalled].

LENGTH   [9 to 14 words a message. Two sends back to back is normal.]
OPENERS  [No greeting. Start on the thing they just said.]
         [First name only when I'm being warm.]
FORMAT   [Lowercase starts. Line breaks instead of commas.]
         [Emoji: 💪 or 🙂 only, one in ten, never first line.]
         [End on a question unless I'm confirming a time.]
PRICE    [Number bare, no cushion, then one question about fit.]
NO       [Say it plainly and give the reason.]

NEVER SAY
[reach out, circle back, investment, journey, at your convenience]
Never promise a result, a timeline, or a number on the scale.
Never invent a discount, a deadline, or a spot count.

HAND TO ME
Grief, illness, money trouble, or a call we already had.
```

Keep old copies, so a bad week is one file away from undone, and keep the offer facts in that top line so a price change is a one-line edit.

## Step 3: Why do worked examples beat more rules?

Rules tell the model what to do. Examples show it. Paste in eight to ten whole exchanges (the lead's message and your real reply), labelled by type, pulled from the set you gathered in step one. That's enough for a DM setter, because you only really send a handful of kinds of message. More does keep helping: Google DeepMind's 2024 paper on many-shot in-context learning found the output kept improving as examples went from a handful into the hundreds, though every extra one costs something to run. For DMs the gains flatten long before the cost does.

```text
EXAMPLE 3: naming the price
LEAD: ok so what's the damage 😅
ME: GBP 900 for the 12 weeks
ME: does that land anywhere near what you had in mind?
```

## A worked example: Marcus, and eight before/after pairs

Marcus is a composite, built from patterns coaches describe rather than one person, and his numbers are illustrative. He's in Manchester, sells a 12-week 1:1 program at GBP 900, and gets 35 to 45 DMs a week off Reels about training plateaus. His corpus: 27 replies across the eight types, a 22-line spec, nine examples.

Week one he rewrote 22 of about 30 drafts, and three edits kept repeating: the model opened with the lead's first name, wrote "Great question!" before answering, and explained the program before the number. Those became three spec lines, and by week three he was editing six drafts in thirty. Same message twice below.

**1. First reply off a Reel comment**

> "Hi there! Thanks so much for your interest in my coaching program."

> "right, the plateau thing. how long have you been stuck at the same numbers?"

**2. Qualifying before pitching**

> "Could you share more about your routine and your goals?"

> "how many days a week are you lifting right now?"

**3. Naming the price**

> "The investment for my 12-week transformation programme is GBP 900."

> "GBP 900 for the 12 weeks"

> "does that land anywhere near what you had in mind?"

**4. "It's too expensive"**

> "I completely understand budget concerns! Many clients find the value outweighs the cost."

> "fair. what were you expecting it to be?"

**5. "Let me think about it"**

> "Of course, take all the time you need! Let me know if you have questions."

> "sure. what's the bit you're not sure about?"

**6. Following up after three quiet days**

> "Just following up to see if you're still interested!"

> "hey, you still thinking about it or has life taken over? either's fine"

**7. Turning someone down**

> "Thank you for your interest! Unfortunately this isn't the right fit right now."

> "honestly not the right time for you. six weeks of just showing up would do more"

**8. A question you can't answer**

> "That's a great question! While I'm not a medical professional, I'd suggest your doctor."

> "that one's a doctor question, not mine. worth asking before we start anything"

Pairs 4 and 5 map to [objection handling in DMs](/blog/objection-handling-in-dms) and [how to handle "let me think about it"](/blog/how-to-handle-let-me-think-about-it), pair 3 to [the price conversation in DMs](/blog/the-price-conversation-in-dms).

## How to train ChatGPT to write like me: the setup inside a project

Put the three pieces where ChatGPT reads them every time. A project is the cleanest place. It keeps chats, files and instructions together, and project instructions apply only inside that project, overriding your global custom instructions (OpenAI Help Center).

1. Create a new project and name it after the job, such as "DM replies".
2. Paste your style spec into the project instructions, under project settings.
3. Add your samples and worked examples as a project file, one exchange per block, labelled by message type.
4. Paste a lead's message into a chat in that project and ask for one reply. Correct it in the chat, then copy any fix you make twice into the instructions.

Custom instructions, under Settings and Personalization, work too. They apply to every chat, though, so your DM rules would shape everything else you write. For ready-made wording, see these [ChatGPT prompts for Instagram DMs](/blog/chatgpt-prompts-for-instagram-dms).

ChatGPT still drafts one message at a time while you copy and paste. That's fine for learning how you write, and [whether ChatGPT can respond to Instagram DMs](/blog/can-chatgpt-respond-to-instagram-dms) covers why it stops there. For live DMs, the same spec and examples move into a tool that drafts inside the inbox.

## Step 4: How do you correct drafts with review on while you train?

Turn the model loose on real DMs, but switch on review for the channels you're training. People want a hand on the wheel wherever AI is involved, and 61% of Americans said they want more control over how it's used in their lives, up six points in a year (Pew Research Center, 2025). In SetLuca, auto-send is on by default and you can switch any channel to review. Each reply then waits for you to approve, tweak, or rewrite before it sends. The corrections are the training data, so log them as you go. The third column is the point.

| What you changed | Why | Where it belongs |
| --- | --- | --- |
| Cut "Great question!" | Not something you'd type | Never-say list |
| Removed the name from the opener | You only use names warmly | Opener rule |
| Moved the price above the explanation | You answer first | Price rule |
| Split four lines into two sends | Your real rhythm | Length rule |

Any edit you make twice in a week becomes a spec line. Any edit that survives the new rule becomes an example instead, because a rule the model can't apply needs showing, not stating.

**Chart:** How sales professionals use AI: 43% use AI at work, 47% use generative AI to help write sales content, and 22% use it to qualify leads, per HubSpot's 2024 State of AI in Sales report. (Source: HubSpot State of AI in Sales, 2024.)

## How do you keep DM replies accurate?

Keep every fact out of the model's imagination. Prices, start dates, spots, links and what's included come from you, not from the samples. Matching how you write gets a reply read, and the facts decide whether it costs you the sale. Four rules cover most of it:

- **Facts live in one line.** The offer, price and length sit at the top of the spec. Update it the day anything changes, then fix every example that names the old number.
- **Links get pasted, never written.** Keep your booking and checkout links in the spec, so the model copies them instead of guessing a URL.
- **Availability comes from your calendar.** If the model can't see your calendar, it shouldn't offer a time. It should ask you first.
- **Anything it can't confirm goes to you.** "Let me check and come back to you" beats a confident wrong answer every time.

The other accuracy failure is judgment. A lead writes "I've barely trained since my mum got ill," and a draft that matches your writing perfectly asks how many days a week they're lifting. The style was right, the judgment was missing, and that thread is over.

Fixing it takes three steps. Send grief, illness, and money trouble to yourself before the model drafts anything; that's the last block in the spec above. Take over any thread where a lead has told you something personal. And if a tin-eared message did go out, reply yourself within the hour and own it plainly, without explaining the software. People forgive a person correcting themselves.

SetLuca drafts and sends DM replies the way you write. On Messenger, it also reads your Page's past replies to learn your writing style. Auto-send is on by default, and you can switch any channel to review whenever you want to read drafts first.

## Step 5: How do you run the iteration loop and catch drift?

The loop is draft, compare, correct, retrain, on a schedule. Without the schedule, training quietly turns into editing every message by hand forever, which is the thing you bought the tool to stop doing.

| When | What you do | Roughly |
| --- | --- | --- |
| Every draft, weeks 1 to 2 | Edit in the queue, log what repeats | As you go |
| End of week 1 | Promote repeated edits into spec lines | 15 min |
| End of week 2 | Swap better approved drafts into the bank | 15 min |
| End of week 4 | Run the blind read | 20 min |
| Monthly after | Cold-read 20 approved drafts, mark what's off | 10 min |

Drift then arrives on a trigger rather than a timer. Four of them force a retrain: a price change (fix the spec line and every price example that day), a second channel (run a separate spec; the [multi-channel DM playbook](/blog/multi-channel-dm-playbook) covers why), a one-season corpus, and any tool upgrade.

## How do you know the training worked?

Run a blind read. Shuffle ten AI drafts with ten messages you genuinely sent, strip the labels, and ask someone who knows how you text to pick the AI ones. Ten right out of twenty is chance; fifteen or more means the spec is missing something their reasons will name. Then branch on the result:

- **Wrong in one message type only.** Add four or five samples of that type.
- **The same wrong move everywhere.** A missing spec rule, not a missing sample.
- **The rule exists and it keeps breaking it.** Replace the rule with an example.
- **Still rewriting most drafts, or one-word edits on half.** Review stays on for everything.
- **Two clean weeks plus a blind read at 10 to 12.** Switch review off for that channel, and keep reading price and objection threads yourself.
- **A draft made a claim you wouldn't make.** Stop until the never-say list is fixed.

## Which edge cases should you decide now?

- **You don't have 25 sendable DMs yet.** Write the spec from memory, start with four or five examples, and keep review on until the corpus fills. Don't pad with invented messages. A fabricated sample teaches a fabricated style.
- **Half your samples were written by a setter.** Build from one person only. A mixed set produces a third writing style neither of you recognises.
- **Your offer changed halfway through.** Keep the older samples for rhythm, delete every one that names the old price or length. Your writing style survives a change of offer. Facts don't.
- **You genuinely write long.** Set the length rule to your real median, not a "short DMs" rule you read somewhere.

## Why isn't the training loop improving your drafts?

| Symptom | Likely cause | Fix |
| --- | --- | --- |
| Drafts got worse after new samples | Last batch wasn't how you write, or was someone else's | Roll back a spec version, add samples one type at a time |
| First replies land, objections don't | Coverage gap in the corpus | Add three or four objection samples, including one you lost |
| The spec grows, drafts stiffen | Too many rules, and some of them clash | Cap the spec at 40 lines, move the fine detail into examples |
| Style is right, facts are wrong | Offer details scattered through the rules | Keep facts in the header line, update the day they change |
| Drafts read like your best writing | You only fed polished samples | Add three ordinary busy-day replies |

## Which mistakes stall training AI on your writing style?

1. **Writing the spec before collecting the samples.** Adjectives first gets you "friendly and professional," the prompt every coach writes and the reason every draft reads the same.
2. **Rewriting drafts outside the tool.** Typing your reply straight into Instagram feels faster and teaches the model nothing. The edit has to happen where it gets recorded.
3. **Feeding only the threads that closed.** The follow-up that got ignored and the lead you turned down teach range a win-only corpus can't.
4. **Treating training as setup.** A spec written once in September describes a coach who no longer exists by March.

## One honest limit

Training gets the writing right, but it can't invent what the model was never told. If a lead brings up a joke from your last call, or something they said on their intake form, the AI won't know unless it's there in the thread. It won't make up a memory, and you wouldn't want it to.

## Should you tell leads an AI helped write your replies?

Yes, and don't wait to be asked. Never deny AI involvement, and tell people at the start of the conversation, not only when they ask. In California, SB 1001 has made it unlawful since July 2019 to use a bot to mislead people about its artificial identity in order to sell goods or services or influence a vote. Meta's messaging policy asks for disclosure at the beginning of a conversation where the law requires it, and recommends it everywhere else. New York City proposed a rule on October 20, 2026 that would require disclosure at the beginning of an automated customer service interaction. It's proposed, not in force. Since August 2, 2026, the EU AI Act goes further. For AI that talks to people directly, the EU standard is disclosure at the first interaction.

Article 50 of the Act puts that duty on providers, the companies that build AI systems that interact directly with people. Those systems must be designed so people know they're talking to AI, unless it's obvious. The [European Commission's Article 50 FAQ](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act) says the notice comes from the start of the first interaction, in a clear way. It also reaches providers outside the EU when their AI's output is used in the EU.

The rule wasn't delayed, either. The EU's AI Omnibus pushed back the high-risk rules, not these transparency duties (Goodwin, via Mondaq). The Commission's only grace period, to December 2, 2026, covers machine-readable labels on AI-generated content from systems on the market before August 2.

A coach using an AI tool is what the Act calls a deployer, so this duty sits with the tool's provider. Don't assume your tool adds a notice, though. If you have EU leads and replies send on their own, check what a lead sees first, and add a plain line if nothing's there. This is general information, not legal advice.

Saying it costs you nothing, and hiding it costs you the thread the moment someone works it out. Reply only to people who messaged you first and you stay on the right side of the platforms too, which [is Instagram DM automation safe](/blog/is-instagram-dm-automation-safe) covers in full. The platform rules are in [what Meta allows for DM automation](/blog/what-meta-allows-for-dm-automation).

## What should you do first?

Start with the samples, today. The short answer to "how to train ChatGPT to write like me" is one evening of setup and two weeks of corrections. Pick 25 to 30 real messages, write the spec, and label eight to ten examples. Keep the facts line current and the hard threads yours, and the drafts end up accurate as well as recognisably yours.

When you're ready to move it into your inbox, [how to set up an AI setter](/blog/how-to-set-up-an-ai-setter) walks through the channels. For the role itself, read [what an AI DM setter is](/blog/what-is-an-ai-dm-setter). Every [SetLuca plan](/pricing) has auto-send on by default and lets you switch any channel to review while you train.


## FAQ

### How many samples do I need to train ChatGPT to write like me?

Around 25 to 30 real replies, chosen for coverage rather than quality. Take five or six first replies, four qualifying questions, three price conversations, three objections, three follow-ups to quiet leads, and two where you turned someone down. Push to 40 or 50 if you also want to count patterns into a fingerprint sheet.

### Can ChatGPT learn my writing style permanently?

Close enough. Put your style spec in a project's instructions and your samples in its files, so every chat in that project starts from them. Custom instructions, under Settings and Personalization, apply to all your chats instead. Either way, update the spec whenever you correct the same thing twice, or the drafts drift back to generic.

### Does this work for DMs and email?

Yes, but keep a separate spec for each. Your DMs are probably short, lowercase and question-led, while your emails run longer and more formal. Train each on its own samples, because one mixed set teaches a style that fits neither. The examples and the correction loop work the same way in both.

### Should I tell people AI helped write my replies?

Tell them up front and never deny it. EU AI Act Article 50, applying since August 2, 2026, wants disclosure from the first interaction. Meta asks for it at the start where the law requires it, and a New York City rule proposed on October 20, 2026 would too. California's SB 1001 bars bots that mislead people to sell.

### How long until the drafts match how I write?

Usually a week or two of real messages. Early drafts need heavy editing while the model learns from your corrections. As you approve and tweak, the edits shrink. There's no fixed number. It depends on how much you correct and how consistent your own writing is across samples.

### What should a style spec include?

Ten things: message length, opener habit, capitalization, punctuation, emoji use, whether you end on questions, how you say no, how you name a price, your signature phrases, and a never-say list. Keep it under 40 lines. Google DeepMind's 2024 many-shot research found examples keep improving output, so add eight to ten real exchanges alongside the rules.


## Sources

1. [HubSpot: State of AI in Sales 2024. Retrieved October 26, 2026.](https://blog.hubspot.com/sales/state-of-ai-sales)
2. [Twilio: 2025 State of Customer Engagement. Retrieved October 26, 2026.](https://www.twilio.com/en-us/press/releases/socer-2025)
3. [Agarwal et al. (Google DeepMind): "Many-Shot In-Context Learning" (2024)](https://arxiv.org/abs/2404.11018)
4. [Pew Research Center: "AI in Americans' Lives: Awareness, Experiences and Attitudes" (September 2025). Retrieved October 26, 2026.](https://www.pewresearch.org/science/2025/09/17/ai-in-americans-lives-awareness-experiences-and-attitudes/)
5. [European Commission: "Transparency obligations under Article 50 of the AI Act" FAQ (July 24, 2026; retrieved October 7, 2026)](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act)
6. [Goodwin via Mondaq: "Not Delayed, Not Deferred: EU AI Act Transparency Obligations Are Now In Force" (August 5, 2026; retrieved October 7, 2026)](https://www.mondaq.com/uk/compliance/1826042/not-delayed-not-deferred-eu-ai-act-transparency-obligations-are-now-in-force)
7. [California Business and Professions Code §17940 and §17941 (SB 1001, Bots: disclosure; operative July 1, 2019; purchase or sale of goods or services, or influencing a vote). Retrieved October 22, 2026.](https://law.justia.com/codes/california/code-bpc/division-7/part-3/chapter-6/section-17941/)
8. [Meta for Developers: Messenger Platform and Instagram Messaging policy, "Informing users about your automated experience" (disclose at the beginning of a conversation when required by applicable law; recommended otherwise). Retrieved October 21, 2026.](https://developers.facebook.com/docs/messenger-platform/policy/policy-overview/)
9. [NYC Department of Consumer and Worker Protection: "Mamdani Administration Proposes Ban on AI Customer Service Doom Loops" (October 20, 2026; proposed rule, not in force; proposed 6 RCNY 5-312(a) requires disclosure at the beginning of a covered interaction). Retrieved October 22, 2026.](https://www.nyc.gov/site/dca/news/072-26/mamdani-administration-proposes-ban-ai-customer-service-doom-loops)
10. [OpenAI Help Center: "Projects in ChatGPT" (project instructions apply only within a project and override global custom instructions). Retrieved October 22, 2026.](https://help.openai.com/en/articles/10169521-projects-in-chatgpt)

---

Published by SetLuca.