The AI Writing Tools That Actually Sounds Like You

We tested eight AI writing tools across six distinct coaching voices. Warm and nurturing. Direct and challenging. Analytical and structured. Conversational and informal. Corporate and precise. Spiritual and intuitive.

The results were not what we expected.

The gap between tools was far smaller than anticipated. What created the biggest difference in output quality had almost nothing to do with which tool was used. It had everything to do with how the tool was prompted before a single word of coaching content was written.

Here is what the testing found, what it means for your practice, and the exact prompting technique that changes everything.

Going into the test the assumption was that Claude would outperform ChatGPT, ChatGPT would outperform Gemini, and the gap between them would be significant enough to make a clear recommendation. That is broadly what happens when you test these tools on generic business writing tasks.

For coaching-specific content the picture was completely different. When given no context about the coach's voice, all eight tools produced output that sounded broadly similar. Competent. Professional. Completely generic. The kind of writing that could have come from any coach in any niche anywhere in the world.

When given a detailed voice brief before writing anything, the gap between tools nearly disappeared. Claude produced the most nuanced output on average. But ChatGPT 4o and Gemini Advanced were close enough that the difference would not be noticeable to a coaching client reading a session recap or a welcome email.

The tool matters less than you think. The brief matters more than almost anything else.

Each tool was tested on four coaching writing tasks: a session recap, a welcome email, a LinkedIn post, and a client-facing 90-day roadmap summary. Each task was run twice: once with no voice context and once with the full voice brief below.

Claude (Sonnet)  Best overall for nuance

Consistently produced the most contextually aware output when given a detailed voice brief. Handles subtle emotional register better than the others. Best choice for client-facing documents where tone matters most.

ChatGPT 4o  Best for speed and iteration

Slightly less nuanced than Claude on emotional register but faster to iterate. Excellent for first drafts that you plan to edit. The voice brief made a significant difference here.

Gemini Advanced  Best for structured documents

Strongest performer on structured outputs like scope of work documents and session frameworks. Less reliable on warm or conversational tone. Good second choice if you already use Google Workspace.

Claude (Haiku)  Good for short-form content

Faster and cheaper than Sonnet. Performs well on short tasks like email subject lines, social captions, and intake questionnaire questions. Not recommended for longer client-facing documents.

Jasper  Not recommended for coaching

Built for marketing copy, not coaching content. Every output regardless of the voice brief had a promotional undertone that felt out of place in a coaching context. Deleted after week one.

Copy.ai  Not recommended

Similar issue to Jasper. Output defaulted to marketing language. The voice brief helped slightly but the underlying model consistently pushed toward persuasion rather than conversation.

Writesonic  Inconsistent

Some strong outputs on social content. Unreliable on longer documents. Not worth the learning curve when Claude and ChatGPT are more consistent across all task types.

Notion AI  Useful if you already use Notion

Not a standalone recommendation but if your practice is already built inside Notion it is capable enough for session notes and internal documents. Not suitable for client-facing writing without significant editing.

Use Claude as your primary tool for client-facing documents where tone matters. Use ChatGPT 4o for iteration and drafting when speed matters more than nuance. Everything else is optional. Do not use a tool you have to fight to make sound like you.

Every tool performed significantly better when given this brief before any writing task was started. Run this once at the beginning of a new conversation and keep it as a saved template you paste at the start of every session:

That last instruction, do not start writing anything yet, is important. It stops the AI from immediately producing output based on an incomplete understanding of your voice. The confirmation step forces it to process the brief before acting on it.

The words you never use list is the most powerful part of this brief. AI models default to high-frequency coaching vocabulary. Telling the tool explicitly which words you reject removes the most generic layer of the output before you have written a single word.

If you are not sure which words to include in your never use list, run this prompt first:

Read the list. Pick the ones that make you wince. Those are your never use words. Add them to your voice brief. Every piece of content you produce from that point forward will be immediately distinguishable from the generic output most coaches are publishing.

The coaches who produced the most consistently authentic AI output shared one habit that had nothing to do with prompting technique: they edited every output before sending it.

Not heavily. Not rewriting from scratch. One pass, five minutes, reading it out loud once. The coaches who sent AI output directly without editing were identifiable within two sentences. The coaches who did one editing pass were not.

AI writing tools are a first draft engine, not a publishing engine. That distinction is the difference between content that sounds like you and content that sounds like everyone else who is using the same tools.

Before using any new AI tool for client-facing content, run this calibration test. It takes three minutes and tells you immediately whether the tool can handle your voice or whether it will fight you every time.

Read all three versions. If any of them could have been written by a motivational poster, the tool needs a stronger voice brief before you trust it with client content. If one of the three versions sounds genuinely like something you would say, that is the tone register this tool handles best. Use it for that type of content only until you have calibrated it further.

Three things moving in the coaching + AI world this week.

Coaching clients are increasingly able to identify unedited AI output and the association with inauthenticity is damaging. Research into client trust in professional services shows that perceived authenticity is now weighted more heavily than credentials in initial engagement decisions. Coaches who publish unedited AI content are not saving time. They are spending trust they have not yet earned back.

The pattern emerging among high-performing coaching content creators is consistent: AI produces the structure, the coach provides the voice through editing. Not rewriting. Editing. One pass that adds the specific detail, replaces the generic phrase, and removes the word that no one in their right mind would actually say to a client. That split is producing content that is both efficient to create and authentic enough to build an audience.

Coaching clients are beginning to compare notes. In high-achieving professional networks where multiple people use coaches, the quality of client-facing documents is discussed. Coaches who send personalised, well-structured documents are being referred. Coaches who send documents that read like they came from a template are not. The AI tools covered in this issue close that gap entirely when used with the voice brief technique above.

Five issues in.

The AI writing tools issue was pushed back from Issue #5's original schedule because the testing needed more time to be worth publishing. Eight tools is not a comprehensive market review. It is a practical working test based on the tasks coaches actually do every week. The verdicts above reflect that specific lens, not a general technology assessment.

One thing the testing process confirmed: the coaches who get the most out of these tools are not the ones who use them most. They are the ones who invested thirty minutes at the start building a voice brief they use consistently across every tool and every task. That thirty minutes compounds every time they open a new conversation.

If you run the voice brief prompt from this issue and it produces something that sounds genuinely like you, hit reply and tell us which tool you used and which part of the brief made the biggest difference. That reply directly shapes what gets covered in future issues.

See you next Tuesday.
The Coaching Operator