Why We Built an AI-Powered Writing Style Analyser (And How It's Changing Client Communications)

When you're managing communications for multiple executives, voice consistency becomes a scale problem. Each client has developed their communication style over years - the way they structure cause-and-effect relationships, how they introduce contrasting ideas, whether they lead with conclusions or build up to them gradually.

These patterns matter more than most people realise. They're what make a CEO's bylined article feel authentic rather than ghostwritten. They're what ensure an executive's crisis response sounds like them, not like a committee-drafted statement.

The subtle differences that separate a CEO who says "We need to pivot our strategy" from one who says "Given current market conditions and emerging customer preferences, we're implementing strategic adjustments to our approach."

After years of briefing team members on client styles, sharing writing samples, and hoping everyone would pick up on the patterns, we wanted to experiment with creating a tool that could help us get closer to someone else's style in the first draft.

Fortunately, around 30 years ago, my mum wrote a book on how children learn to read and write, and her initial insights got us started down a rabbit hole of deeper research that ultimately led us to create an analysis tool we've named the Writing Style Fingerprint.

Our Approach: Reverse-Engineering Communication Patterns

Rather than relying on subjective style guides or theoretical preferences, we decided to reverse-engineer the writing process itself. If we could identify the specific, measurable elements that make someone's writing distinctive, we could teach AI systems to replicate those patterns consistently across all content.

Working with linguistic research and our own observations from years of client work, we selected 60 distinct dimensions where writers make different choices. These weren't arbitrary categories - they were the specific decision points that create a distinctive style and tone of voice.

The style dimensions that actually matter:

Sentence Structure and Information Flow

For sequential events, some executives integrate steps within sentences:

"After collecting data, analysing results, and consulting experts, we implemented changes."

While others separate them:

"First, we collected data. Then we analysed results. Finally, we consulted experts."

Information ordering reveals another critical pattern

Do they lead with conclusions:

"Our revenue increased 40%. Here's why..."

Or build up to them:

"Market conditions improved, customer retention increased, and as a result, our revenue increased 40%"

Voice and Tone Patterns

Personal versus impersonal perspective creates dramatically different feelings that accumulate across an entire piece of content:

"We found that customers prefer..."

vs.

"Research indicates that customers prefer..."

Conversational interjections reveal personality:

"Honestly, this changes everything"

vs.

"This represents a significant change."

Some executives use these naturally; others avoid them entirely.

Process and Instruction Style

Action sequences show up differently depending on the writer:

"Download the file, extract data, analyse results"

vs.

"First download the file. Then extract the data. Finally, analyse the results."

Conditional statements have distinct patterns:

"If trends continue, we'll exceed targets"

vs.

"We'll exceed targets if current trends continue."

The placement of conditions relative to results creates different emphasis.

Emphasis and Contrast Techniques

Contrast markers vary significantly and affect rhythm and emphasis:

"Sales dropped, BUT we stayed profitable"

vs.

"While sales dropped, we stayed profitable."

Emphasis devices range from subtle to pronounced:

"This really matters"

vs.

"This matters significantly"

vs.

"This, above all else, truly matters."

Vocabulary and Register Choices

Formality levels span a wide spectrum and signal different audiences and contexts:

"The individual perspired heavily"

vs.

"The person sweated a lot." These choices

While technical versus plain language reveals communication philosophy:

"Implement comprehensive restructuring measures"

vs.

"Make big changes to how we work."

- And dozens more dimensions covering everything from punctuation patterns to how they integrate evidence into arguments.

The Technical Complexity of Analysing 60 Dimensions of Language

The biggest technical hurdle was creating a system that could reliably identify patterns across 60 different dimensions without human intervention. Above you have only 6. Some patterns are obvious - like whether someone uses numbered lists or integrates information within paragraphs. Others are subtle - like the specific placement of modifying phrases or the preferred structure for presenting contrasting information.

To solve for this, we developed weighted scoring algorithms that consider pattern frequency across multiple documents, context-specific variations, industry-specific conventions, and confidence thresholds for classification and transformation.

What We Discovered About Writing

Building this system taught us several things about writing and communication that surprised us:

Style is more systematic than we thought

Even seemingly "natural" writing follows consistent patterns that can be measured and replicated. The randomness we perceived was actually systematic variation.

Industry context matters significantly

PR writing has specific conventions that needed to be factored into our analysis. Generic style analysis misses these crucial contextual elements.

Confidence scoring is absolutely critical

Not every style dimension applies to every writer, and we needed reliable ways to distinguish between strong patterns and occasional variations.

Examples are essential for application

The system doesn't just identify patterns - it captures specific examples that demonstrate each stylistic choice, making transformation rules concrete and actionable.

The Broader Implications

This project has created a shift in how we think about AI writing assistance. Instead of generic AI that produces bland, templated content, we're data that combined with AI can authentically replicate individual communication styles and ensure it is close enough to the company's corporate messaging style. We can also protect against plagiarism, or demonstrate authenticity.

The implications extend well beyond PR. In the weeks since I posted about this on LinkedIn, we've been approached by publishing houses that want to fingerprint an author's existing work to know what new content has been AI-generated, and universities that want to create style fingerprints during student intakes so that they too can understand how much of a person's work bears their own style or that of an AI.

AI-Assisted Writing In The Future

The Writing Style Analysis has fundamentally changed how we approach client communications. It's not just a tool - it's a new methodology for ensuring authentic, consistent voice across all touch points.

It creates a significant evolution in how we can serve clients. We're no longer just managing their communications—we're becoming guardians of their authentic voice, ensuring it resonates consistently across every interaction.

This isn't about replacing human creativity or judgement. It's about augmenting our ability to maintain the authentic voice that makes communications genuinely effective.

If you're interested in learning more about how MVPR's Writing Style Analyzer could transform your communications strategy, we'd love to hear from you.