.jpg)
Why AI Content Sounds Like Nobody. How a Claude Project Fixes It.
Most AI-generated content has a tell. It's not wrong, exactly. It's just nobody. Competent, neutral, vaguely helpful, and completely interchangeable with everything else out there. If your brand has spent any real time developing a distinct voice, that's a problem.
The fix isn't a better prompt. I learned that the hard way. The fix is a system that carries your brand's knowledge into every piece of content automatically, so you're not re-explaining who you are every time you open a new chat window.
Here's how I build that system using Claude Projects, and how you can do the same thing.
Why Prompts Alone Don't Work
A prompt is a one-time briefing. You explain the brand, the audience, the tone, and the AI does its best with what you gave it. Then you close the tab, and the next conversation starts from zero again. A different team member writes a different prompt. The output drifts. Six months in, you have content that technically follows the brand guide but doesn't actually sound like the same company.
The problem isn't the AI. It's that brand voice lives in context, and context doesn't survive between conversations unless you build somewhere for it to live.
What Claude Projects Make Possible
Claude Projects let you create a persistent shared context that travels with every conversation inside that project. You load the brand knowledge in once. From that point on, every team member who opens the project starts from the same informed foundation. No re-briefing, no drift, no "this doesn't sound like us" edits.
The goal is to build something that functions less like a style guide and more like an editorial brain.
What Actually Goes Into the Project
This is where most people underinvest. They paste in a few bullet points about tone and wonder why the output still feels generic. Real brand voice isn't a list of adjectives. Here's what actually makes the difference:
Start with what already exists. There's no need to build brand knowledge from scratch. The starting point is what the brand has already decided about itself: the existing brand guide, voice guidelines, strategic priorities for the year. A sitemap with links to the main pages of the live website goes in too, so the Claude Project treats current, on-brand, site copy as the source of truth. Blog content stays aligned with how the brand describes itself today, not two years ago.
The brand's soul, not just its voice. A brand guide tells you how a brand sounds. A founder story tells you why it exists at all. That means understanding the specific gap the founder set out to close, who they built it for, and what they believed was missing before they showed up. That origin doesn't appear in every blog post. But it shapes what the writing reaches for, what problems feel worth solving, and what the brand would never say. Without it, AI-generated content is technically on-brand but emotionally hollow.
Audience definitions with distinct messaging angles. If a brand serves more than one audience, vague guidance produces copy that tries to speak to everyone and lands with no one. Each audience gets its own definition: their situation, their pain point, what they need to hear. Every brief specifies exactly one. Single-audience focus on every piece.
Project instructions that function as a full editorial system. This is the part most people skip entirely. Beyond uploaded documents, the project needs a formal set of written instructions covering everything required to produce publish-ready content. In practice that means: writing style rules with specific patterns to avoid, a named content formula for each post type matched to the right use case, a complete CMS field map so every output comes out ready to publish, audience-specific messaging by campaign period, how to write a brief, and a naming convention for keeping conversations organized as the team scales. The instructions don't describe a vague editorial direction. They define a repeatable system that any team member can open and use on day one.
A repeatable content structure, chosen per post, not improvised. Rather than hoping the format comes out right, the project instructions include a set of named formulas, each suited to a different content goal. The thinking behind this approach was shaped by the work of LaShay Lewis, a B2B content strategist known for building structured, conversion-focused content frameworks. Adapted into five formulas matched to different content types:
- Formula A (PAS). Problem, Agitate, Solve. Best for educational and audience-focused posts where the reader needs to feel understood before they'll engage with a solution.
- Formula B (Benefit / Current Way / Better Way). Built for product and program content, with a benefit-led structure that stays credible without tipping into hype.
- Formula C (Pillar / Skyscraper). Long-form SEO content with a table of contents, structured H2s, and early CTA placement. Built to rank and to hold attention.
- Formula D (Before / After / Bridge). For transformation stories, founder content, and culture posts where the journey is the point.
- Formula E (Listicle with Narrative). Scannable, shareable content that still earns a real read. Structured list items with enough context to be useful, not just clickable.
Every brief specifies one formula. The AI doesn't guess at structure. It follows the one assigned. Want the full breakdown of each formula? Get in touch and I'll share the complete framework.
Reference documents as supporting context. Beyond the written instructions, the brand guide, example posts that hit the right note, and any relevant reference materials go into the project. Claude draws from all of them without being prompted.
How the Workflow Runs Day to Day
Once the project is set up, using it is straightforward. Anyone on the team opens the project, starts a new conversation, and provides a brief. The brief covers four things: content type, audience, the angle to lead with, and which formula to use.
That's it. The brand knowledge doesn't need to be re-explained. The structure doesn't need to be re-taught. The writing rules are already there.
The editorial question shifts too. Before this kind of system, reviewing a draft meant asking: does this sound like us? That's slow and subjective. With the project in place, the question becomes: does this get the angle right? Faster, more specific, and far more useful for a team producing content at scale.
Where the Webflow Claude Connector Changes Everything
Writing the content is one half of the job. Getting it into the CMS is the other, and for most teams, that second half is where time quietly disappears.
Webflow now has an official Claude connector, and it closes that gap in a meaningful way. Once connected, Claude can work directly inside your Webflow site: creating and updating CMS collections, adding and modifying fields, writing metadata, managing bulk content changes across multiple collections, and helping apply internal links where they add context.
In practice, that means content coming out of the Claude Project, already formatted to match your CMS fields, can go straight to publishing without a manual copy-paste relay. The brand voice system handles what gets written. The connector handles where it goes.
Setting it up takes a few minutes entirely within Claude. You connect your Webflow account, grant access to the relevant sites or workspaces, and configure how much autonomy Claude has. It can act automatically or ask for approval before making changes. That last setting matters for production sites. The option to require confirmation before anything is published means the speed benefit doesn't come at the cost of control.
Who This Is For
You don't need an enterprise content team for this to be worth building. If your brand has something distinct to say, a real point of view, a specific audience, a voice that's actually yours, and you're using AI to produce content, you're leaving value on the table every time you start from scratch.
A small business owner who needs consistent thought leadership. A marketing manager who wants drafts that are already thinking in the right direction. A designer or agency managing content for multiple clients. The upfront work is the same: get the brand knowledge right, anchor the project to the live site so nothing drifts, and write the instructions with enough specificity that every draft comes out sounding like the brand. A human review is always the final step. That's what takes it from good to published.
The goal isn't AI that writes for you. It's AI that writes like you.
