In a recent LinkedIn post, Charlie Hills 🦩 discusses how a structured set of Claude files can make AI-generated graphics more consistent with a brand. This article reports on Hills’s post and examines the workflow he describes; it is not written by Hills.
“I’m sick of sloppy AI graphics on my LinkedIn feed.”
Hills presents the setup as a practical response to a familiar frustration: graphics that look inconsistent or poorly finished. He says his approach relies on four files and folders that give Claude a repeatable visual brief.
A brand system built from four parts
According to Hills, the workflow begins with a reference file containing brand colors and fonts. A separate rules file instructs Claude to apply those choices each time, while a design file documents the broader brand system. An examples folder supplies samples of the creator’s strongest existing work.
“A reference file stores your colours and fonts.”
The distinction between these components matters in Hills’s account. Rather than asking the AI to infer a brand from a single prompt, the files spell out visual standards and provide examples. Hills’s post suggests that this preparation is what makes the process easier to repeat, although it does not detail how the files are formatted or how the system handles outputs that miss the mark.
From repeated revisions to a better setup
Hills also points to his own design background as context for the process. He describes himself as “fussy” about the results and says that, when he began in April, producing a graphic could take hundreds of variations. He now characterizes his “AI OS for LinkedIn” as substantially better.
“When I started in April, one graphic could take hundreds of variations.”
That comparison frames the post’s central argument: the quality and efficiency of AI-assisted design depend not only on the model, but also on the instructions and brand materials supplied to it. Hills does not provide metrics for the time saved or examples of before-and-after graphics, so the improvement is presented as his experience rather than a measured benchmark.
Setup as the condition for ease
In Hills’s view, the system makes creating graphics easier once the groundwork is in place. His post is both a description of that process and a promotion for a paid resource: he directs readers to a full AI OS for LinkedIn at charliehills.ai/aios and says enrollment closes Sunday, 8 November. The post does not specify the year for that deadline.
Hills’s takeaway is that consistent AI graphics come from giving the tool a clear visual rulebook and examples to follow. His account may interest professionals who use AI for brand content, particularly those weighing the effort of building a reusable setup against repeated one-off revisions.
📝 About This Content
This article is based on insights shared by Charlie Hills 🦩 on LinkedIn.
📅 Originally posted on November 4, 2026 | View original post on LinkedIn →