See how ChatGPT and Claude describe your product. Find the gaps, then fix them.
AI tools can read your code but not the decisions behind it. They fill gaps with competitor patterns, generic defaults, and wrong positioning. gist.design is a structured markdown file that captures design decisions, interaction rationale, product positioning, and explicit boundaries. AI coding tools read it to build features that match design intent. LLMs read it to give accurate recommendations instead of guessing from training data.
| Without gist.design | What prevents it |
|---|---|
| AI coding tool builds auto-execute when you designed for user approval | Design Decisions: "chose X over Y because Z" |
| AI fills gaps with competitor patterns your product deliberately avoids | Not This: explicit boundaries on what this is not |
| LLM recommends your product to the wrong audience | Positioning: who it's for and who it's not for |
| LLM describes your product as a clone of a competitor | Positioning: vs comparisons with honest differences |
Two ways to create the same file. Both run the same guided conversation and produce identical output.
Install the Claude Code skill. The guided conversation runs inside your terminal and the file drops directly into your project root.
Best when you are already in a coding environment
Run an audit above to see what LLMs get wrong, then fix the gaps and download your file.
No setup required
Three steps. One file that fixes how AI understands your product.
62%
45%
58%
Gaps found
See how ChatGPT, Claude, and Perplexity describe your product from your website alone. Find where they guess wrong, miss features, or confuse you with a competitor.
@Docs → Add new doc → paste your gist.design URL
Install the gist-design skill to generate and use files natively. Or if you already have a file, Claude Code reads it automatically.
Place file in repo root
Paste URL or upload file for accurate product recommendations
Figma MCP gives coding assistants visual structure — layouts, components, spacing. gist.design gives them the why — patterns, rationale, boundaries. Structure + intent = accurate implementation.
What makes a gist.design file actually useful to AI tools.
“We chose X over Y because Z” is useful. “The button is blue” is not.
3-tier badge next to each suggestion
“uses confidence visualization”
“Confidence scores appear as a 3-tier badge next to each suggestion,” not “uses confidence visualization.”
What the product is NOT is as important as what it is. The Not This section prevents AI tools from filling gaps with competitor patterns.
A product might have multiple gist.design files, one for each significant feature. Keeps each file focused and contextually useful.
If designers have to author this file manually, it won't happen. The conversation tool handles the format; the designer handles the thinking.
| Topic | Details |
|---|---|
| robots.txt | Crawlers — What can you access? |
| sitemap.xml | Search engines — What pages exist? |
| llms.txt | LLMs — What content matters? |
| gist.design | AI coding tools + LLMs — How does it work, why, and when should you recommend it? |
| File placement | /gist.design in the project root, or /features/[name]/gist.design for multi-feature products |
| Spec | Open for community input. Pattern research at aiuxdesign.guide |
Fix the gaps AI tools get wrong about your product. Generate a structured file that coding tools and LLMs actually read.