Someone can spend ten minutes reading a portfolio and still ask an AI, “What does this person actually do?”
That is a reasonable question. A personal site has pages, navigation, case studies, and lots of context. A chat window starts blank.
I added a small button to my homepage: Copy context and ask AI.
It copies one plain-text file. The visitor can paste it into ChatGPT, Claude, Gemini, or another AI app, then ask a question about my work.
The feature is tiny. The reason for it is not.
A recommendation needs good source material
People increasingly ask AI tools questions that used to go to a search engine or a friend:
- Who could help us review an AI vendor?
- Who has worked on backend systems and product delivery?
- Is this consultant a fit for a difficult migration?
- What should I ask before I book a call?
An AI answer is only as good as the information it can use. If the user gives it a vague name and a homepage URL, the answer may miss the useful details or fill gaps with guesses.
The copied context gives it a better starting point. It says what I do, where I have worked, what GattyWorks offers, the kinds of problems I am usually a fit for, and the work I am less likely to take on.
It also gives the AI a simple role: introduce my work, acknowledge the visitor’s question, add relevant context, and ask one focused follow-up question.
That last part matters. I do not want a copied prompt that turns every chat into a sales pitch.
The file is plain text on purpose
The button fetches /ai-context.txt and copies its contents to the clipboard.
const response = await fetch("/ai-context.txt");
await navigator.clipboard.writeText(await response.text());The file is public, readable, and easy to change. It does not depend on a chat widget, an API key, or a particular model.
Plain text also makes the source easy to inspect. A visitor can see exactly what the model receives before they paste it.
That is better than hiding a prompt inside a polished interface and hoping it behaves well.
What is inside the context
The file includes four useful groups of facts:
- A short introduction: backend and infrastructure work, product building, practical AI, and GattyWorks.
- The services: audits, architecture, focused sites, MVPs, AI workflows, and technical direction.
- Fit guidance: examples of teams and problems where a conversation may be useful.
- Limits: no invented projects, results, pricing, schedules, or promises.
The limits are as important as the services list. A good answer can say, “I do not know,” instead of turning a missing detail into a confident claim.
Reach is not the same as attention
I am not treating this as an SEO trick or a way to force an AI recommendation.
It cannot guarantee a mention, a lead, or a better ranking. It can make a relevant conversation easier to start with accurate material.
That is the kind of reach I want: fewer confused introductions and more useful questions.
The same principle shaped the site’s agent-readiness work. The site now serves clear HTML, offers Markdown when an agent requests it, and explains where to find the important pages. The copy button is the interactive version of that idea. It hands a visitor a compact, inspectable starting point.
What I would copy on another site
If I added this to another portfolio, I would start with a text file that answers these questions:
Who are you?
What work do you do?
Who is a good fit?
What should the AI avoid guessing?
What is one useful next question?Keep it short enough to read. Remove claims that cannot be checked. Update it when the work changes.
The useful outcome is simple: someone who is already curious can ask a better question without rebuilding your context from scratch.