Don't Make the Website Dumber

A website should leave room for a person to have a point of view.

I design hospitality spaces. I think about the order in which a room reveals itself. Where your eye goes. How color changes as you move. What a material feels like before you know its name. None of that begins with a list of keywords.

I want my website to have the same freedom.

But now there is another audience. Before a person finds the studio, a search engine or AI assistant may have read an abbreviated version of it. It might be asked who designed a place, what kind of work we do, or whether we are worth considering. The introduction happens before anyone reaches the photographs.

The temptation is to make everything easier for the machine. Shorter descriptions. Predictable headings. A simpler vocabulary. The same explanation repeated in ten places. Eventually the site might become very easy to summarize and very hard to remember.

I don't think that is a worthwhile trade.

The person who chooses what we see

The shift is real enough to require attention, even if its consequences are still unfolding.

In research published on July 22, 2025, Pew Research Center's Athena Chapekis and Anna Lieb examined browsing data collected in March 2025. People clicked traditional search results in 8% of Google visits with an AI summary, compared with 15% of visits without one. The difference is an observed association, not proof that the summary caused it. [SRC-1] Still, it tells us something about where the first encounter with an idea may now take place.

The people and systems filtering information can influence which work is encountered at all.

That creates an unusual responsibility for anyone making something individual. If we ignore the new intermediaries, they may have too little accurate information. If we write entirely for them, we risk flattening what made the work worth finding.

Our response has been to separate expression from evidence, without separating the truth.

One website, two kinds of attention

We are building the visible KPHD website as an experience. Photography, movement, pacing, language and project narratives belong there. A project should unfold the way we want it to unfold. Some things are better felt before they are explained.

Behind that presentation sits what I call a shadow layer. The name is slightly misleading. It isn't secret, and it should never be a different story reserved for bots. It is a quiet, public, accessible layer of information: who made the work, what the scope was, where it happened, when it was published, and which original sources support those facts.

Think of a restaurant. The dining room is not improved by turning it into the inventory sheet. But the inventory sheet still matters. It allows the people running the room to know exactly what they have.

Our website needs both disciplines.

For us, Framer carries the designed experience. The site also needs ordinary, readable HTML, sensible page titles, descriptive links, project information and a linked press index. These are not aesthetic directions. They are ways of making the same work accurately legible to another kind of reader.

I don't want two competing websites. I want one distinctive site with a reliable public record.

What the press already knows

We discovered that some of our best evidence was sitting outside the website.

Jessica Ritz wrote about the studio in Metropolis in December 2022. [SRC-4] Edie Cohen covered our Equator Coffees work for Interior Design in April 2022. [SRC-5] These articles are not descriptions we manufactured for a new search strategy. They are independently published records of the work.

The task now is modest but important: connect them properly.

On a relevant project page, one carefully chosen and correctly attributed passage can sit beside the design story. A separate, restrained press index can identify the article, author, publication date and original link. An interview with a client should be identified as a client's account. A contributed press release shouldn't masquerade as independent reporting. Neither should a magazine's words be rewritten as though they were ours.

That distinction is part of authority. An accurate citation is more useful than an impressive but careless claim.

Why GitHub belongs in this conversation

GitHub is usually presented as a place for software. I have begun to see another use for it: a versioned public reference shelf.

Our public GitHub profile is a starting point. The next proposed step is a small public repository holding concise, verified studio and project reference material. Not a second portfolio. Not copied magazine articles. Not private drawings, client files or unpublished work.

A useful repository might hold a plain-language README, structured project summaries where publication is permitted, a list of original press links, attribution notes and correction history. Each record would point back to the canonical project on our own website and outward to its original sources.

Why does this matter? A repository makes the factual record readable, linkable and revisable. It gives corrections a history rather than allowing old descriptions to circulate indefinitely. GitHub's own documentation treats the README as the place to explain what a project is, why it exists and who maintains it. [SRC-3]

A GitHub repository does not become an independent endorsement just because it is public. It cannot lend us the authority of a journalist, an original photograph or a built project. Its value is narrower: clarity, provenance and maintenance.

That is useful enough.

A better kind of optimization

There is another encouraging detail. Google's guidance for AI search features says that ordinary search fundamentals still apply. It does not require a special AI-only file or an entirely separate website. Important content should be accessible as text, structured data should agree with what people can see, and the underlying page should be indexable. None of this guarantees an AI citation or a ranking. [SRC-2]

Google also recommends original, helpful, people-first content. [SRC-6] That is not an argument for stripping the personality out of creative work. It is an argument for leaving the expertise in.

The practical sequence is simple. Make the thing you want people to experience. Then give its factual claims a stable public home. Link to the real sources. Record corrections. Make sure the machine-readable material says nothing the human-facing website cannot support.

And keep the editorial decisions with the person whose work is being represented.

A search system may be able to establish that a café contains a particular material, or that a publication covered a project. It cannot decide what the design should mean to its author.

That decision is still ours.

Keep the part that cannot be compressed

This is not an argument against clarity. It is an argument against confusing clarity with sameness.

I want an AI assistant to get the facts right. I want it to know what the studio designed, to recognize the difference between our words and a journalist's, and to find the right project when someone asks.

I do not need it to become the author of the work.

A website should help someone discover us. When they arrive, it should still feel like us.

The factual layer can do its quiet work. The creative layer can remain strange, specific, atmospheric, even unresolved.

We can make the work easier to find without making it easier to forget.