On the other hand, a person (an actual human) doing the same research would have skimmed four or five.
Multiply that across everyone who has quietly replaced searching with asking, and the math tips pretty quickly. The crawlers and agents have become the majority audience of the open web now.
That means some growing share of the visits to this site are a machine reading on behalf of someone who will never actually see the page. The layout I fussed over, the type scale, the way a paragraph lands after a section break, the images, etc.
All of it performed for a parser that converts the whole thing to plain text, json, or markdown in a second and moves on.
The Visit Without a Visitor
A while back I wrote about the click disappearing, about search engines answering questions on the results page so the visit never happens. This is the stranger sequel.
The visit happens, the pageview registers, the analytics line goes up and to the right, and there's no person behind any of it it seems.
I mean, the agent reads everything. It reads the whole archive in the time a human reader spends on a single article.
Then it extracts the two sentences relevant to whatever its user asked, blends them with extractions from forty other sites, and presents a tidy synthesis. The user gets their answer, and the agent gets its completion.
The site gets a pageview that has a very different meaning now.
Traffic used to be a proxy for readership, and readership used to be a proxy for attention, and attention was the thing you were actually trying to earn. The proxies have come apart.
A chart that once meant people are reading this now means something closer to machines found this useful as an ingredient. Those are different accomplishments, and only one of them is the reason I write.
The Website Becomes an Ingredient
The honest way to describe what's happening to websites is that they're shifting from destinations to sources. Less a place someone goes to, and more like a thing someone's tools consult. Think of it like an API with some prose on top.
This arc started well before agents though.
Search snippets lifted answers out of pages many many years ago, social cards reduced articles to a headline and a thumbnail. Heck even social feed algorithms decided which fragments got attention.
Every step moved value off the page and into someone else's interface. Agents complete the move, and the page becomes raw material for an answer assembled elsewhere, under someone else's brand, with your name attached loosely (or not at all).
I find this strange to sit with as someone who has cared about websites since 2005. After all, the craft of making a page worth visiting assumed a visitor. When the primary consumer becomes a machine, the craft starts to look like set dressing for an audience that stopped showing up, while a very different audience just files through and takes notes.
Optimization Temptation
The industry response is already forming, and it rhymes with everything I wrote about in the slop floor.
There's a new acronym, answer-engine optimization, and a growing pile of advice about structuring your content for agent legibility. Schema markup everywhere, key points in extractable chunks, prose flattened into question-and-answer blocks so the machines can lift it cleanly.
And the commerce side is even further along still, with whole payment protocols shipping so that shopping agents can transact without a human ever loading the actual storefront.
Restructuring your writing for machine extraction is the same bargain as restructuring it for search ranking, and it costs the same currency. The texture, the digressions, the voice, the parts of writing that make a human want to stay get kinda sanded off in favor of the parts a parser can use.
You can absolutely make your site easier for agents to read. Every step in that direction makes it less worth a person's visit, and the people were the point, at least to me.
I keep coming back to the same conclusion I reached about SEO years ago:
The systems will change again, the optimization has a short shelf life, and the voice you sacrifice to feed them has a long one. So the website stays written for people, even as the people become the minority of its readers.
Why I Keep Betting on Email
This is a big part of why email keeps getting more of my attention while the websites feel more and more like public infrastructure.
Email is the channel where a human still definitely opens the thing, at least for now.
Every address on the lists I run, whether at DailyPhotoTips, TheDailyPreset, DailyTips.dev, or the newsletter attached to this site, was given voluntarily by a person who wanted the emails to arrive.
The open is a choice someone makes with their thumb on a Sunday afternoon, and the replies come from real people with actual names, asking real questions about a real thing they're working on, which is a kind of feedback the website's analytics have never once produced.
The reading happens in a private space with no ranking algorithm deciding whether my send deserves distribution, no agent strip-mining it for someone else's interface, no intermediary at all between the send button and their inbox.
The economics of that relationship are different in kind, and I've written before about how long it takes to build. The list runs on permission instead of discovery, on retention instead of ranking, on people who chose you instead of strangers an algorithm might or might not route past your work.
When I publish to the web now, I'm increasingly publishing to a commons that machines harvest. When I send an email, I'm writing to people who asked.
The honest caveat is sitting right there in the phrase "for now at least." Inbox AI already exists and Gmail will happily compress a newsletter into a one-line summary, and the assistants that read email on people's behalf are only getting more capable.
The day is probably coming when some real fraction of my sends get read first, or only, by a machine doing triage.
Even then, the structural difference holds.
An agent summarizing an email its owner chose to receive sits inside a relationship. The person subscribed, the sender is known, the summary happens in service of something the human already wanted.
An agent harvesting a public page involves no relationship at all. If machines are going to intermediate everything eventually, I'd rather be intermediated inside a permission structure than outside one.
The subscription survives the summarizer better than the pageview survives the crawler.
Writing for the Minority
So the plan, such as it is, stays simple.
The website remains, written for people, knowing most of its visits now come from machines and making peace with that. The essays stay essays instead of becoming structured answer-fodder.
And the email lists keep getting the real investment, because that's where the humans verifiably are.
There's something clarifying about the whole shift, honestly.
For twenty years, web publishing let you avoid knowing exactly who your audience was. Traffic was abstract, readers were numbers, and you could tell yourself stories about reach.
The agents have ended the ambiguity. The machines read the website, and the people read the email. Knowing which audience is which makes it a lot easier to decide where the care goes.