Content marketing is splitting in two. Human attention is concentrating inside a handful of walled gardens: TikTok, Instagram, YouTube, Substack. Each one runs a feed built around the individual user, and those feeds are extremely good at deciding what you should watch next. Meanwhile, the open web is turning into infrastructure for AI. It has fewer and fewer human visitors, and its main readers are becoming AI models, which come back to the same pages again and again and cite what they find.
The internet now looks like two cities with different residents, and brand content has a different job in each. The gardens are where you persuade people and earn their affection, because that's where people spend their time. The open web is where you set down the facts about your company, because that's where models learn what to say about you. A lot of content underperforms because teams still treat these as one job, even though the two audiences now read in different places and respond to different formats.
The move into the gardens started before AI
People went into the gardens because the gardens pair the right format with the right distribution. Video is easier to take in than text, and the recommendation feeds are built to keep you watching. That combination was already pulling attention inward. Now that search engines write the answer themselves, people have even less reason to click through to a website, so the shift is happening faster.
The open web still matters. Its main reader has changed.
The open web's new reader
As people browse less, models read more. Model crawlers fetch public pages like docs, FAQs, product details and comparisons over and over, and what they read becomes part of what models know and tell people. The open web is the part of the internet models can reach cheaply and reliably, while access to the gardens is fragmented and depends on contracts. So the open web is becoming a reference library, and models are its most regular readers.
Writing for that reader takes a different style: factual, unambiguous, current and easy to cite.
People and models reward different things
We already know what people can't resist: video, short and long, served up by a strong feed. Brands are responding by moving marginal effort out of classic SEO and into creator-led programs, serialized video and newsletters that build a community. Long-form analysis still has readers, and more of it now lives inside the gardens, often behind a subscription.
People reward narrative, pace, charisma and a sense of community. Models reward precision, structure, coverage and freshness, and they do best with clean, structured pages that answer questions directly. One piece of content rarely serves both well. Trying to make it do so usually produces something that satisfies neither.
What to make for each audience
Inside the walls, make high-production media for people. Start with the formats the feeds amplify, which means YouTube and TikTok first, then clips adapted to the native format of the other platforms. Use newsletters and podcasts as channels you own, where you can go deeper and publish on a steady cadence. Measure what people do: watch time, completion, saves, replies, community growth and the conversions that follow.
In the open, publish detailed reference pages that a model can read without guessing: product specs, FAQs, implementation guides, evaluations, policies and comparisons. Give each page a stable URL and clear headings, use tables where they help and structured data where it fits, and update aggressively. Write pages a model can quote accurately. The goal is for models to reach for your pages by default whenever they need a fact they can trust. Run these pages like product documentation or a knowledge base, edited for accuracy and breadth. Teams used to churning out SEO copy will find it a different discipline.
What changes for the team
The editorial calendar splits in two: campaigns for people, in video and editorial inside the gardens, and reference material for models on the open web.
Production splits too, with a separate voice and format for each. Inside the walls, the work is cinematic and led by creators. In the open, it's concise and literal.
Measurement gets harder. When a model gives someone what they need inside the conversation, that person has less reason to click, so expect fewer clean paths back to your site. For the open web, track leading indicators: how often model crawlers visit your pages, how often models cite or mention you, and the lift in branded search and direct traffic that follows those citations after a lag.
Govern your open-web reference pages the way you'd govern product documentation, with SLAs for keeping them current. When a model reads your pages, it should find today's pricing, your current positioning and every caveat a buyer needs to know.
Meet each audience where it reads
People will keep spending most of their attention where the feed is strong and the video is good. Models will keep spending most of theirs on the public pages they can reach easily and often. The content strategy that fits is to make great video for people inside the gardens and great reference material for models in the open. The sooner you accept the split, the less effort you waste, and the more often both people and models come back to you by default.