How search is changing.
A running account of how AI is rewriting search — original thinking, live experiments, and the lessons that surface in the work itself.
94% of B2B SaaS Homepages Prove They're Credible. Only 13% Say Who They're For.
We ran 100 B2B SaaS homepages through the same six-question test. Nearly all prove they're credible; almost none say who they're for — and directness falls off a cliff as companies scale.
What Actually Happens When AI Reads Your Website
A model doesn't read your website — it takes it apart. Pages become chunks, chunks become vectors, and somewhere in that translation it decides which sentences it trusts enough to repeat. We followed one page through the whole process, and what came back changed how we write.
The Trust You Don't Control
Models weigh what the rest of the web says about you at least as much as what you say about yourself. On mentions as borrowed credibility.
The Paragraph Is the New Page
AI doesn't retrieve pages; it retrieves passages. Why the paragraph is now the unit that decides whether you're found.
Two Kinds of Being Found
Traditional search rewards pages; AI rewards passages it trusts. Where the two still overlap — and where optimizing for one quietly costs you the other.
How to Refresh Old Content for LLMs
Outdated pages don't sit quietly. AI systems pull them, restate them as current fact, and hand them to your customers with confidence. Here's how to refresh old content so LLMs cite the right version of you.
How a Machine Sees Meaning
Before AI can decide what to cite, it has to grasp what your words mean. A plain-language tour of embeddings — the step that turns writing into something a model can weigh.
How AI Decides Who to Believe
Every AI answer is a chain of small trust decisions. We break down the signals behind a citation — and the ones you can actually influence.
The Unglamorous Half of Visibility
Crawlability, structure, speed — the invisible groundwork that decides whether any of the smart work ever gets seen.
No articles under this topic yet.
Thinking on AI search, in your inbox.
Occasional and genuinely useful — sent when we have something worth saying, never to fill a calendar.
