Writing for the answer, not the click

Around 60 percent of Google searches now end without a click. AI Overviews and chat answers have commoditized the generic informational blog post, which means a large share of what content teams have been producing for a decade no longer has a job.

What survives has some combination of four properties. Every asset worth building should pass at least two.

The four traits

1. Proprietary. Original data, original research, firsthand testing, or an insider view a model cannot reconstruct.
2. Experience-based. First-person reporting, hands-on testing, real customer accounts, actual workflows.
3. Niche-focused. Narrow enough that people seek it by name rather than by keyword.
4. Task-completing. The page is where an action happens, not just where information is delivered.

What to build, in order

Build these first. An owned audience, meaning an email or SMS list, because it reduces dependence on Google and drives branded search back to you. And transaction pages, which survive even as agents start doing the shopping.

Build these when you have something proprietary. Original research from your own data. Community spaces where the users make the content. Video or podcast where the host is the brand. In-depth reviews with repeatable testing protocols. And brand pages, meaning about, editorial policy, trust, press, and case studies, which are cheap to produce and foundational to how both Google and language models recognize you as an entity.

Build these carefully. Directories and databases, which only work if the data is proprietary and fresh. Expert perspective anchored in experience someone else can't claim. Templates, useful but easy to reproduce, strongest when tied to a product. Case studies, valuable only when the before and after numbers and the actual process are exposed. Original reporting, if you can be fast or exclusive. Documentation, required if you have a product, and heavily referenced by models.

Avoid these as your primary strategy. Guides and explainers, which AI summaries have replaced unless yours carry proprietary data. FAQs and glossaries, useful only if yours becomes the canonical industry definition. Lists and roundups, which now require original testing and a real scoring rubric to clear the bar.

The decision rule

Before starting any piece, name which type it is. If it lands in the weak tier and you have no defensible moat for it, don't ship it. Find a different angle instead of a better draft.

One correction worth making early: most case studies read like ads and are wasted effort. The ones that work expose the metrics and the process, including what went wrong.

Formatting so a model can lift it

The four traits decide what to write. These three rules decide whether an AI answer can cleanly extract it.

1. Answer in the first 150 words. Lead with the answer rather than the windup. This is the single highest-return retrofit on most existing pages.
2. Every section maps to a real question. Structure around what people actually ask rather than an outline that made sense to you.
3. Every section stands alone. Each block should be extractable and complete without the rest of the page. Citation happens at the passage level, not the page level.

Those three make a reasonable pre-publish check on any page you were going to ship anyway.