Six Pillar Pages, Not Sixty Blog Posts

So the click is largely gone. What’s left is the citation, and most B2B sites are indexed without ever being cited.

Here’s a number from that same study that should impact your content plan: 88% of the AI summaries cited three or more sources.

Three. Selected from every indexed page on the topic. Your job is no longer to be findable. Your job is to be one of the three.

Sixty posts do not make you one of the three. Sixty posts give you sixty candidates, each holding only a fraction of the actual answer.

Where the sixty came from

Nobody sat down and decided sixty was the right number of pages. Sixty is agency math, I can reveal: Five posts a month on a retainer, twelve months, and you get sixty. The number is a deliverable count, but not necessarily driven by a strategy. It exists because an agency needed something countable to invoice against and a VP needed something countable to put on a board slide.

I’ll call it calendar cosplay. The costume of a content strategy: publish schedule, editorial calendar, color-coded sheet, quarterly theme. All the visible signals of the thing, none of the mechanics. Without a coherent strategy that addresses things like topic clustering, thought leadership or a post’s role in the funnel, hat comes out the other end is craptent, competently produced and hardly carrying any weight in your GTM or demand gen program.

What’s the tell? The fact that nobody can name which page is supposed to win which query. Ask a marketing team which URL should be cited when a buyer asks an AI engine what’s the difference between contract lifecycle management and e-signature and you get a pause, then a Google Sheet, then a link to a post from March called “5 Contract Trends to Watch.”

What the machine is doing with your library

Google has documented how both AI Overviews and AI Mode use a query fan-out technique: the system issues multiple related searches across subtopics and data sources, then assembles a response from what comes back.

Read that from a process mindset: One buyer question becomes eight retrieval passes across adjacent subtopics. A single page that covers the category definition, the failure mode, the comparison set, and the evaluation criteria at real depth can surface on several of those passes. Sixty posts that each cover one thin slice surface on just one pass apiece, and lose that pass to whoever went deeper.

The Princeton GEO study out of KDD 2024 tested roughly 10,000 queries and found that classical keyword-density signals had minimal influence on whether a source got cited. What moved the needle was epistemic weight: adding statistics, citing sources, quoting credible authorities. Visibility gains ran up to about 40%.

That finding represent the entirety of what generative engine optimization actually amounts to, and it’s all about resource allocation. Statistics, citations, and named sources cost editorial time. You can afford to put that time into six pages. Most marketers can’t afford to put it into sixty, which is why the sixty posts end up as me-too opinions with a stock photo. Ugh.

Google’s own May 2025 guidance says roughly the same thing in the mildest possible language: focus on unique, valuable content for people and make sure structured data matches what’s actually visible on the page.

Your sixty posts are competing with you

The thin posts you push out don’t just fail to get cited. They actively suppress the page that should.

We call it distractor debt. A retrieval system pulls semantically adjacent content, and “5 Contract Trends to Watch” is semantically adjacent to almost every contract query a buyer could ask. It’s close enough to get retrieved and empty enough to answer nothing. It occupies the slot your good page needed.

Multiply that by fifty-nine siblings and your own archive becomes the noise floor your best work has to climb out of.

Google added scaled content abuse to its spam policies in March 2024, and made the definition deliberately method-agnostic. It doesn’t matter whether a human or a model wrote the pages. What matters is when you’ve pubbed many pages, unoriginal, with little value. Most B2B blogs aren’t spam by intent. But a fair number are spam by outcome.

This is, I might note, a tough thing to confront for a guy who made a lot of his living doing content execution of exactly those kinds of packed calendars. But it’s the truth. There’s still a place for blogposts, sure. But you’ve got to make sure they’re aligned with pillar messaging, scrupulously.

The six

Six pillar pages, each written to be the source an engine quotes when a specific class of buyer question comes up.

What this category actually is. A plain definition of the thing you sell, written so a machine can lift two sentences and be correct. Not your product. The category. If the web can’t agree on what you are, nothing downstream works.

The failure mode, in the buyer’s language. What breaks when they don’t have this, described the way they describe it internally at 4pm on a Thursday. Not “inefficiencies in the procurement workflow.” Something closer to “legal is the reason nothing closes in December.” Get this one wrong and your best prospects leave in about five seconds, having decided you’re describing someone else’s problem.

You against the four names on their shortlist. Honest comparison, including the two scenarios where a competitor is the better call. Comparison queries are the highest-intent thing an AI engine gets asked, and third-party review sites currently own most of them because vendors won’t write an honest one.

How to evaluate one of these without getting sold. The criteria page. Buyers ask AI engines for evaluation frameworks constantly. Whoever supplies the framework shapes the shortlist.

What it costs, what it takes, and what breaks. Implementation reality. Timelines, prerequisites, the integration that always goes sideways. This is the page that answers the anxiety nobody addresses on a pricing page.

The proof, with methodology attached. Results, with the baseline stated and the measurement window named. A number without a method is decoration. The GEO research is explicit that statistics and citations improve citation likelihood, and a stat you can’t source is a stat a careful engine has no reason to trust.

Notice what isn’t on that list. Trends posts. Roundups. Anything with a number in the headline that could be swapped for a different number without changing the piece.

The before and after

Before: “Our AI-powered platform streamlines the entire contract lifecycle, giving legal teams the visibility they need to move faster.”

Forty-eight retrieval systems read that and learn nothing they can quote.

After: “Contract lifecycle management software stores executed agreements in a searchable repository and tracks obligations, renewal dates, and approval status. It differs from e-signature tools, which handle execution but not what happens to the agreement afterward. Legal teams typically adopt it after missing a renewal deadline.”

Nothing clever happened there. Somebody just wrote down what the thing is, in a form an answer engine can lift without rewriting.

What to do with the sixty?

Keep the handful with earned inbound links and fold their substance into the relevant pillar. Redirect the rest. Delete the ones that were filler on the day they published and have not improved with age.

Crownpeak’s blog traffic grew 230% over a year when the content stopped chasing calendar slots and started answering the questions their buyers actually asked. Mitratech went from content sourcing 0% of pipeline to 40%. Same principle in both, formalized as the way to run a content system, and the rest of the numbers are here: fewer pages, each one built to be the answer rather than adjacent to it.

Open your CMS and sort by pageviews. Count how many of your published pages you’d be comfortable having an AI engine quote verbatim to a buyer who has never heard of you. If the answer is under six, you don’t have a publishing problem. You have six pages to write, and an AI Visibility Audit will tell you which six.