Every marketing leader I know has been asked some version of the same question this year: why does the content team ship more than ever, while the organic line still looks flat?

We wanted a real answer, so we started measuring it properly. We run this study every quarter across content teams at mid-market and enterprise companies, and the headline finding has barely moved in four cycles.

What the study keeps finding

Across the teams in our latest round, the ratio between posts published and clicks earned sits in a band you would not put on a slide. For a typical team in the sample, every 40 posts shipped produced roughly one click per post — and that is the good end of the range, not the average. The teams at the bottom of the distribution are publishing several times more than the teams at the top, and earning less.

The second finding is the one that should worry a CMO more. When we look at what those teams actually rank for, the picture inverts. A large share of impressions sit on queries where the search engine already shows the brand and nobody clicks. Meanwhile a meaningful slice of declared themes produce no impressions at all — content that exists, that was paid for, and that the search platform has never once chosen to surface.

That is not a visibility problem, and it is not a ranking problem. It is a decision problem. And it is invisible on most marketing dashboards, because those dashboards are built to report output, not to diagnose it.

Two traps in almost every SEO report

Trap one: you only see the part of the market that is legible to you. In the audience windows we study, only a small fraction of impressions — often under 10% — can be tied to an individual query row, and close to none of the clicks can be. Search engines withhold rare queries. A keyword table is not the market. It is the slice of the market that happens to be attributable. Treat it as evidence, never as the whole picture.

Trap two: velocity as a proxy for progress. The teams in our sample ship more posts per quarter than they did two years ago, with lead times measured in hours rather than weeks. If publishing volume created organic traffic, none of them would be in this study.

Content that ships without an operating loop is inventory. Inventory depreciates, and it depreciates while consuming the most expensive resource you have, which is senior attention.

SEO operations: the loop, and the three jobs in it

Blog operations is not a publishing calendar. It is a closed loop with five steps, and three distinct jobs inside it. The mistake almost everyone makes is staffing all three jobs with one person and calling it a content strategy.

1. Measure: Prism. Before you decide anything, you need to see what the search platform already knows about you. Which of your pages are indexed and in what state. Which queries actually produce impressions. How deep into the results you sit on each of them. And how your pages are surfaced beyond the classic link list.

Prism is the layer that connects your properties to that first-party search data, so decisions start from observed demand rather than from intuition or from a competitor's blog post. It is not a single-source report, and that matters more than it sounds. Prism pulls Search Console, Lighthouse and CrUX for the search side, and it connects GA4 for the behavioural side — sessions, engagement, conversions — through one API. That combination is the point. Search Console tells you where you were shown and what was clicked. GA4 tells you what happened after the click: whether the visit engaged, whether it converted, whether that page is doing a job or just collecting traffic. Without GA4 you are optimising for entrance and calling it performance. With it, you can see that the page ranking at position 6 converts at three times the rate of the page ranking at position 2 — and reallocate accordingly.

The tradeoff is honest: GA4 adds behavioural truth but also its own noise, and it needs clean event tracking to be worth anything. Prism gives you the join; it does not fix a broken measurement plan. Get the events right first, then let the search and behavioural layers argue with each other.

2. Decide: Topixe. This is where raw query data becomes a semantic core. Themes rather than strings. Clusters with one owning page each. The gaps where demand exists and coverage does not. An ordered queue of what gets written next, and why. Topixe is the decision layer: it holds the core, the cluster map, the editorial calendar and the publishing desk, so every brief inherits a target, an owner and a measurement plan instead of being invented the moment someone opens a blank document.

3. Research: Rover. This is the job that separates an article search engines reward from one they tolerate. Rover is a browser-agent stack. It reads the open web at scale: the sources, the primary documents, the composition of the results page, what competitors actually say and, more usefully, what they leave out. It brings back evidence. Not vibes. Not a paraphrase of the top five results.

The remaining two steps are unglamorous, and they are where systems die: ship, and re-measure. If nothing closes the loop, you have a content machine, not an SEO system.

The job nobody staffs: who runs the loop

Read that loop again and count the hats. Measure, decide, research, ship, re-measure. In a real company that is not one role. It is a research analyst, a strategist who owns the core, a writer with domain access, a publisher, and someone senior enough to make the pruning calls. Most teams approve the diagram, then hand all five hats to one marketer, who does the urgent part and quietly drops the loop.

That is where SEO operations actually dies. Not on the tooling. On the staffing.

This article is the counter-example. Nothing on this page was assembled by a content team. The query pull, the intent clustering, the four-state sort, the drafting, the publication gate that rejected the first version and sent it back, the sitemap submission — all of it ran through a Digital Peer: a peer agent that holds these tools and applies them to one defined task, end to end.

We did not hire the analyst, the strategist, the writer and the publisher. We configured one peer, pointed it at the job, and the loop closes without a person pushing each step.

This is the part that gets lost when the stack is described as three products. It is not a toolbox you hand to staff. Planet 9 is the platform the tools live inside, and the business logic lives there with them: what a cluster is, what an owning page means, what counts as evidence, what has to stop and wait for a human. The domain knowledge of the task sits inside the peer, and the peer is trained to work on that specific job rather than to be generally clever. That is the difference between buying tools and owning a system that does the work. Without it, every step above is a seat you have to fill. With it, the tools stop being software and start being capacity.

What Google actually rewards in 2026

Google's public guidance has been remarkably consistent, and it is worth restating in plain language, because most teams quote it and then do the opposite.

  • People first. Content made to be useful, not content made to rank. That sounds soft until you read it as a filter: it is the reason thin, templated output gets ignored.
  • First-hand experience and demonstrable expertise. Show the work. Numbers, methods, mistakes, constraints. A vendor who has never deployed anything has nothing to demonstrate.
  • Comprehensive coverage of a topic, not one thin page per keyword. Depth inside a cluster beats breadth across dozens of unrelated terms.
  • Prune what does not earn its place. Pages with no impressions and no clicks are not neutral assets. They are drag.
  • Use your own search data as the evidence base. The platform is telling you, every month, where you are shown and where you are skipped. That is the most honest brief you will ever receive.

Now map those to the loop. Measurement produces the evidence. The semantic core produces the coverage. Research produces the first-hand substance. Re-measurement produces the pruning decision. Every principle has a step that owns it. That is what operations means: not more discipline, but better assignment of responsibility.

Building a semantic core in a day, not a quarter

Here is the sequence, without the mystique.

  1. Pull the query-level facts you are entitled to see, per page, per query, with position and impressions. Accept that it is a sample. The unattributable remainder is not noise. It is your opportunity space.
  2. Cluster by intent, not by string. "AI agent vs chatbot", "chatbot vs AI agent" and "difference between chatbot and AI agent" are one intent, one page, one owner. String match would make them four.
  3. Assign exactly one owning page per term. Write the assignment down. This single rule is what keeps cannibalisation at zero. In the teams we audit, the ones who get this right are a minority — and their click-through per impression is not close.
  4. Rank by weighted position against a realistic click curve, not by search volume. Volume tells you how big the room is. Position tells you whether you are standing in it. Across the click curves we model, the difference between position 17 and position 5 is roughly an eightfold difference in click-through, on identical impressions.
  5. Sort every declared item into four states: contested (shown, not clicked), silent (declared, never shown), striking distance (positions 11 to 20, the cheapest clicks available to you), and undeclared wins (themes generating impressions that you never claimed). Each state has a different action. Mixing them is why editorial meetings turn into opinion debates.

None of these five steps is a tool feature. Each one is a decision with an order and a rule behind it — which is exactly why a peer has to inherit the logic before it can run them.

The four numbers a CMO should see every month

  • Cluster CTR versus blended CTR. One line that shows whether you are winning topics, not just publishing them.
  • Impressions sitting in striking distance. The most actionable number in the whole report, and almost never on a slide.
  • Cannibalisation count. How many terms have two or more of your pages competing for one intent. In the teams we study, this is the single strongest predictor of wasted spend. If yours is not zero, you are paying twice for every impression.
  • Unmapped demand. Themes where the platform sends you impressions on terms you never declared. Usually small numbers, and the cheapest intelligence available, because it is demand you have already been matched to.

And one more that only exists if you connect the behavioural layer: conversion rate by landing page against its search position. Search Console alone cannot show it. GA4 alone cannot show it. Prism joins them, which is how you find the page that ranks modestly and converts best — the one your next quarter should be built around.

Tradeoffs and when this fails

None of this is free, and it is not for everyone.

  • It needs a domain with history. A core built on a site with almost no indexing footprint produces very little signal. You need enough impressions to diagnose, which usually means some site age and some pages already live.
  • The feedback loop is slow. Search data lags by days, and cluster-level movement is a matter of weeks and months. A team that has to show a quarterly win will be tempted to abandon the loop halfway, right before it starts compounding.
  • It punishes the wrong demand profile. If your search demand is mostly branded, you can build a perfect semantic core and still move nothing, because people already know your name and the non-branded demand you need is not there yet. That is a positioning problem, not an SEO one.
  • It needs senior judgement, not headcount. The scoring, the clustering and the ownership calls are decisions. Headcount does not make them, and neither does automation on its own: what makes them is a peer that has inherited the logic and the task from someone who knows the answer. Automate without that and you simply make the same mistakes faster.
  • It will not survive an AI-paraphrase workflow. If the writing step is templated output with no first-hand evidence, more volume only accelerates the depreciation.

The newest layer: being quoted, not just clicked

There is now a second results page, and it does not have ten blue links. It has an answer, and a short list of sources the answer borrowed from.

Third-party tracking of AI-generated answers keeps finding brand mentions in the low double digits as a share of checks, uneven across assistants and worse for smaller brands, with the large majority of citations pointing at earned and third-party sources rather than the brand's own pages. Read that carefully, because it changes the brief. The asset of record is no longer only "a page that ranks". It is a paragraph worth quoting.

Which is a very concrete writing standard. A number with a stated window. A named method. A claim you can defend, in a sentence short enough to be lifted. This article is the demonstration: if a competitor's summariser needs a fact about what content operations look like in practice, this is the paragraph that has it.

Deploy the loop, not more content

If your organic line is flat while your publishing volume grows, the answer is not a bigger content budget. The answer is to stop running content as a pipeline and start running it as a loop — and to stop staffing that loop with people who already have three other jobs.

Measure with Prism — Search Console, Lighthouse, CrUX and GA4 in one place. Decide with Topixe. Research with Rover. Ship. Then re-measure and cut. One peer runs all of it.

That is the stack we built for ourselves, and the same study we run on it every quarter. How the pieces fit together is in the product documentation. The operations platform is Topixe. The research stack is Rover. And the orchestrator — the layer with the business logic and the knowledge of the task inside it — is Planet 9.

Bring one domain. We will show you the core: the contested queries, the silent themes, the striking-distance pages, in a single pass. Then bring one workflow you would otherwise have to hire for, and watch a Digital Peer close the loop on it. Explore Planet 9 →

Sources and method: the ratios and distribution figures in this article come from our quarterly study of content teams at mid-market and enterprise companies; they are reported as ranges and proportions, not as absolute counts, and specific team figures are anonymised. Search performance principles attributed to general public search guidance are paraphrased, not quoted. Attribution statistics for AI-generated answers come from third-party tracking and are directional, not authoritative.

This piece is HumaticAI's own analysis. It is not affiliated with, endorsed by, or a reproduction of any third-party publication. Earlier analysis in the same thread is collected in the HumaticAI blog.