Accountable for
what they do not own.

Run the desk backwards and a pattern appears that none of the individual findings states.

Feature 27: the structure of the search market is being redrawn by a court, on a timetable nobody in marketing influences. Feature 28: whether you are readable at all is set in a file usually owned by engineering. Feature 30: whether your content exists in the served document depends on a build pipeline most marketing teams cannot inspect. And on the Content desk, Feature 23 established that 84% of AI citations come from earned media - material published by third parties.

Now ask what an SEO function actually controls.

The right column decides the outcome. The left column gets the review.

Why this is not a complaint

Every discipline works inside constraints it does not set. Paid media does not control the auction. Sales does not control the product. The observation only matters if the accountability structure has stopped matching the causal structure, and here it has, in a specific and recent way.

Under conventional search the alignment was reasonable. The dominant inputs - content quality, site structure, links - sat close to or inside the SEO function. A team that did its job well generally saw the result, and where it did not, the reason was usually diagnosable within the same remit.

What changed is that the decisive inputs migrated outward at the same time as the measurable output degraded. Feature 21 on the Content desk documents a 33% fall in search referral. So the function is now judged on a declining number produced largely by decisions taken elsewhere.

That is a structurally unfair position, and more importantly it is one that produces bad decisions. A team measured on something it cannot move will optimise the part it can - which means more on-page work, more published pages, more keyword coverage - precisely the activity the rest of this edition suggests has the weakest returns.

The brief that cannot be written

There is a specific version of this that shows up in every organisation trying to respond seriously.

Feature 23 established that citations come overwhelmingly from earned media, and that the most reliable route in is publishing original data that third parties reference. So the correct instruction to an SEO function is something like: cause our organisation to produce findings other people want to cite.

That is not a brief. It is not commissionable, not schedulable, and not deliverable by the function receiving it. It requires access to proprietary data, permission to publish it, subject-matter time from people with other jobs, and a tolerance for being wrong in public. None of those sit inside a search function's authority.

This is the mechanism by which organisations that understand the problem still respond to it wrongly. The correct action is not available to the person who understands it, so the available action gets taken instead.

What the function is actually good for now

The useful reframing is that a modern search function is closer to an internal diagnostic and coordination role than to a production one - and that this is a promotion in substance if it is recognised.

The accountability that would follow

If the inputs moved, the measures should move with them, and the honest version is uncomfortable for everyone.

An SEO function should be held to eligibility, not to outcome. Is the site readable to the crawlers we have chosen to permit. Does the served document contain the content. Are the two markets reported separately. Is there a citation baseline. Those are all within the function's control or its power to escalate, and all are currently unmeasured in most organisations.

Citation outcomes should be held jointly with whoever owns third-party publication. Feature 23's 84% is not a search function's number. Holding one team to it while another controls it produces the misallocation without producing the result.

We are aware this reads as an argument for the function to be judged on inputs rather than results, which is usually a bad idea and is often how underperformance is defended. The distinction is that these particular inputs are necessary conditions - a site serving twenty-eight nodes cannot rank or be cited no matter how good the rest is. Necessary conditions are the right thing to hold someone to when they hold nothing else.

The failure this produces, specifically

An accountability structure misaligned with a causal structure does not produce inaction. It produces a predictable and identifiable pattern of misdirected effort, and it is worth naming because it is visible in most organisations right now.

A team measured on organic visibility, controlling only on-page work, will do more on-page work. That is not irrational; it is the only lever within reach. So the response to a channel contracting for structural reasons becomes: publish more pages, cover more keywords, refresh more existing content, produce more of the material the Content desk's Feature 21 identified as most exposed to substitution.

The observable symptom of this misalignment is a content calendar that grew in the same year the channel shrank.

Three secondary effects follow, and each is commonly misdiagnosed as a performance problem.

• Escalations stop happening. A team that has raised rendering or crawler policy twice without resolution stops raising it, and the necessary conditions quietly go unmet.
• The wrong work gets defended. When output volume is the only demonstrable activity, it becomes the thing presented in reviews - which entrenches it further.
• Capable people leave. The most diagnostically able practitioners are precisely the ones who can see that the decisive inputs sit elsewhere, and they tire of being measured on them first.

None of this is measurable and we are not presenting it as research. It is a mechanism consistent with the ownership structure in Figure 01, offered because naming it is usually enough for someone to recognise whether it is happening to them.

What to do about it

Write down who owns each input in Figure 01. Not who should - who does, by name. Most organisations have never done this and the exercise itself produces the argument.

Escalate the two you do not own and cannot work without. Crawler policy and rendering. Both are binary, both sit outside marketing, and both remove you from consideration entirely when wrong.

Convert the uncommissionable brief into a question. A search function cannot commission original research. It can arrive with the ten questions its market asks that nobody has answered with data, which is a considerably more persuasive document than a content calendar.

Propose the joint measure before someone imposes the single one. Citation outcomes shared with communications is a better structure than either function owning it alone, and it is much easier to establish before a bad quarter than after one.

Measure the necessary conditions monthly. Rendering, crawler eligibility, canonical integrity, index coverage. Cheap, binary, and the only things in this feature that are entirely within reach.