The session was never the point. - Feature 26 | Content Intelligence

Assemble the desk and one problem appears five times.

Search referral fell 33%. Ranking predicts 38% of AI Overview citations, down from 76%. Four citations in five come from pages outside the ranked set. The dominant channel has no paid entrance. Budgets rose 26% into a contracting delivery mechanism.

Each of those is usually reported as a distribution problem. They are also, jointly, a measurement problem, and the measurement problem is the one an individual organisation can actually fix this quarter.

What sessions were standing in for

For about twenty years, a content function could be governed by traffic. Sessions rose, the function was working. Sessions fell, it was not. The relationship held well enough that almost nobody examined it.

It held because of a chain: a person with a question searched, found your page, read it, formed an impression of you, and some proportion later bought. Sessions sat in the middle of that chain and were the only part of it that was cheap to count.

Sessions were never the objective. They were the measurable link in a chain whose actual product was an impression formed in someone's head.

The chain broke in the middle. The answer now arrives without the visit - but the impression still forms, and it forms out of material the reader may never trace back to you.

This is the awkward part. A generated answer that draws on your research, in a category where a buyer is deciding, does the work the session used to do. It just does it without telling you, without a log entry, and often without naming you.

Four things content actually produces

If sessions were a proxy, it is worth being explicit about what they were a proxy for. Four candidates, and each is measurable - none as cheaply as a session.

Figure 01

What the content function is actually for

And what each output can be measured by now that traffic has stopped serving as the universal proxy.

Being found when someone is looking. The classic job. Now split between ranked search, which still serves transactional intent, and citation, which Feature 22 shows is a substantially different game. Measure: citation presence in your category's questions, plus retained search traffic on transactional queries - reported separately, because they behave differently.

Being credible when someone is deciding. The impression the session used to deliver. This survives the traffic decline entirely, because it does not require a visit - a buyer who encounters your material through a generated answer, a colleague or a third-party reference still forms it. Measure: win rate at proposed price, discount depth, and how often your material is named unprompted in sales conversations.

Giving the sales function something to send. Almost never counted, frequently the highest-value output. A piece that closes deals while earning no traffic reads as a failure on every standard dashboard. Measure: internal usage - which assets are actually sent, by whom, at which stage.

Building the asset you own. The list, per Feature 25. Content is the acquisition mechanism for the one channel nobody else can reprice. Measure: subscriber acquisition attributable to a piece, and engagement retention over time rather than list size.

This decomposition is ours. No source proposes it and none of the four measures is standard practice. Three of the four are harder and slower to collect than a session count, which is the honest reason the industry standardised on traffic in the first place.

Why the wrong metric survives a channel collapse

A metric that has stopped working does not usually get replaced. It gets defended, because it is embedded in things that are expensive to change.

Sessions are in the dashboard, the quarterly review, the agency scorecard, and often the compensation plan. Replacing them means renegotiating all four with people who did not read the research and have no reason to believe the channel changed rather than the team underperforming.

The result is predictable and visible across the industry right now: teams reporting failure every month against a metric that is falling for reasons entirely outside their control, while the outputs that still work go uncounted because no one built a field for them.

Figure 02

Two reporting regimes, same content function

What the choice of metric does to the same body of work.

Governed by sessions

The standard dashboard in 2026

Reports declineA third of the channel is gone. The chart falls regardless of the work.

Suggests no actionNothing the team can do restores a click the answer engine absorbed.

Misprices the archivePages that build credibility but earn no clicks read as waste and get cut.

Rewards volumeMore pieces, more chances at traffic - exactly when production got cheap and distribution got hard.

Governed by what content produces

Available now, at real cost

Separates the four outputsFound, credible, usable by sales, list-building. Each can move independently.

Survives the channel changeThree of the four do not depend on a visit at all.

Makes trade-offs visibleDepth versus volume becomes a decision with evidence rather than a preference.

Slower and partly qualitativeWhich is the real reason it is not standard, and worth admitting.

This comparison is ours and it is an argument, not a finding. We have no evidence that organisations measuring this way outperform those that do not, and we are not going to imply otherwise. It is offered as a reasoned response to a documented change.

The honest difficulty

The reason traffic won as a metric is not stupidity. It was cheap, automatic, comparable across organisations and available daily. Every replacement proposed above is worse on all four counts.

Citation presence has no agreed sampling frame and no validated link to revenue, as Feature 22 sets out. Credibility effects show up in win rate and discount depth on a lag of months. Sales asset usage requires the sales function to log something they currently do not. Subscriber attribution is contested in every organisation that attempts it.

Anyone selling you a clean replacement dashboard is selling something that does not exist. What is available is a worse instrument pointed at the right thing, replacing a good instrument pointed at something that stopped happening.

That is still the better trade, but it should be made with the cost stated.

Whose interest this serves

Marketing Legendary operates a content practice and publishes original research. This feature argues that content should be measured on credibility and citation rather than traffic - which is precisely the argument that makes our kind of work look valuable and cheap-volume content look wasteful. It is the most self-serving conclusion available on this desk. We have tried to earn it by stating the difficulty of the alternative honestly and by declining to propose a metric anyone has validated. The reader should weigh it accordingly.

Starting with the cheapest one

Of the four outputs in Figure 01, sales asset usage is the one to instrument first. Not because it is the most important - credibility probably is - but because it is the only one measurable this quarter with tools you already own, and because it produces the fastest internal argument.

The mechanism is unglamorous. Every asset the sales function sends to a prospect is a vote, cast by someone with no stake in content's performance review, about which material actually helps close business. That signal exists in most organisations already and is almost never collected.

Three things it reveals immediately, and all three tend to surprise the content team:

Which pieces get used. Usually a small number, often not the ones with the most traffic, frequently including something written years ago that nobody has updated because it does not rank.
What is missing. The gaps show up as sales people writing their own material, which is the clearest possible statement of unmet need and is usually invisible to marketing.
Where in the cycle it matters. Assets used late, at evaluation and objection-handling, do work that traffic reporting cannot see because the prospect was already in conversation.

A page that closes deals and earns no traffic reads as a failure on every standard dashboard. That is not a measurement gap. It is a measurement error.

The internal effect matters as much as the data. A content function that can name which assets the sales team relies on has a different conversation about budget than one defending a falling traffic chart - and it is a conversation about contribution rather than about volume.

What not to do with it

Two failure modes are common enough to name.

Do not turn usage into a target. The moment asset usage is a metric someone is measured on, it will be gamed by producing more assets and encouraging their circulation. The signal is valuable precisely because the people generating it are indifferent to it.

Do not conclude that unused assets are worthless. Material that builds credibility with a reader who never becomes a tracked opportunity is doing output 02, not output 03. Cutting everything with low sales usage would optimise a single output and destroy the others - which is the same error as governing by sessions, with a different denominator.

What to do about it

Report sessions as context, not as the headline. Keep the number. Move it. A metric falling for structural reasons belongs beside the market data that explains it, not at the top of a page that implies the team caused it.

Pick one of the four outputs and instrument it properly this quarter. Not all four. The most common failure is designing a complete new measurement framework and shipping none of it. Sales asset usage is usually the cheapest to start and the most immediately persuasive internally.

Split transactional search from informational search in reporting. One is largely intact; the other is where the 33% went. Reporting them together produces a decline that cannot be acted on and hides the part that still works.

Establish a citation baseline now. Whatever its flaws as a target, you cannot observe a change without a starting point, and the cost of establishing one falls every quarter you wait - along with the value of the comparison.

Write down what you think content is for, before you choose what to measure. This is the same argument Feature 20 makes about design quality. A measurement framework is a statement of purpose with numbers attached, and organisations that skip the statement end up measuring whatever their tooling makes easy.

Figure 03

The Content desk, Q1 2026

Every headline figure from the six features, assembled.

Finding	Value	Feature
Distribution
Google traffic to publishers, global	−33%	21
Google traffic to publishers, US	−38%	21
Organic CTR where an AI Overview appears	−61%	21
Searches returning an AI Overview	48%	21
Zero-click, mobile	77%	21
Conflicting zero-click figures on AIO queries	83% / 38%	21
Citation
AI Overview citations from top-10 rankers	38%	22
Same, prior measurement	76%	22
LLM citations from outside Google top 100	80%	22
AI citations from earned media	84%	23
AI citations from paid or advertorial	0.3%	23
AI search visits, Q1 2026	27.4bn, +42.8%	22
Economics
Content share of marketing budget	26%	24
B2B marketers increasing content spend	61%	24
Increasing AI tooling / owned media / paid	45 / 32 / 25%	24
Email ROI claims, cost basis disclosed	None	25
Not established anywhere on this desk
True zero-click rate	-	Sources conflict
Citation share linked to revenue	-	No research
Return per published piece	-	No research
Composition of the earned-media 84%	-	Not broken out

Figures are assembled from separate research with different samples, periods and methods. The final group is the honest state of the evidence: four questions that matter more than most of the answers above them, and that nobody has published.

How we did this

Where this comes from

This feature is a synthesis. Every figure is carried from Features 21 to 25 of this desk with its original grading intact. No new external source is introduced.

What's ours, not the source's

The proxy argument, the four-output decomposition, the two-regime comparison and the difficulty section are entirely ours. No source proposes any of them.

Not validated

None of the four proposed measures has published validation against commercial outcome. We say so in the body rather than in a footnote.

Interest

Disclosed prominently. This is the most self-serving conclusion on the desk and we have said so in the text.

What this doesn't prove

That the proposed measures work. None is validated. They are a reasoned response to a documented change, not a tested framework.
That organisations measuring this way outperform. No comparative evidence exists and we did not look for a favourable study to cite.
That sessions are worthless. They remain a real signal for transactional intent and a useful diagnostic. The argument is against their use as the governing metric.
That the four outputs are exhaustive. It is our decomposition. Others are possible and we would expect a good one to differ.
Anything about the size of the credibility effect. We assert it survives the traffic decline because it does not require a visit. We cannot quantify it and no source does.
That the industry will change. Metrics embedded in compensation plans are durable well past the point of usefulness, and nothing here suggests that will not happen again.

Sources for this feature

Features 21, 22, 23, 24 and 25 of this edition, and Feature 20 of the Creative desk. Another feature in this edition - all external sourcing and grading carried from those features

End of the Content desk

Q1 2026 · Six Features

Marketing Legendary publishes quarterly. The next edition follows.