Somewhere in your 2027 commercial plan is an assumption that your buyer finds you by searching. Find the page it is written on. In most plans, it is not written anywhere, and that is exactly what makes it dangerous.
The path lost a step.
The old sequence had four steps. Search, results, your page, decision. Three of those four involved something you produced, and one of them was a page you owned entirely.
The new sequence has three. Ask, answer, decision. The step that disappeared is the one you owned.
This is not a prediction, and it is not evenly distributed. In some categories it has barely begun. In others, a majority of early-stage evaluation now completes without a click or search. The point is not that it has happened everywhere. The point is that almost nobody has measured whether it has happened in their category, while every number in their plan assumes it has not.
What inherits this assumption
More of the plan than most people realize, because the assumption sits underneath rather than inside the numbers.
• Traffic targets, which assume a click that may no longer occur
• Content budget, which is justified by the traffic it produces rather than by the citation it earns
• Cost per lead, which is calculated from a funnel that has lost a stage
• Conversion assumptions, because a buyer arriving pre-informed converts differently from one arriving cold
• Sales cycle length, which shortens when the buyer has already formed a position and lengthens when that position favors a competitor
• Headcount, where demand generation roles are sized against a lead volume model
None of those numbers is wrong. All of them rest on a premise that has not been checked.
The signature to look for in your own analytics
There is a specific pattern that indicates this is already happening to you, and it is visible with data you already hold.
Flat or declining sessions, combined with stable or rising qualified inbound.
That combination is counterintuitive, and it is the clearest in-house evidence available. It means fewer people are arriving at your site, but the ones who do arrive are further along and better informed. They have completed the early evaluation somewhere you cannot see, and they are contacting you having already formed a position.
The usual interpretation of falling sessions is that the content program is underperforming, which frequently leads to the content budget being cut at precisely the moment citation work should be funded. Segment the two signals before drawing that conclusion.
The second signature
Sales reporting that buyers arrive with a formed view of the competitive landscape that nobody in your organization supplied. If your representatives are increasingly spending the first call correcting an impression rather than building one, that impression was formed at a layer you are not present in.
How to produce a defensible number in one afternoon
You do not need a research program. You need twenty questions and two engines.
Step one: assemble the twenty questions
Your twenty highest-intent buyer questions phrased the way a buyer would phrase them rather than the way your category describes itself. Take them from sales call notes, not from a keyword tool. The phrasing matters more than the topic.
Step two: run each on two engines, twice, on separate days
Two runs, because variance between runs is real and a single-run number will be challenged by the first person who tries to replicate it. Record the variance rather than averaging it away.
Step three: count complete answers
How many produce an answer complete enough that a reasonable buyer would not click through to any source. Be strict about this. An answer that names three companies and gives no detail is not complete. An answer that explains the trade-off and names a recommendation is.
Step four: state it as a range
That percentage, expressed as a range across engines rather than a point estimate, is the share of your funnel that no longer passes through anything you own.
An afternoon. And unlike almost any other number in your plan, it can be reproduced by anyone who doubts it, which is what makes it hold up in a room.
What to do with the number
Do not lead with a recommendation. Lead with the assumption. Write one line into the plan document: this plan assumes that approximately X percent of category buying questions still route through a click, reviewed on a named date by a named owner.
That single line does three things. It converts an invisible premise into a stated one. It gives the number a date and an owner. And it creates the condition under which the plan can be revisited on evidence rather than abandoned on politics, which is the mechanism described in the piece on revisit triggers.
It also reframes the conversation. A marketing leader arguing that AI search matters are making a case. A commercial leader stating that a named percentage of the funnel has moved outside the organization’s control, with a review date attached, is describing an operating risk. Those two statements receive very different treatment in the same room.
The objection, and why it is partly right
The predictable objection is that the measurement is crude. Twenty questions is a small sample, two engines is not all engines, and the definition of a complete answer involves judgment.
All of that is true, and this is not a reason to skip it. The alternative is not a better number, but no number. An unstated assumption carried forward into a plan that will be executed for twelve months.
State the limitations alongside the finding. A crude number with its weaknesses named is more useful and more credible than a precise-looking number with its weaknesses hidden, and considerably more useful than silence.
What this does not mean
It does not mean abandoning search. The click-through path is still substantial in most categories and will remain so for years. It does not mean the answer is more content. In many categories, the constraint is entity resolution rather than content volume, and publishing more into a fragmented entity produces nothing.
It means writing down an assumption that is currently invisible, so that when it moves, somebody notices. That is the entire ask; it costs an afternoon, and the window closes when the plan locks.
Frequently Asked Questions
How has the B2B buyer path changed with AI engines?
The traditional path had four steps: search, results, a visit to a company page, and a decision. The emerging path has three: ask, answer, decision. The step that disappeared is the company-owned page, which means a share of early-stage evaluation now completes without any contact with material the company controls.
How do I measure how much of my funnel no longer passes through my site?
Take your twenty highest-intent buyer questions phrased as buyers phrase them, run each on at least two AI engines twice on separate days, and count how many produce an answer complete enough that a reasonable buyer would not click through. Express the result as a range rather than a point estimate.
What analytics pattern indicates this is already happening?
Flat or declining sessions combined with stable or rising qualified inbound. Fewer people arrive, but those who do are further along and better informed, having completed early evaluation somewhere the company cannot observe.
Why is falling traffic often misdiagnosed?
Because declining sessions are usually read as content underperformance, which leads to the content budget being cut at the point when citation work should instead be funded. Segmenting sessions from qualified inbound before concluding avoids this.
Which numbers in a commercial plan inherit this assumption?
Traffic targets, content budget justification, cost per lead, conversion assumptions, sales cycle length, and demand generation headcount. The assumption sits underneath these numbers rather than inside them, which is why it is rarely stated.
Is a twenty-question sample defensible?
It is crude, and it is reproducible, which matters more. State the limitations alongside the finding. A crude number with named weaknesses is more useful and more credible than no number at all.
Does this mean abandoning search?
No. The click-through path remains substantial in most categories. The purpose is to write down an assumption that is currently invisible so that when it moves, somebody notices.
How should the finding be written into a plan?
As a single stated assumption with a percentage range, a review date, and a named owner, recorded in the plan document itself. This converts an invisible premise into one that can be reviewed on evidence.
External Sources