Expense, or Asset

For a decade the question has had a settled shape. A budget meeting reaches the line marked search, and someone asks which channel returns more for the money — organic or paid. The answer gets assembled from cost-per-click, cost-per-lead, a blended target, a recommended split. It is a reasonable question, asked in good faith, and it has produced a great deal of careful analysis.

It is also the wrong question. Or more precisely: it quietly assumes something that isn't true — that organic and paid search are two prices for the same thing, and the work is to find the cheaper one.

They are not the same thing. They behave differently over time, they leave behind different residue, and they sit in different places on the only ledger that matters by the end of the year. Comparing their cost-per-click is like comparing the monthly cost of leasing a car against the monthly cost of owning one and concluding the lease is cheaper. This month, it is. That was never the point.

Paid search buys a position at the top of the page for exactly as long as the money flows. It is fast, precise, and entirely contingent. Stop the spend on a Tuesday and the position is gone by Wednesday — not diminished, not matured, gone, reverted to the state it held before the first dollar. Whatever was paid bought visits, and the visits have been spent.

Organic search behaves the other way around. A position earned through substance and structure keeps returning visitors after the work that earned it has stopped. It does not switch off when attention turns elsewhere. It holds, and — maintained — it tends to gather rather than fade, because each piece that earns its place lends standing to the next.

That single distinction reorganizes the whole decision. The moment the two channels are seen as different kinds of thing rather than two prices for one thing, "which is cheaper" stops being answerable in the terms it was asked — because one of them is a recurring expense and the other is something that accrues. You can compare the cost of two expenses. You cannot honestly compare an expense against an asset by lining up their monthly invoices.

Rent has a direction, and the direction is up. The average cost of a Google search ad has risen every year for five years running — most recently to around $5.26 a click, from $4.66 the year before, across more than sixteen thousand campaigns. The trajectory has carried into 2026. This is the ordinary physics of an auction for a fixed amount of space: as more bidders compete for the same positions, the clearing price climbs.

So the rented position does not hold its price. It costs more each year to stand in the same spot — and none of that spend converts into anything you keep. A decade of paid search, run well, leaves you precisely where you began the morning you stop: at zero, with a portfolio of receipts. The spend was not wrong. It simply never became anything.

This is why optimizing the cost-per-click, while worth doing, cannot answer the larger question. A more efficient rent is still rent. Lowering the price of attention you do not keep changes the size of the recurring line, not its nature.

The earned position runs on the opposite logic. Its cost is concentrated up front — the work to deserve the ranking arrives before the return does, and the return arrives on a lag measured in months rather than afternoons. That lag is real, and it is the honest cost of the asset. It is also the whole of the difference.

Because once the position exists, the visits it returns are no longer billed. The marginal cost of the next visitor to an earned page trends toward nothing, while the marginal cost of the next paid visitor is fixed at the auction price, forever. One curve bends down as it matures; the other runs flat and high for as long as you are willing to pay it. A page that earned its standing two years ago can still be returning qualified visitors today, having been paid for once.

This is not a claim that organic is free. It is the observation that organic front-loads its cost into something that lasts, where paid spreads its cost across a result that does not. Same dollars, different residue. One leaves you with receipts; the other leaves you with a position.

There is a quieter consequence, and it is the one that decides the matter.

The healthiest things in a business are the ones you protect — positions already working, that you maintain and defend because they return value reliably. Below those sit the things you are still building, where the presence isn't there yet and the work is to open it. Every channel a business runs lives somewhere on that line: something to hold, or something to build.

Rented attention can never move from one to the other. It cannot graduate into a position you protect, because it has no persistence to defend — the instant the spend stops, there is nothing there to maintain. Paid search is permanently a thing you are building, never a thing you have built. It opens a door and holds it open with your hand against it; the day your hand comes away, the door closes. An earned position is a door that stays open on its own.

That is the asymmetry beneath the cost question. Not that one channel is cheaper, but that only one of them is ever finished — only one can move from the column of perpetual effort into the column of things that simply keep working.

This distinction is sharpening, not softening, as search itself changes.

As AI-generated summaries take up more of the results page, two things move at once. The rented surface grows more contested and more expensive, because the space available to buy is shrinking against rising demand. And the free click grows scarcer: when an AI summary sits atop the results, people follow a link to a website roughly half as often — about 8% of the time, against 15% when no summary appears — and they click a source named inside the summary only around 1% of the time, on Pew Research Center's reading of real browsing in March 2025.

It would be easy to read that as a threat to organic. It is closer to the reverse. The position that holds its value in a summarized landscape is the one substantial enough to be cited as the source — and that is an earned asset, not a rented placement. You cannot buy your way into being the thing the machine quotes; you can only become it. The shift makes rented attention pricier and less certain at the very moment it makes durable substance more valuable. We take this up directly in a companion piece on AI visibility — there, too, the asset is the substance worth referencing.

The new landscape rewards exactly the thing that compounds. That is worth sitting with before the next budget is set.

Step back from the channels for a moment, and the real distinction is not about search at all.

A business that funds its growth entirely through rented attention is running to stand still. It can be a very well-run race — efficient bids, sharp creative, disciplined targeting — and it will still end each year holding nothing it did not hold the year before, because every dollar bought a result that was consumed on arrival.

The discipline, then, is not choosing the cheaper channel. It is knowing, for each dollar, whether it buys a result you rent or builds a position you keep — and then funding both on purpose, with clear eyes about which is which. That clarity is the entire skill. It is also surprisingly rare, because the monthly invoice for paid is visible and immediate while the slow accrual of an earned position is neither — and what is easy to see tends to get managed, while what is easy to keep tends to get deferred.

None of this is a case against paid search. Paid is the correct instrument for a specific set of jobs, and for those jobs nothing else comes close: speed when a launch cannot wait, reach into demand you have not yet earned a position for, a controlled test of a message before content is committed to it, capture of a time-bound moment that will be gone before any organic position could mature. Used for those, paid is not an expense pretending to be an asset — it is an expense doing exactly the honest work of an expense.

The error is never using paid. The error is mistaking a rented result for a built one — funding the recurring expense year after year and recording it, in the place where it counts, as an investment. Both belong in a serious plan. They simply belong in different columns, and the whole of the intelligence is in not confusing them.