The 2027 plan is on the table and paid search is up again. Your team did nothing wrong. Conversion rates held and the agency found savings in the second quarter. Cost per opportunity still rose, after a 2025 in which Google's average cost per click climbed 7% (Alphabet 10-K, SEC), and the CFO wants to know why the same work costs more every year.
The answer is awkward to say in that room. The price moved because someone else sets it. Most of the demand in your plan does not belong to you. You rent it from Google, Meta, LinkedIn, and now ChatGPT, by the click or by the thousand, at a rate they control.
That matters more in this cycle than it did five years ago, because the landlords are raising the rent and, in AI answers, removing the door.
Who sets the price
Ben Thompson named the mechanism in 2015. The value chain in any consumer market has three parts, "suppliers, distributors, and consumers/users," and the internet moved power from the companies that control supply to the companies that own the user relationship (Stratechery, 2015). Once a distributor owns that relationship at scale, "suppliers can be commoditized leaving consumers/users as a first order priority."
Two years later he tightened the definition. Aggregators have a direct relationship with users, serve each additional user at zero marginal cost, and acquire each new user or supplier more cheaply as the network grows. For everyone else, Thompson's conclusion is that "suppliers will come onto the aggregator's platform on the aggregator's terms" (Stratechery, 2017).
You are the supplier in that sentence. Google owns the person typing the query. Meta owns the feed. When a buyer reaches you through either one, the aggregator introduced you, and it will charge you again for the next introduction.
The invoice is public. Alphabet's Google Search and other revenue went from $175.0 billion in 2023 to $224.5 billion in 2025 (Alphabet 10-K, SEC). Paid clicks rose 6% that year, on top of the higher price per click. In the second quarter of 2026 the same line grew from $54.2 billion to $63.3 billion year over year (Alphabet 10-Q, SEC). Meta's average price per ad rose 12% in the same quarter (Meta, July 2026).
Your account team controls none of those inputs. Alphabet's own filing names the factors that move its click prices, including "advertiser competition for keywords" and its own changes to ad formats, delivery, and policy (Alphabet 10-K, SEC).
The answer engine keeps the click
Search used to send traffic away, and that was the deal. Google aggregated the demand and passed a share of it to the open web.
The share keeps shrinking. In 2024, for every 1,000 US Google searches, 360 clicks went to the open web, meaning a site Google does not own, reached through a link nobody paid Google for (SparkToro, 2024). In the first four months of 2026, 68% of US Google searches ended without a click, measured on a different panel than the 2024 study (SparkToro, 2026). Pew tracked the browsing of 900 US adults in March 2025. Users who encountered an AI summary clicked a traditional search result link on 8% of visits, and users who did not encounter one clicked on 15% (Pew Research Center, July 2025). Links inside the summary got clicked on 1% of visits to pages that showed one.
An AI answer is aggregation carried to its end state. The aggregator reads every supplier, writes the answer itself, and keeps the user on its own page. Your content becomes an input to its product. ChatGPT reached 900 million weekly users in February 2026 (TechCrunch), and OpenAI had already said in January that ads were coming to its free and Go tiers (SiliconANGLE). The newest aggregator follows the same sequence as the old ones, gathering the users first and selling access to them second.
The Rent You Can't Negotiate covered the retail version, where the retailer charges you to reach customers you already won. The aggregator version reaches further up the funnel, to buyers who have never heard of you, and AI agents that complete the purchase are the step after that (Marketing to Machines). The planning question is what share of your pipeline sits behind one of these tolls, and what is left when you stop paying it.
What you keep when you stop paying
There is one test for every demand source in the plan. Imagine the spend off for a quarter and ask what is left.
Non-brand paid search leaves nothing. The click was the aggregator's customer, rented to you for one visit. Paid social leaves a little, if the ad built memory. A buyer who types your company's name into the search bar is a different asset. The aggregator still sits between you, but it did not create the intent and has a harder time steering it to a competitor.
That is the escape in Thompson's own terms. Aggregators commoditize suppliers that users cannot tell apart. A supplier the user asks for by name is the hardest input for the aggregator to swap out. Fractl tracked 1,010,848 high-volume keywords across 379 brands and found that 90% of the search volume was non-branded, "the exact query type that AI chatbots can easily replace" (Fractl, June 2026). Its conclusion was that branded query share "is the asset that holds." For most companies the branded pool is the smaller one, which is why it is worth growing on purpose.
Brand demand still carries a toll, a smaller one. Google does not restrict the use of trademarks as keywords (Google Ads policy), so a competitor can bid on your name and you pay to defend it. I treat brand-term spend as insurance on an asset you own. It is still cheaper than renting the buyer who never heard of you.
The second escape reverses the aggregator's own advantage. Thompson's aggregators win because serving one more user costs them nothing. An email list or a customer base that opens what you send gives you the same economics at small scale. Reaching your list a second time costs close to nothing, while reaching the same buyers through an auction costs full price every time. You become the aggregator of your own customers. Distribution Is the Last Moat made the strategic case for owned reach. The aggregator math adds the cost case.
In Partnership with HubSpot
A CRM so smart, it updates itself.
The Rent-to-Own Ratio
The marketing version of Aggregation Theory needs one number a CFO can track. I call it the Rent-to-Own Ratio. It is the share of new pipeline that arrived through rented or leased demand.
Rented demand is any source where an aggregator owns the user and sets the unit price. Owned demand is any source where the buyer came to you by name or through a channel you control. The table sorts the usual sources.
Demand source | Rent or own | Unit price set by | Left after spend stops |
Non-brand paid search | Rent | Google's auction | Nothing |
Paid social prospecting | Rent | Meta or LinkedIn auction | Some memory, if creative is distinctive |
AI answer ads | Rent | OpenAI | Nothing |
Marketplace and retail media | Rent | The retailer | Nothing |
Organic non-brand search | Lease | Google's ranking and AI answers | Shrinking traffic |
Paid brand search | Own, with insurance | Competitors bidding on you | Most of the demand |
Organic brand and direct | Own | Nobody | The demand |
Email, community, customer base | Own | Nobody | The audience |
Referrals and word of mouth | Own | Nobody | The demand |
Organic non-brand search sits in its own row because you do not pay per click, but Google sets the terms and the AI answer keeps more of the click every year. It counts on the rented side of the ratio.
This default is my position, and no study backs it. Past roughly $20 million in revenue, rented demand should produce no more than 60% of new pipeline. Above that line, a CPC increase you do not control becomes a margin problem you own. Run higher in two cases. An early-stage company with no name yet has nothing to own, so renting is the only way to start. A category where buyers only shop inside a marketplace may never cross the line, and the job there is to keep the ratio from getting worse.
The second default is a capture rate. Every rented click is a chance to convert rent into ownership. Track the share of paid visitors who join something you own, such as an email list or a trial, within 30 days. My default is to raise it two points a quarter from today's baseline. Exempt bottom-funnel campaigns built to catch buyers already shopping, and judge those on CAC payback instead.
Moving the ratio in the 2027 plan
Calculating the ratio takes an afternoon. Pull the last four quarters of new pipeline by original source. Sort each source into rent, lease, or own using the table, then divide rent plus lease by the total.
Bring two lines to the planning meeting. The first is the ratio today and the target for the end of 2027. My default is five points of movement a year, because brand demand compounds slowly and a faster target forces you to relabel sources until the number looks right. The second is the rent increase you expect to absorb anyway. Use the platforms' own disclosures for that estimate. Google's cost per click rose 7% in 2025 and Meta's price per ad rose 12% in the second quarter of 2026, before anyone touched a bid.
Rented demand is a variable cost with an escalator the landlord controls. Owned demand behaves like an asset. You pay to build it, and it keeps producing after the spend stops. That is the same capex language from Brand Is a Balance Sheet Item, applied to where the pipeline comes from. Fund the brand and audience lines out of the non-brand rent line, a few points a year, and report both the ratio and blended CAC together so the CFO sees the trade as it happens.
Before your next planning review, put the ratio on one slide with last year's number beside it. If nobody in the room has seen it before, the plan has been renting most of the pipeline without anyone deciding to.
James