The Economics of Being Found: Discovery Costs and Who Pays Them

A Toll That Predates the Internet

Every transaction between a buyer who does not yet know a seller and a seller who wants to be known requires closing a distance, and closing that distance has always carried a cost. For roughly two centuries in market economies, that cost has been paid by the seller and collected by whoever controls the channel through which attention flows. The history of that cost is not primarily a story about advertising becoming more sophisticated. It is a story about the price of being found moving through three distinct economic forms, with the identity of the toll collector changing each time the form changed.

In the first form, running from the 1830s through most of the twentieth century, discovery was bought as a fixed cost. A business purchased space or airtime in bulk, at a rate a publisher or broadcaster set, with no reliable way to know which of those purchases reached a genuine prospective buyer and which reached no one who could ever buy anything. In the second form, opened in 1998 by a small Pasadena start-up and refined two years later by a young search company named Google, discovery became a variable, transaction-priced cost. A business paid only when a specific person clicked, at a price set in real time by a live auction, and that auction-set price became the input to a single ratio, customer acquisition cost, that has governed venture-backed growth for most of the past decade, climbing for most of that time. In the third and newest form, now taking shape as search results give way to written answers, the cost of discovery is becoming something the first two forms never were: unpriced. For most businesses there is no invoice to pay and no bid to place for being named inside a generated answer.

At each of these transitions the party who captured the toll also changed, and each new collector held more of its market, and disclosed less about how the price was actually set, than the one before it. That pattern, more than any single figure inside it, is the argument this study makes.

The Penny Press Invents the Toll

The economic machinery this history traces begins with a price cut. When Benjamin Day launched The Sun in New York City on September 3, 1833, he sold it for one cent, far below the six cents most papers charged, and the arithmetic only worked because advertising, not the reader, covered the difference. The paper said as much about itself: its stated purpose was to lay the news "before the public, at a price within the means of everyone, all the news of the day, and at the same time offer an advantageous medium for advertisements." The Sun's circulation grew fast enough that its presses went from printing 4,000 copies an hour in 1840 to 18,000 an hour by 1851. The penny press did not invent advertising, but it did something more specific: it made a media owner's audience, not the advertiser's product, the thing actually being sold, with the advertiser paying the toll that subsidized everyone else's access to information.

The first person to formalize collecting that toll as a stand-alone business was Volney B. Palmer, who opened an office in Philadelphia in 1841 and is generally credited as the founder of the first American advertising agency. Palmer's business, though, ran in the opposite direction from what the word "agency" implies today. He bought space in bulk from newspaper and magazine publishers at a discount, then resold slices of it to businesses that wanted to advertise, keeping the markup. He worked for the publishers, not the advertisers, and had no hand in writing the advertisement itself. By 1849 his network reached from Pennsylvania to Alabama. The first toll collector in this history, in other words, was aligned with the seller of space, not the buyer of visibility.

That alignment did not shift until 1875, when N.W. Ayer & Son, founded in Philadelphia in 1869 by twenty-one-year-old Francis Wayland Ayer, introduced what the agency called the open contract. Under it, Ayer worked exclusively for the advertiser for an agreed period, negotiating the lowest price it could get from a publisher on the client's behalf, with the client paying the actual media cost plus a disclosed commission that eventually standardized around 15 percent industry-wide. Business historians have called the open contract the single most consequential structural change in the agency business, because it was the first time an advertiser's agent was paid to represent the advertiser's interest rather than the publisher's. What the open contract did not change is worth stating plainly: the underlying price of space was still fixed by the publisher, still bought in bulk according to a rate card, and still carried no mechanism for connecting a specific ad exposure to a specific sale.

Wanamaker's Confession and an Unmeasurable Spend

The best-known line about advertising's economics is also its most unreliable citation. "Half the money I spend on advertising is wasted, the trouble is I don't know which half" is almost always attributed to Philadelphia department-store owner John Wanamaker. The earliest documented appearance of anything like it, according to the researchers at Quote Investigator, is a 1919 speech by the Reverend Roy L. Smith at a Bible conference in Indiana, in which Smith credited the line to Wanamaker, who was then still alive and would not die for another three years. No direct citation of Wanamaker himself saying or writing it has ever been located, in his own papers or anywhere else, and the same line has separately been attributed over the decades to the British soap magnate William Hesketh Lever. The honest position is that the quote's authorship is unresolved and probably unresolvable.

What is not in dispute is the economic condition the line describes, and why it was true regardless of who first said it. Mass-media advertising in this era was sold as an undifferentiated exposure to a publication's total circulation or a broadcast's total audience, priced by a rate card set once and applied to everyone who bought that slot. There was no return signal at the level of an individual reader or listener: a business could know how many copies of a magazine were printed, and could not know how many of the people who saw its advertisement inside that magazine ever became a customer. Under those conditions, "half is wasted" is not really a joke about ignorance. It is close to a literal description of buying an average outcome across a mass audience with no way to isolate which member of that audience the money actually moved.

Procter & Gamble's Ivory soap campaign is a concrete instance of the model. The company introduced Ivory in 1879 and ran what is generally regarded as one of the earliest national U.S. advertising campaigns to build it into a brand, built around the claim that the soap was "99 44/100 percent pure"; company founder Harley Procter pushed to expand that spending nationally through newspapers starting in 1882. The campaign worked, in the sense that Ivory became one of the most durable brand names of the following century, but the mechanism by which it worked, repeated exposure across a national readership, was exactly the fixed, unmeasured spend the Wanamaker line describes. A business could choose to buy more of that exposure or less of it. It could not, in any rigorous sense, buy the specific reader who was about to make a purchase.

Mail Order Tries to Close the Loop

The first serious attempt to close the measurement gap did not come from the newspaper business at all. It came from mail order. Richard Warren Sears and Alvah Curtis Roebuck founded their mail-order company in 1886 and issued their first catalog in 1888. Unlike a newspaper advertisement bought in bulk against an undifferentiated readership, a mail-order catalog was sent to a specific, named list, and it sold directly against an itemized, testimonial-backed offer rather than a brand impression, with the order that came back in the mail serving as a direct, attributable record of what worked. In 1892 the company tested the model directly: it mailed 8,000 postcards carrying a simulated handwritten message and received roughly 2,000 orders back, a documented, computable response rate decades before the phrase "customer acquisition cost" existed in any business vocabulary.

Sears built on the model with "Send No Money" offers that let a customer order goods from the catalog and pay only on receipt, and with product descriptions that included buyer testimonials, an unusual practice at the time that competitors avoided because it consumed too much page space. By 1916 the company was mailing more than 50 million catalogs a year. The limitation of the model is as important as its innovation: mail order and, later, the redeemable coupon, closed the measurement loop only for a list a business already owned, rented, or built through a prior order. Neither did anything for the much larger, much harder problem of a shopper who was actively looking for a category of business they did not yet know existed, the person searching for a plumber rather than opening a mailer from one. That gap, between a controlled list a seller already owns and an open search a buyer is actively running, is exactly the gap the auction described in the next section was built to close, more than a century later.

The Auction: Discovery Becomes a Real-Time Priced Market

CAC Becomes the Central Number, and the Price of Admission Keeps Rising

The Answer Era: A Cost With No Posted Price

The newest shift in this history is structural rather than incremental: search results are increasingly giving way to a single written answer, and the economics of appearing inside that answer do not yet resemble an auction at all. Pew Research Center tracked the actual browsing behavior of 900 U.S. adults across 68,879 real Google searches in March 2025 and found that when a model-written summary appeared, which happened in about 18 percent of the searches in the sample, users clicked through to a traditional web result only 8 percent of the time, against 15 percent when no summary appeared. Clicking a link inside the summary itself happened in just 1 percent of all visits, and users abandoned the search session entirely more often after seeing a summary, 26 percent of the time, than without one, 16 percent. That is a measured finding about one sampled month, published July 22, 2025, not a projection about the future.

The forward-looking figures are, honestly, projections, and this study treats them as such rather than as settled fact. Gartner forecast in a February 19, 2024 press release that traditional search engine query volume would fall 25 percent by 2026 as generative tools substitute for search; whether that specific figure lands on schedule is genuinely contested and, as of this writing, unresolved. McKinsey's October 2025 research estimated that roughly half of consumers already seek out AI-powered search intentionally, and that AI-influenced consumer spending could reach $750 billion by 2028, while warning that unprepared brands could see traditional search traffic fall 20 to 50 percent. Bain & Company separately estimated that a large majority of consumers now rely on summaries produced by a machine for a meaningful share of their searches. These are consultancy estimates about a market that has not finished forming, not measured outcomes, and the gap between Pew's tracked panel data and these forecasts is itself worth noticing.

What is more interesting than any traffic forecast is that the mechanism for buying a citation, as distinct from buying an ad placed near one, does not currently exist in the way the keyword auction existed. Google has said its AI Overviews draw citations from organic web content and has kept paid search ads structurally separate from the generated summary itself. A study published July 15, 2026 by the research firm SE Ranking, reported by Search Engine Land, examined Google's AI Mode ads directly against the citations appearing in the same AI Mode responses and found that only 11.53 percent of advertiser domains paying for an AI Mode ad on a given keyword were also cited as a source for that keyword, and that roughly 85 percent of advertisers paying for AI Mode ads did not appear anywhere in organic results, let alone the citations, for the keywords they were bidding on. Paying for placement next to the answer and being named inside the answer turned out to be two largely disconnected systems.

There is a real, if partial, correlation worth naming honestly alongside that disconnection. Separate tracking by the SEO platform BrightEdge found that the overlap between AI Overview citations and a business's own top-100 organic search ranking grew from roughly 32 percent to 54 percent across a sixteen-month window running from February through July 2025, meaning existing organic strength is correlated with, though far from a guarantee of, being cited. That overlap varies enormously by category, from about 1 percentage point in e-commerce to more than 53 points in education, according to the same tracking. Correlation with organic rank is not a purchase mechanism, though, because no engine has published a rate card, an auction, or even a stable rule set for what earns inclusion.

That is the honest, if unsatisfying, center of this study. A business can still buy a click at a known, bid-set price. It cannot yet buy a citation with any comparable certainty, only a set of after-the-fact correlations that researchers are still mapping. The cost of being found in the answer era is real, and Pew's own tracked data shows it is already absorbing a rising share of search behavior, but it is currently unpriced in the specific sense that matters most to a business trying to budget for it: no seller, however much it is willing to spend, can currently purchase a guaranteed place inside the answer the way it could purchase the top slot in a 2003 keyword auction. That is a difference in kind, not only in degree, from everything that came before it in this history.

Who Has Captured the Rent, Three Times

Lining up the three eras against who captured the toll makes the pattern visible. In the fixed era, the toll went to the space broker and the media owner, the newspaper Volney Palmer bought discounted space from and resold at a markup, though N.W. Ayer's 1875 open contract at least realigned the toll collector's loyalty toward the advertiser paying the bill. In the auction era, the toll increasingly went to a small number of platforms running both sides of the market at once, a marketplace and a leading seller inside it, with Google and Meta together still holding roughly half of U.S. digital ad revenue even after several years of decline, and with two separate federal courts finding that Google had crossed from winning that position fairly into defending it illegally. In the emerging answer era, the toll, if there is one yet, is held by the engine that writes the summary, and unlike the newspaper or the auction, that engine currently collects nothing in a transparent, priced way for granting a citation, because it has not built, or has not yet chosen to build, a market for one at all.

The pattern across all three transitions is not just that concentration increased, though it did: a single newspaper always competed with other newspapers for an advertiser's budget, Google and Meta compete with a shrinking but real set of alternatives, and the leading answer engines are, for now, a small handful of companies each running its own undisclosed selection process. The pattern is also that disclosure fell at every step. A rate card is public. An auction's clearing price is at least visible to the bidder who paid it. A generated answer's citation logic has not been published as a rule set at all, only inferred after the fact by researchers measuring which businesses happened to get named.

Whether the answer era eventually re-prices itself into a new auction, a subscription, or something with no precedent in this history is not answerable from the record available in 2026, and this study makes no claim that it is. What the historical arc does establish reliably is that the cost of being found has never gone to zero, and it has never stayed in the same economic form for more than a few decades. The open question this history leaves is not what the number will turn out to be. It is what a business can actually measure and change while that number is still being decided by someone else.