When Buying Becomes a Conversation, What Happens to the Advertising Model?

Over an extended conversation, an AI assistant can learn far more about what a buyer wants than any single search query reveals, and still be harder to make money from. Those two facts now sit side by side in Google’s own results.

On Alphabet’s third-quarter 2026 earnings call, Sundar Pichai said AI was driving “overall query growth, including commercial queries,” and that query growth was the highest in five years. On the same call, chief business officer Philipp Schindler said Search revenue growth had slowed for several reasons, currency the largest. Another was that “the strong growth in AI Mode queries had an impact because of the lower monetization we were seeing in AI Mode.”

More intent, less revenue per question. It stops looking like a paradox once you separate three things search used to deliver as one.

Understanding, Influencing, Monetizing

Understanding intent means knowing what someone is trying to do. Influencing the decision means shaping which options they consider and which they choose. Monetizing intent means getting paid by a business for access to that person at that moment.

In traditional search, all three happened in one place. The query revealed intent, the results page shaped what was considered, and the ads on that page were where payment could happen. Understanding, influence and revenue sat on the same screen.

Conversation separates them. An AI system can understand far more over a dozen exchanges than any single query reveals, and its answer can shape the decision more directly than a list of links. Neither automatically creates a moment someone will pay for.

How a Search Became an Auction

A query is a discrete, declared signal of intent. For each one, the system decides whether advertising is relevant, and many show no ads. When ads are eligible, advertisers compete in an auction for a few positions, and the winners earn an impression. Many advertisers pay when someone clicks, and that interaction can lead to a website visit or another measurable action, with subsequent conversions attributed to the advertisement.

So the query and the paid event are different things: the query is the opportunity the system evaluates; the click, or sometimes the impression, is what gets priced. Twenty searches give the system twenty chances to find a commercial moment, and some become paid clicks. Advertisers keep paying because they can see what they bought: the query, the click, and often the sale.

One Trip, Many Questions

Imagine a couple planning a guided walking holiday in the Dolomites next September. In a search engine, the decision might unfold over weeks of separate queries: when to go, guided or self-guided, which operators run small groups. Each is a fresh chance for an ad.

In a conversation, it can become one thread. They describe their dates, budget and dislike of large groups. The assistant rules out self-guided options and narrows the field to three operators. Then they ask:

Which of these three has the most flexible cancellation terms, and which would you choose for a couple who want a quieter route?

The system now understands their intent far better than any single query could show. But most exchanges that got them here weren’t natural places for an ad, and by the time the question became commercial, the shortlist already existed.

Where an Ad Fits in a Conversation

An ad inside a conversation can be useful. If the couple ask for operators with availability on their dates and a relevant one appears, clearly labelled, it adds an option that fits.

It becomes disruptive when it interrupts the reasoning, like a large-group tour advertised after the couple has ruled such tours out.

It undermines trust when the user can’t tell whether the answer itself is for sale, which is why platforms draw a hard line. OpenAI says “Ads do not influence the answers ChatGPT gives you”, and that ads are labelled as sponsored and kept visually separate from the answer. Google labels its AI Mode ads “Sponsored.”

That line protects the answer, not the buyer from persuasion. An ad is meant to influence the person who sees it, and a labelled ad that persuades the couple to consider a fourth operator is working as intended. Separation doesn’t make advertising ineffective; it changes what the ad competes with. On a results page, ads sat among options the buyer still had to evaluate; in a conversation, the ad sits next to a recommendation the buyer trusts. The most influential part of the interaction, the answer, is the part no one can buy.

What Google Has Disclosed

Google’s 28 October announcement expands ads in AI Mode to ten more markets. In the US, it is launching multiturn ads “to allow ads in follow-up queries in AI Mode.” Google also reports, from its own research, that AI Mode queries are on average three times longer than traditional searches, and, from its own internal US data over two weeks, that “click-through rates are already similar between sponsored and organic product list items within AI Mode.”

Each finding has a specific scope. Three times longer measures query length, not how many conventional searches a conversation replaces. Similar click-through rates show people engage with sponsored product listings about as readily as organic ones. They don’t show how many ad opportunities a conversation produces, how many are filled or what each is worth, and revenue depends on all of those.

On the earnings call, with Search and other revenue up 12% to $63.3 billion, Schindler said “the majority of AI Mode queries are incremental, creating more monetization opportunities over time,” and that many are “upper-funnel queries.” He also said “we feel very good about our trajectory here.”

Whether AI Mode will eventually earn as much per question is open. Google hasn’t said how large the gap is or broken it down, but the structural pressures are clear. A conversation gathers intent into fewer, richer exchanges, much of its value lies in an answer advertising can’t enter, and the click search relied on happens later or not at all.

From Selling Queries to Selling Decisions

Search spread a decision across many small questions, each a separate opportunity to show an ad. Conversation can consolidate the decision into fewer, richer interactions. The platform may understand more about the buyer while having fewer natural opportunities to monetize that understanding through conventional advertising.

This is the difference between value created and value captured. The couple saves time and avoids a poor fit. The operator they choose gets a better-qualified customer. The platform earns trust and return visits. None of that is revenue unless there is an event someone will pay for.

The other side of the ledger is cost. Every conversational answer is generated, turn by turn, and that carries inference and infrastructure costs. Monetization has to be judged against them: the question isn’t only whether a conversation earns as much as the searches it absorbs, but whether it earns enough to cover what it costs to run.

Attribution makes pricing harder. In search, a click tied the ad to the visit and often the sale. In a conversation, the evaluation happens before any click, and no one can easily say which exchange earned the sale. Advertisers pay most readily for what they can measure, and the most valuable work in a conversation is the hardest to measure.

So the unit of monetization is under pressure to move from the query to the journey: far more valuable, and much harder to price, auction and attribute with the tools search built.

Six Ways to Pay for a Conversation

Model	When it pays	Attribution depends on	Trust depends on	Willingness to pay depends on
Contextual ads	Alongside a relevant exchange	Linking impressions or clicks to later sales	Relevance, labelling, timing	Reach and measurable response
Decision-stage ads	When the buyer is choosing	Proximity to the purchase	Clear separation from the recommendation	Likelihood of a sale, and whether it’s incremental
Qualified leads	When a buyer asks to be connected	Clear, if the hand-off is recorded	Who gets referred, and why	Lead quality and the value of a customer
Transaction compensation	When a purchase completes	Clear, if the platform sees the sale	Fees not affecting the recommendation	Margin, and whether the purchase can happen in the conversation
Subscriptions	Continuously, from the user	Not needed	The user being the customer	Not advertiser revenue
Hybrid	Several of the above	The mix	Rules keeping each stream out of the answer	The mix

Subscriptions aren’t advertising, but they pay for the conversation from the user’s side and reduce how much advertising has to carry.

Search sold many opportunities across a journey, most early and relatively cheap. A conversation reaches a late, qualified moment: three operators, fixed dates, a buyer ready to book. That buyer is worth more to whoever wins the booking. It doesn’t follow that an ad shown to them is worth more.

Four things are easy to run together: what the booking is worth to an operator; how much the ad changes the chance of winning it; what the operator can bid, given that increment, its margins and its budget; and the price the platform obtains, which depends on competition for the placement. A buyer close to purchase raises the first. It doesn’t necessarily raise the second, and what advertisers can bid rests on the second.

The three kinds of advertiser show why:

• Seeking initial consideration. An operator appearing while the couple is still deciding between guided and self-guided. Further from the sale and cheaper, but the ad can still shape what gets considered.
• Trying to enter the shortlist. A fourth operator, after the assistant has named three. The ad could change the outcome, but it must overcome a recommendation the buyer has just accepted, so its chance of converting is lower.
• Converting a buyer already considering it. One of the three. The buyer is most likely to book with it, so the ad adds least. It may pay to protect the sale, much as brand bidding works in search, but has limited reason to pay much for a booking it might win anyway.

The opportunity with the greatest likelihood of a purchase isn’t necessarily the one with the greatest ability to change the purchase. For one late opportunity to replace many early ones, the platform has to match qualified moments with advertisers who can still change the outcome, can afford to pay for it and face enough competition to sustain the price.

The Value of a Qualified Buyer Is Not the Price of an Ad

• Potential customer value
• Illustrative achievable ad price

Relative commercial value / advertising price

1. Exploration
2. Evaluation
3. Decision

Progress through the buying conversation

A shortlist doesn’t necessarily mean fewer bidders. What changes is the order of events: in every case, the platform’s answer shaped the shortlist before any advertiser bid.

What the Publishers Are Seeing

Publishers feel the shift first, though not equally. On Reddit’s third-quarter call, Steve Huffman said search referral traffic was “down meaningfully” in the quarter and that during October it had “dropped by more than 1 million referrals on average per week,” with logged-out web traffic hit hardest.

Asked about the cause, Huffman described Reddit’s hypothesis as “twofold.” The biggest factor, he said, was AI Overviews and AI Mode: “at the minimum it pushes everyone down the search page,” and “the experiences themselves are not traffic drivers.” The result, in his words: “Google has really shifted in their relationship with the rest of the Internet.” The second, he said, was that Reddit’s machine-translated content appears to be ranked lower outside the US.

Referral pressure and business performance are different measures: Reddit’s revenue grew 53% in the quarter, and logged-in and app users account for over 80% of its monetization. Its response is to become a direct destination, testing logged-in Reddit outside the US, which Huffman said improves ad targeting and protects access to “Reddit’s extremely valuable corpus of data,” which it licenses. As a site Google surfaced heavily for years, Reddit doesn’t establish what is happening to other publishers.

When Elimination Happens Before the Click

For businesses competing to be considered, the familiar path was search, click, visit, lead, transaction, and each step was visible. In a conversation, a business can be evaluated and eliminated without anyone visiting its website. The Dolomites operator that lost on cancellation terms never saw the buyer. Being cited doesn’t establish influence either; a business can be mentioned and then ruled out in the next exchange. Visibility and outcome were always different things. In conversation they move further apart.

The platform being paid to monetize the interaction is, increasingly, the intermediary that decides which businesses are considered.

The Case That It Works Out

There’s a serious argument that this is a transition, not a structural loss. Search ads were once unproven too. Google says most AI Mode queries are incremental and many come early in the buying process, which could mean more commercial moments, not fewer. Sponsored product listings already earn click-through rates similar to organic ones. And qualified opportunities have long commanded high prices in categories such as insurance and legal services.

For that to hold, several things would need to be true:

• Platforms find reliable decision moments across a conversation.
• Advertisers gain measurement that doesn’t depend on clicks.
• Users accept clearly separated ads at the point of decision.
• Revenue per opportunity rises enough to cover fewer opportunities and higher serving costs.
• Or subscriptions and transaction revenue fill the gap.

None is implausible, and Google has been changing Search from the inside for some time. But each requires building something search never needed, which is why the outcome is open even though the mechanism isn’t.

A Valuable Conversation Isn’t Automatically a Profitable One

A platform can understand a buyer better, influence the decision more and still earn less from the interaction. The three no longer coincide. Understanding builds up across the conversation. Influence sits in the answer, which has to stay independent to be trusted. Monetization has to happen beside that answer, at moments the platform can identify and an advertiser can measure.

That leaves the advertiser in a different position. In search, an advertiser bought a place on the page where the buyer’s options were assembled. In a conversation, it buys access to a buyer whose options the platform’s answer has already shaped, and it has to persuade alongside a recommendation it can’t pay to change.

The value of a commercial conversation is not the advertising revenue it carries. And the platform trying to capture that revenue is increasingly the one deciding which businesses the buyer considers before any advertiser gets a chance.