What’s an ad impression worth in an AI conversation?

There was a time, within the past few years in fact, when a decent search query looked like this:

Laptops under 60,000, good memory and battery, exchange offer.

Today, more and more search queries are starting to look like this:

Hey, I’m looking for a new laptop. My budget is up to a lakh, and I need some decent to heavy performance because I need it for both gaming and light work, because most of the major stuff is done on the office one. Also, a decent battery because I travel a lot. So, good anti-viruses and security also. Also, I have my old laptop I want to exchange, everything is fine, it just crashed when I clicked…

In that earlier time (of three years ago), search queries were globally handled largely by one player, and it wasn’t Ask Jeeves. Since then, the latter query is one that could have been typed across a variety of platforms and surfaces, and met with a wide variety of outcomes.

In that earlier time of fewer words and fewer platforms, intent was easier to package. An advertiser could bid against a keyword, measure an impression, a click-through rate and cost-per-click, follow that click towards a conversion, and calculate return on ad spend. Imperfect signals, certainly, but familiar ones, refined over decades of search advertising.

The latter query contains far more information. There is a budget, use case, purchase intent, lifestyle, product requirements and even an existing device available for exchange. But which of those signals should determine the ad? Is the user shopping for a laptop, comparing one, researching one, or merely asking an assistant for advice? And if an ad appears halfway through that conversation, what exactly has the advertiser bought: an impression, a slice of attention, a signal of intent, or proximity to a purchase?

A keyword has boundaries. A conversation has a trajectory.

And that trajectory is beginning to attract serious money.

Advertisers spent an estimated $244.9 billion on search advertising globally in 2025 and another $413 billion on social media, according to WPP Media. In India alone, Dentsu estimated paid search at ₹16,581 crore and social media at ₹21,057 crore last year.

These are mature markets built around relatively familiar currencies. Search monetises declared intent. Social monetises audiences, affinities and behaviour.

Conversational AI may be trying to monetise something rather messier.

Nikhil Kumar, media and tech leader and former CGMO of Affle, is not convinced that simply placing an ad inside a high-intent conversation should make it premium inventory.

“I wouldn’t automatically put a premium on an impression simply because it sits inside a high-intent conversation. The premium should come from the quality of the signal and its ability to drive a better outcome,” he said.

A conversation may reveal far more about a user than a keyword or social interest, Kumar noted, but advertisers still need to understand how that intent is derived, how time-sensitive it is and whether it actually produces better outcomes. “Ultimately, the value may sit less in the impression and more in the intelligence behind the impression.”

Subhash Pais, Founder and CEO of SutrAi, takes that argument a little further, musing perhaps the problem is not that conversational impressions need a new price. Perhaps advertisers are trying to price the wrong thing.

A search for “best running shoes” tells an advertiser something. A conversation in which the same user says they have ₹10,000, run three times a week, have flat feet and need shoes for a half-marathon in six weeks tells the platform far more: budget, use case, constraint, timing and potentially purchase readiness.

“That is a fundamentally different advertising environment,” Pais said.

The distinction matters because conversational AI is hardly a niche interface anymore. Sensor Tower estimated ChatGPT crossed one billion monthly active app users in May 2026. On August 11, Google said the Gemini app had also surpassed one billion monthly users. Similarweb, meanwhile, found generative-AI platforms averaged 9.5 billion monthly visits between June 2025 and May 2026.

The audience has arrived before advertisers have agreed on what that audience is worth. Pais argues that the answer may lie in what he calls the “decision moment”.

“Someone can have a long conversation about buying a car without ever intending to purchase one. Conversely, a very short conversation could represent an extremely valuable decision moment,” he said.

More words, in other words, do not necessarily mean more intent. “The premium, therefore, should be earned by decision quality, not by the novelty of the medium,” Pais said.

His formulation is neat: “Search monetises the query, social monetises the audience, but conversational AI has the potential to monetise the decision moment.”

Ad sales maven and APAC veteran Taranjeet Singh broadly agrees, but narrows the commercial test further.

“In search, you know what someone is searching for. In a conversation, you potentially have a much richer understanding of what they are trying to solve, why they are considering something and where they are in the decision journey,” he said.

That does not necessarily make every conversational impression more valuable. “I would pay a premium for better intent and proximity to purchase, not just because the inventory happens to be inside an AI conversation,” noted Singh.

That “proximity to purchase” is important because the incumbent advertising machines have spent years translating their particular signals into enormous businesses.

Google Search and other advertising revenue reached $63.3 billion in Q2 2026. Meta advertising revenue reached $59.4 billion over the same quarter. For the full year, EMARKETER forecasts Meta’s global net advertising revenue at $243.46 billion, slightly ahead of Google’s $239.54 billion in digital advertising revenue.

Conversational AI therefore does not merely have to invent an advertising product. It has to persuade advertisers that the additional information embedded inside a conversation improves on systems into which hundreds of billions of dollars already flow.

Vipin Yadav, Vice President and Head of Marketing at DriveX, frames the distinction as one between proximity to a consumer and proximity to a decision.

“In search, consumers express what they want; in conversational AI, the platform can potentially understand why they want it, their constraints and the trade-offs they are considering,” he said, adding, “The real currency could shift from impressions to ‘decision value’.”

Which sounds compelling until Ambika Sharma, Founder and Chief Strategist at Pulp Strategy, points out that the advertiser may not even be delivering the most influential commercial message in the conversation.

The assistant is.

“The ad runs below the answer. It cannot touch the answer. And the buyer trusts the answer, not the sponsored line under it,” Sharma said.

Consider the problem.

A consumer asks an AI assistant which laptop to buy. The conversation reveals budget, requirements, travel habits, performance needs and a device available for exchange. By almost any definition, this looks like extremely valuable commercial intent.

Then the assistant recommends a rival.

“So, an impression next to a reply that names your rival is high intent spent against you,” Sharma said. “The number to price is not the intent. It is what the assistant recommends in that chat, and whether you are in it.”

That complicates the easy assumption that richer intent must mean more valuable inventory.

Prashant Puri, Co-Founder and CEO of AdLift, lands somewhere between the optimists and sceptics. A high-intent AI conversation can command a premium, he said, but only where platforms demonstrate stronger intent and measurable business value than search or social. “Ultimately, AI inventory should be valued not just on impressions or clicks, but on its ability to influence consideration and conversion.”

Which brings the industry back to a strangely basic question: what is the unit being bought?

OpenAI’s emerging advertising stack looks familiar enough on the surface: CPM and CPC buying, campaign budgets, ad groups, auctions and contextual signals. But the thing surrounding the advertisement is fundamentally different from the page, keyword or feed that preceded it.

A conversation can contain stated needs, unstated constraints, objections, timing, budgets, alternatives and changing levels of purchase intent. It can move towards a decision. It can move away from one. It can recommend a competitor.

Whether advertisers can actually verify that “decision value”, rather than simply taking the platform’s word for it, is another problem altogether. Come back for Part Two tomorrow.

Which brings us back to that laptop. Everything was fine. It just crashed when the user clicked…

Whether the next word is “update”, “advertisement”, “attachment” or “a link in an email from my bank” may determine whether the AI is looking at product information, troubleshooting, cybersecurity or a context in which the ad perhaps should not appear at all.

That is the problem with pricing a conversation.

Search spent two decades teaching advertisers what a keyword was worth. Social taught them how to price an audience. Conversational AI may now be asking them to put a value on a decision while it is still being made.