Interview · · 7 min read

Distribution decides the money. Whoever touches the consumer keeps it. The basics still win.

Ari Paparo, Founder and CEO of Marketecture Media and author of Yield, argues that advertising has always been a fight over distribution, and that the profits now accrue to whoever sits closest to the consumer, whether that is ChatGPT, Gemini or a future agent. He walks through the collapse in search referral traffic, why OpenAI ads already skew mid funnel, and why the real scale is far smaller than the hype suggests. He is sharp on the lost opportunity of 2018, the absence of any breakout AI-first ad tech company, and why the basics still win.

Ari Paparo, Founder & CEO, Marketecture Media. Interviewed by Justin Cooke.

Episode page
Share
Ari Paparo, Founder & CEO at Marketecture Media
Ari Paparo, Marketecture Media

Listen while you read:The AI Surface

0:0034:00

Advertising has always been a fight over distribution

The setup.

The story of advertising is always about distribution, who controls distribution to the end consumer. It goes back to newspapers, who held a monopoly on local distribution for 50 or 100 years. So the profits should accrue to whoever touches the consumer, whether that is Google and Gemini, ChatGPT, or ultimately an agent of some kind. The reason it holds is a very distributed world of millions of advertisers. Where the newspaper era had 100 advertisers, now there are a million. They are fragmented, they will spend wherever they can, and if you are next to the consumer, you are going to get it.

Search referral traffic has fallen 40 to 80 percent in two years

On the collapse.

Search traffic has declined by something like 40 to 60 to maybe even 80 percent over the last two years, depending on category. The worst affected are the ones who were just aggregating data without adding much value, and there is a cat and mouse game with AI crawlers, using tools like Cloudflare to block them or extract money. The more unique the data and the more it is a genuine utility, the safer you are. Fantasy football is a media business AI cannot replace, because you still have loyalty to the source. A review site living on affiliate fees has to think hard about traffic, outbound clicks and those fees.

On data over pages.

The biggest difference between media and tech is that media thinks in terms of pages, content and eyeballs, while tech thinks in terms of data and APIs. Agents like Muse from Meta and Dots from ChatGPT want to get data on behalf of the consumer. So publishers need to think that way. Instead of a review that only lives on an HTML page, what if those reviews, including affiliate links, could be extracted in structured form for a Claude or ChatGPT to read? Maybe the revenue model survives and you are still adding real value.

OpenAI ads are mid funnel, with 80 percent new customers

On advertising inside assistants.

ChatGPT is showing ads based on what people ask. As a first-order analogy it is search advertising in a different form, but it is richer, there is far more context in a prompt than a search term. I interviewed the CEO of Criteo, an early OpenAI partner, and he said something like 80 percent of purchases from OpenAI ads were new customers. That means it is not bottom funnel, it is mid funnel, which is a really big difference.

From Business of Marketing

Most companies try to reach senior decision makers. We give them a reason to engage.

Partner with Us →

A billion-dollar goal, a 150 million run rate

On the real scale.

On paid advertising in AI, the honest answer is we do not really know. There is essentially one participant, OpenAI, which has stated a goal of a billion dollars in revenue for the year, but recent news put the run rate closer to 150 million, so hitting a billion looks unlikely. They are speeding down the highway while changing the tires. Google has Gemini answers but is not necessarily putting ads inside them yet, and how Google Search transitions to AI is probably more important in the long run. Remember too that early adopters are the ones paying, and if you pay you do not get the ads, so OpenAI is trying to saturate the paid and free markets at once, which is trickier than Netflix's sequence.

Google built, bought and bullied, but the build did most of the work

On the book.

The build was the most significant. In 2005 and 2006 programmatic advertising was a total unknown, a remnant, last-ditch way for publishers to make money. The idea it would become the pipes for all digital advertising, from streaming TV to mobile to desktop, was unrealistic. Google did quite a job building AdX into the market leader, which did involve some bullying, but we do not talk enough about DV360, the buying platform that became effectively number one in the open web. They supported the open web with those dollars, even if they had sharp elbows and some competitors ended up with black eyes.

On the natural monopoly.

Search turned out to be a natural monopoly, where the winner is determined by scale, so there can only be one. The data flowing to Google made its search better every day until no one could catch up. The real question now is whether a consumer AI environment is unstoppable, or more like accounting software you can switch between. Chat looks like an open market, you could move between Gemini, ChatGPT, Claude and Perplexity without losing much. But agents know your bank, your email, your preferences, so there could be some really significant consumer lock-in as we move to agents.

2018 was the lost opportunity to change the web

On the broken promises.

The strongest chapter, a story that had not been told, was 2018, when Apple released Apple Reader, Facebook announced Facebook Reader and Google announced AMP within six months of each other. All three had the same vision of moving the web forward, mostly ad-supported, and publishers thought this could be their saviour, a path out of fragmentation, lack of identity and bad ads. But all three broke their promises. Apple was hostile to ads so no one made money, Meta would not build a real partnership and pulled it, Google said the right things but did not do them. That was the lost opportunity. If one had worked, it could have changed the consumer's relationship with content.

No breakout AI company in ad tech, but creative is a tidal wave

On table stakes.

Press releases saying we just added an agent to our boring old SaaS product are so boring, it is table stakes now. If you still want a marketer to log into your UI, fill out forms, use naming conventions, download a report and reformat it for PowerPoint, you lost. You need to be API and MCP enabled today. Put out a press release if you want, but nobody cares, just use the product.

On the winners.

There have not really been any breakout AI companies in ad tech, none of substantial size started in the last five years. The winners are incumbents already in the transaction, a DSP or SSP, using AI to get more efficient. The breakout names are in the adjacent creative space, ElevenLabs, Twelve Labs, Higgsfield, names I had not heard of twelve months ago and now billion-dollar valuations. Generative creative came faster than I expected and is a tidal wave, used in agencies for concepting and final work and widespread on social with no taboo.

Agencies are building because the barrier has fallen

On who builds.

I have not seen much on the brand side. When I ran a DSP, appetite for build lived in the extremely direct-response companies, and fell precipitously once an agency was involved. The interesting build-out is at agencies. Almost every agency now has a CTO with a budget actually building things, and holding companies are buying, LiveRamp, Omnicom buying Flywheel. The barrier to running your own tech has gone down. Agencies stayed away from DSPs and SSPs because they were high-tech and low margin, but creator workflow and monitoring is doable and worth owning.

On what marketers should own.

Whether the ads work is so fundamental it should not change when you change your agency. So invest in attribution, analytics and incrementality, and potentially media mix modelling. In the old world MMM was complicated models on mainframes run once a year. Now Facebook and Google have open-source versions, data flows between platforms, and you can run it weekly. Those are a brand's first investments. Workflow is much less important. And keep focused on the basics, a strong informative website and your reviews on retailers matter enormously as sources of AI data.

The AI is your intern, so learn to automate your own job

On the next generation.

This is the crux of the AI question, whether it eliminates or creates jobs, sometimes called Jevons paradox. For young people, they get the negative side first because you do not need interns, the AI is your intern. But if you come in AI-pilled and know how to automate your own job, you have a big leverage effect, far more than the boomer in the corner office printing his emails. It has never been easier to start an ad agency, one person can have more impact than a team of ten a couple of years ago.

On surviving as a company.

My number one advice is don't go to jail. Number two is don't run out of money, and those can conflict. Honesty and integrity come first. Then understand where you are in the fundraising cycle, there is real negative sentiment against ad businesses in venture capital. Even when Beeswax was almost doubling, I could not do a B round, everyone said it was too ad techy. You cannot assume the next round is there. Companies get into trouble on the sales side, where hiring runs well ahead of revenue and is not managed as tightly, because the cycles are long and you are chasing whales.

Hear the full conversation. Follow Business of Marketing.

10 · The board question

If AI assistants become the surface where consumers get their answers, what is our concrete plan to keep touching the customer directly rather than renting that access from whoever owns the surface?

The Rewrite

Stay close to the conversation.

The conversations and intelligence worth knowing about.

About the guest

Ari Paparo

Founder & CEO, Marketecture Media · The AI Surface

Ari Paparo is Founder and CEO of Marketecture Media, the podcast and newsletter he built to help buyers cut through vendor spin in advertising technology. A long-time operator in the plumbing of the ad market, he is also the author of Yield, an account of how Google came to dominate advertising, and he sits on boards as an entrepreneur and investor.

Episode page Full profile LinkedIn (opens in a new tab)

More interviews

All interviews
Share