I finally finished my strategy write-up coming out of Cannes, an exercise in picking at a million threads. I find myself coming back to one of them repeatedly though, so I thought I'd share it here and see if anyone had answers.
While the Palais was packed and everyone all smiles, I left feeling the advertising space was in an existential crisis, bubbling just under the surface. Americans are reaching all time high's of daily media consumption, which has plateaued at ~11 hours per day, with over 8 hours now dedicated to digital media including streaming, social networks, and gaming, signaling a mature attention economy undergoing massive internal reallocation.
If you are a nerd, and I must here admit to being one, you know what the terms WWDC and I/O are. These are Apple and Google’s respective developer conferences where they announce what everyone needs to pay attention to, and Google’s took place in May just about two months ago. At that event, Google announced a long feared change to how it’s search works- they’re now fully defaulting to their Gemini AI results, and doing even more to obfuscate the traditional search page, the famed ten little blue links. Google promises this will be good for publishers, as it will deliver less but higher qualified traffic to their websites (because it’ll be taking information from their sites and publishing it in it’s AI). Publishers appear skeptical, many in the industry have feared this for years, commonly referring to it as “Google Zero” because traffic from Google, which can account for as much as half a sites visitors, is anticipated to drop to zero. Picture losing half our client base and you’ll basically be in a website publishers shoes.
Regardless of how you feel about AI answers compared to search, the art of appearing in those ten links, SEO (search engine optimization) is an enormous element of advertising today, perhaps 25% of overall marketing spend. It’s an extremely proven tactic and Google’s more or less declared it’s extinction in the very near term. In it’s place the industry is frantically searching for it’s replacement tactic and abbreviation. The strongest candidate at the moment is figuring out how to bias the LLMs the same way our business has been able to bias search results in favor of a given client, the name for this new skill is contested but the likeliest candidate appears to be GEO (Generative Engine Optimization). There was much discussion at Cannes of how marketers of all stripes have rapidly adopted vendor tech to help understand how and when their products appear in generative AI searches. This led to some hopeful statements from the ad men- since LLMs index on high quality written material so much, there continues to be demand for the creative class, at least they tell themselves. Certainly there’s evidence it can be done well and easily: interestingly, because there’s so much junk on the internet the LLM makers tend to pick a few sources to be biased towards, Reddit is one of the primary ones as humans continue to use and comment on it. In the study linked above, researchers found that putting a 13 word snippet about a nonexistent restaurant on reddit led to that restaurant being included in LLM recommendations. The fact it can be done is not evidence for a consistent ability to deliver or measure ROI of GEO though, and this is a huge problem.
The reason SEO accounts for a quarter of all ad dollars is simple, SEO is immediately measurable. Google’s implementation of a cost per click model, where everything had previously been on the number of ads served, immediately and permanently altered how advertising performance is assessed. It allowed marketers to tell their bosses they only paid for what worked, and combining click data with the marketing brand site analytics, which tell you what percentage of people convert from visiting the site to buying the product, and the average value of a transaction, teams were able to calculate and present defensible return on ad spend amounts to their executive teams. For the overwhelming majority of marketers, SEO is much like buying leads, it’s the safest, most defensible line item in a budget. This is what is commonly termed a direct response model, or outcome based pricing. Judging from Cannes, "outcomes" is the word of the year, with executives demanding quantification of every form of spend at the modern enterprise.
In consequence, every other part of the advertising ecology is trying to implement similar pricing structures, so they can become the safe option in a marketers budget. The shift from linear TV to CTV is in part driven by this- when media becomes trackable, as connected television is, it creates the possibility of connecting the TV ads to outcomes data, in a clean room environment. Where the linear TV ad buying business was premised on something like “I can put your product in front of 20 million people”, CTV offers the possibility of knowing how many people who see a snickers ad buy a snickers, and refining the CTV buying strategy to make that even more efficient. In more quantified terms, premium CTV environments command the highest CPMs, often ranging from $30.00 to $80.00 depending on live sports or major event associations, followed by linear TV at $20.00 to $40.00 and programmatic open-market video at $12.00 to $30.00. In spite of this, there's still more room for a swing in valuations – : traditional linear TV generates $0.38 per user-hour versus just $0.07 for independent mobile apps—a 5.43x disparity that signals major budget reallocation opportunities as programmatic systems improve. Maybe that played a role in the Comcast decision to split media and tech?
The changes in TV and search are both in response to broader trends in consumer behavior, which neither Google or the TV companies can dictate. Very simply, CTV is taking over linear because more people watch TV on connected devices than on broadcast. Google is shifting to AI answers instead of search because it’s afraid ChatGPT and Claude will become the default way people look for answers instead of Google. This makes Google very sad, because it makes money on being the default option people come to for answers (for instance it pays Apple 20 billion or so a year to be the default choice on Mac devices).
What’s funny about Google blowing up the search model it collects a fortune on is it’s not at all clear that the chat function is the end state that could replace search, at least in the opinion of the AI companies that have pioneered the presently en vogue interface. At least at ChatGPT, they’re declaring “chat is dead” as of last month, in favor of a super-app, conceptually similar to China’s Wechat, which combines social, shopping, gaming, mobile payments, streaming video etc. There’s been a broader debate about the mode by which humans engage with technology for years- most notably in the VR vs AR space, and concepts of “ambient compute” pushed most extensively by Microsoft. “Watching tv” now happens almost as often on phones as it does on televisions. Similarly we’re wading into a space where “search” becomes as amorphously defined as asking a question of a technology, since virtually every piece of software now has a way to receive questions from you. No one knows what that’ll be- most of Facebook’s value comes from being the place people spend most of their time on the internet, but those guys haven’t got a clue why people spend time there- half their users are on Instagram, and at this point they’ve turned instagram into a tiktok clone. I don’t have numbers but I’d bet good money the vast majority of content seen on “social media” now is not from people the users socialize with. No, what’s clear is attention, whether where people search or how they spend their leisure time, is up for grabs, and most of the dollars in advertising ride on where that happens.
How and where will people search for things is an essential question because not every mechanism for asking a question can be equally well monetized. The phone has been a great middle ground between computers, where information is often displayed in text and can have ads wrapped around it without creating an overly intrusive experience, and televisions, where the ad overtakes the content entirely for it’s duration. In both cases however there’s a structure for advertisers to pay to make a user aware of their product. The new modes of content engagement and consumption don’t show the same possibilities of advertising, at least not yet. Amazon and Google both made bets on smart homes- extensions of the ambient compute philosophy, where you can just shout at the box on your counter and it’ll answer. There’s limited advertising possibilities to that hardware, aside of course from data collection to enhance the performance of the older, reliable media channels. Facebook in contrast is pursuing the augmented reality path, with it’s Orion smart glasses product. For about 800 bucks you can have a little computer screen in your glasses. This is not a form factor ready to accommodate a display banner ad, at least not yet.
And this is where I smell the uncertainty- where will the audience be next year? Where they go, the media dollars go, and even the masters of silicon valley seem entirely uncertain. If you can't put ads in it, you're gonna charge a subscription fee for it. By all accounts Meta may introduce a subscription fee for it's glasses- that'll be a real test of what hardware people are willing to subscribe for. Just last week, Anthropic walked back it's decision to limit access to it's best model, Fable, via consumption pricing, because market signals and OpenAI competitive pressures wouldn't allow it. In the last twenty years, I don't think it's ever been less clear how people will use their free time, and how advertisers will reach them during it.
So that's one of the big questions tapping me on the shoulder and demanding answering. Coming out of Cannes, I do know whatever it is, marketers will demand tangible proof it drives outcomes, so I'm very glad of the work we're doing with Decentriq and our cloud based clean room partners like Snowflake. If you think you know the answer, please tell me and let's make some money on it.
Originally published on LinkedIn.