Happy Sunday, Everyone!
I hope you’re all doing well and enjoying the (semi-official) start of summer, also known as Memorial Day. After a few heavy issues on marketing strategy, I thought we’d change course with this week’s issue, and focus on current events + big-picture ad tech trends.
It also just so happens that Google held Marketing Live (part of I/O) this past week. In a surprise to exactly no-one, this was less a product showcase and more an AI infomercial. For two-hours this past Wednesday, advertisers were treated to a firehose of AI-powered everything, from agentic search to predictive bidding to full creative automation.
Depending on your feelings about GenAI, automation and the future of marketing, it was either thrilling or terrifying (or, just maybe, a bit of both).
What’s clear: we’ve hit an inflection point. Much of the last 24 months has centered on experimenting with AI in marketing; the next 24 will be about deploying it and trusting it with progressively more mission-critical functions. But, that trust will not - and can not - come cheap. While it’s tempting to think we’re in the late game of AI, the reality is we’re barely past the opening.
After listening to GML, I’m more convinced than ever before that the most successful marketers will be the ones who know the difference between automation that drives results and hype that drives headlines.
Let’s unpack what actually matters, where the opportunity lies, and what to keep a critical eye on.
AI Mode + AI Overviews Ads Are Coming To SERPs Near You
In a surprise to precisely no one, AI Overviews and the recently-announced AI Mode took center stage for a significant portion of the festivities. Anyone who has followed the AI Overviews trend was likely aware that AIOs have started appearing on progressively more SERPs. This has come with the (fortunate, unfortunate, you decide) side effect of reducing the number of top-of-page ads (usually from 4 to 2, sometimes from 4 to 1). To be blunt, this was never a sustainable solution for Google – reducing the number of ads on high-value SERPs by 50%-75% simply was not viable.
The solution to this particular problem was obvious: insert ad units into AI Overviews – which is exactly what Google announced. The goal of this is to “...transform non-commercial intent into ads that don’t interrupt, but help your customers discover you….” – which is marketing speak for: we’re going to put ads into AIOs, but we’re going to try to make them feel natural and relevant. The same thing will roll out in AI Mode, where relevant, existing ad units (search/shopping) will be inserted into generated results.
Honestly, as an advertiser, this is absolutely wonderful news – the utility of AIOs + AI Mode is incredible. AIOs are already the most-used AI technology in the world, by (at least) an order of magnitude (more likely ~2.5), and opening up more placements is a win for every advertiser (more placements = more supply = lower prices). If these units are paired with AI-enhanced experiences (i.e. more in-SERP lead gen or checkout), all the better.
It’s not all sunshine and rainbows - I have some real concerns about Google’s level of transparency around the precise prompts that trigger ads, as well as how these ads will be tracked/measured – but those are solvable problems.
Why Does This Matter?
This is a real-time reinvention of Google’s core business for an increasingly AI-dominated space. If Google pulls it off (and I think they will), the value of Google Ads is going to increase exponentially. If you’re not already checking your SERPs (particularly for your core/high-value keywords) and noting whether or not AIOs are there, you absolutely should be doing that now. Create a Google Sheet, drop in the KWs, and check each essential SERP. Not only is this helpful for understanding where these ads will appear, but it also tells you which SERPs still have outsized opportunities for traditional search ads (at least in the short run).
AI Max + PMax: Welcome to the Age of “Maxing”
GML was all about the “maxing” – with AI Max for Search Campaigns taking center stage early on. If you’re unfamiliar, AI Max is (essentially) a suite of targeting + creative tools intended to help advertisers better address the myriad of ways that people search for stuff/services/information. At the core, AI Max is a repackaged, maybe-slightly-upgraded Performance Max automation engine, tailored exclusively for keyword-based campaigns.
Activation is pretty simple: turn on one toggle (and you just know that’s going to be an auto-applied recommendation), and you hand over the account structure, keywords, intent signal and the creative to the AI Max machine. There are - allegedly - limits: AI Max will use content from your existing ads, landing pages and assets, and will “learn” from your existing keywords, creatives and active (read: non-excluded) URLs, then use broad match and keywordless matching to find higher performing queries.
Paired with enhancements to Performance Max (now with channel-level reporting and GenAI-generated creative) and the re-worked Demand Gen campaigns (collectively, the “Power Pack”), it’s clear Google wants advertisers to sacrifice control on the altar of scale.
There’s no better evidence of this than Google’s “Smart Bidding Exploration” – a novel(ish) feature which allows advertisers to set “exploration” thresholds for tROAS bidding. In more common terms: it allows you to tell Google how far it can deviate from your set tROAS if a conversion is likely. Used well, this could be a wonderful tool for keyword discovery; used incorrectly, and you’ll quickly find yourself upside down on a campaign.
Why Should You Care?
You're going to spend less time configuring and more time evaluating. Marketers now have to become curators and operators, not button-mashers. From a 30,000’ view, this is a massive positive for marketers. The notion that people can out-trade machines was always a delusion.
That’s not to say this will be an easy or painless transition – quite the contrary. I expect this will cause some significant short-term frustrations, but I genuinely believe it is a net-positive for the industry over the long-term.
Agentic Search & Shopping: The Utility of SERPs Is Increasing
One of the more interesting demos during GML was “agentic search” - a preview of a world where the search engine doesn’t simply return results; they take action. After spending the last ~month with AI Mode, it’s abundantly clear to me that this is the future of search.
Imagine booking a summer vacation. As part of the AI Mode content, there’s a section on the weather at your destination, along with trendy fashions and a small carousel highlighting summer dresses that fit the bill. You browse through the carousel (for what it’s worth, these were impressive demos - allowing a user to see how s/he looked in each option shown while scrolling the carousel), click on one for more information, but get pulled away for something else. A few days later, Google observes that the price on that dress has dropped. It sends an alert, and opens a one-click purchase experience. Discovery, retention, alert, transaction – all in a single interface.
While the primary focus of the demo was eCommerce + retail, the implications of this for B2B and lead generation are (potentially) orders of magnitude greater. The ability to leverage relevant current information to re-surface prior explorations in a context-rich environment is disproportionately valuable for longer-cycle decisions (i.e., home services, legal/law, accounting, home purchase, investment, banking, etc).
There Will Be Pros & Cons
These are high-intent placements in context-rich environments. The conversion potential is massive - but so is the risk of cannibalizing your existing campaigns if Google arbitrarily reroutes traffic.
Fundamentally, agentic ad units come with far more utility than standard placements – there’s a material difference between (for example) seeing a flat-lay shirt on a white background in a shopping carousel, and seeing that same shirt rendered onto your body, along with a new belt, pants and shoes - all with a “get this look delivered before your meeting tomorrow” CTA.
Retailers will pay far more for the second one, the conversion rates will be higher, and this (potentially) opens up an entirely new host of challenges/problems for marketers.
GenAI Gets Creative
This was one of the more interesting announcements from GML - but not for the reasons most think. The big “product” announcement was relatively “meh”: Google Asset Studio now uses GenAI to help smaller teams generate visuals. While the tech was pretty good (the Veo 3 video ad demo was underwhelming; the Creator Hub update was useful, but not earth-shattering), the thing that caught my attention was the application and the implication of the technology:
The Application: in keeping with the trend of the day, it’s clear Google is pulling out all the stops to demonstrate the value and potential of YouTube. I’m sure their teams have heard - loud and clear - from advertisers that creative is the barrier to greater scale, and this product update is aimed squarely at addressing that pain point. It also just so happens that two of the three “power pack” campaign types (PMAX and Demand Gen) are video-heavy - so this might be a not-so-subtle way of Google trying to push more advertisers into video.
The Implication: where this gets interesting is the solution (namely, using GenAI to produce entire creatives) is/was something Google had previously seemed resistant to allowing. It’s pretty clear Google didn’t want YouTube to be chock-full of AI-created content, but this announcement might be a sign that Google is changing its tune, provided that comes with a commensurate increase in YouTube ad revenue.
Brand Partnerships: the final piece of the puzzle, as it pertains to YouTube specifically, was the “Creator Partnerships” rollout. To be quite candid, I love this. It’s a simple, intuitive way for brands to discover, vet and manage new creator partners, then integrate their content into your ad account. If you’re in the creator two-sided-marketplace business, this is likely a five-alarm-fire. I expect Meta will follow suit with a similar product, and all those independent brand/creator matchmaker sites will understand what Orbitz and Travelocity went through when Google Flights launched. The big winners? Platforms, Brands + Creators. The big losers? All those middlemen who just got cut out.
What’s real: This is a productivity win for small-to-mid-size teams lacking internal design capacity. I would not expect Emmy-worthy content from GenAI any time soon, but it is worth trying for concepting and iterative testing.
SynthID: The Most Important (and Under-discussed) Thing At GML
I’m going to go out on a limb here, but I think SynthID might be the single-most-important bit of technology Google announced at GML. This was not a sexy announcement, or one designed to make headlines - but the implications of it are massive.
Let’s start with the obvious: the internet is now a zero-trust environment for all media. Every image, video, and audio clip is guilty until proven innocent. SynthID isn't a tool to establish innocence; it's a tool for one company (Google) to establish a chain of custody for its own creations while setting a standard for the rest of the ecosystem that it can profit from later.
You’re probably wondering: how does this thing actually work?
SynthID is - at its core - a digital watermark of sorts. Unlike a traditional watermark, which is layered on top of an image - and can be cropped, blurred or removed, SynthID is different. It’s woven into the very fabric of an AI-generated image at the moment of its birth.
The best analogy for this is digital DNA. When you ask a model (yours, Google’s, etc.) to generate an asset (image, video, text, sound, etc.), the AI is doing two things simultaneously: it's creating the asset you requested, but it’s doing so in a way that slightly modifies the output during generation. It's not a layer on top; it's a fundamental property of the image's structure.
The Weave (Embedding): Any organization using SynthID can customize it (essentially, create your own “watermark”) by fine-tuning your configuration settings. Every asset generated by that model will then contain your watermark. These changes are imperceptible to the human eye but create a specific, resilient signature. Because the watermark is embedded at this deep level, it's designed to survive common modifications like compression (saving it as a JPEG), color changes, and even minor cropping or text editing.
The Scan (Detection): People can't detect the watermark, and it won’t show up in the image's metadata. You need Google's specific scanner (an API or tool) to analyze the pixels and look for that hidden pattern. The scanner then gives one of three answers:
- Detected: The signature is present. This image was created by AI.
- Not Detected: The signature is not present. This is the critical part: it does not mean the image is real. It simply means it wasn't created by the AI in question, or the watermark has been damaged beyond recognition. It could easily be a deepfake from another source.
- Possibly Detected: The signature is partially present, suggesting the original image was likely generated in part by the AI model, but has been heavily modified.
While most people don’t care about pixel patterns or minor variations to token outputs, I don’t think this is a technology story; it’s a power/control/ecosystem story. SynthID is one of the first legitimate deepfake detectors I’ve seen, and it’s shockingly good.
It’s About Provenance: SynthID’s primary function is to allow Google (or another organization) to put a "Made by [whatever]" stamp on its content. For a CMO, this has one primary benefit: an audit trail. If you are generating thousands of creative assets using GenAI tools, SynthID provides a way to verify they came from your licensed source, helping to manage digital rights and asset inventories. That’s incredibly valuable, especially if someone later wants to claim otherwise.
It’s a Walled Garden Strategy: Why would Google invest so heavily in this? Because it creates a trusted protocol/ecosystem. By providing a (semi)-reliable way to identify their GenAI-created assets, Google is forcing everyone else to use their protocol (no company wants to have 9 different protocols, one for each Google, Meta, Claude, Grok, etc.). It has the side effect of positioning Google’s generative AI tools as more responsible and enterprise-ready than the open-source Wild West. The more the world fears deepfakes, the more valuable a "verifiable" content source becomes. The game is to make you reliant not just on Google’s creation tools, but their verification tools as well.
It’s One Move in a Perpetual Arms Race: SynthID is not foolproof. It never will be. For every brilliant engineer at Google creating a watermark, there's another one in a basement somewhere building a tool designed to strip it away. A sufficiently determined actor will always be able to scrub, scramble, or spoof these signatures. Relying on any single technology for security is a fool’s errand.
Do not outsource your brand's safety or your media intelligence to Google or anyone else. SynthID is a useful, interesting, and necessary component in the new media landscape, but it is not a solution.
Your responsibility is to build resilience. This means creating internal policies for the use and verification of all synthetic media, regardless of its source. It means fostering a culture of healthy skepticism. And it means understanding that in the zero-trust media environment, the only thing you can truly control is your own judgment. All that being said, expect adoption to pick up relatively quickly, especially if a few major players get on board with using it.
The Inevitable Ascension of YouTube: Your Next First Screen
The rise of YouTube is a relatively old story (I wrote about YouTube being the most underpriced impression on the web here) - and to frame this as a simple matter of growing ad revenue is to miss the point entirely. YouTube isn’t becoming the "first screen." - for much of your target audience, it likely already is. YouTube is the world’s most watched video platform. It is the only platform that is (more or less) universally loved by Gen-A and Boomers alike. It’s where plumbers turn for how-tos on repairing your sink, toddlers watch Bluey and millennials figure out how compound interest works.
Linear television is slowly dying. The broadcast networks are a graveyard of reruns and reality shows propped up by an aging (and dying) core audience. The much-hyped streaming wars have resulted not in a single victor, but in a fractured, frustrating landscape of a dozen subscriptions, each with its own walled garden of content. The consumer, fatigued by choice and cost, has retreated to the one place that aggregates everything: YouTube. It is the default commons for global attention.
What Google showcased at GML was not a series of ad formats, but the building blocks of a new commercial infrastructure layered directly onto these commons. Interactive storefronts, in-video shopping powered by real-time inventory, and AI-driven creative that adapts to viewer context - this is the transformation of a media channel into a transaction engine.
What this means for marketers is both exciting and (potentially) quite terrifying:
- Your Creative Is Obsolete: That 30-second spot from your TV buy is not just ineffective on YouTube; it is actively alienating. The lean-forward, active-choice environment of YouTube punishes passive, interruption-based advertising. Success is no longer about broadcast-quality production; it's about context and credibility. YouTube ad inventory will start looking more and more like Meta’s, with more hyper-relevant ads powered by progressively smarter algorithms.
- The CPM Tsunami Is Coming: Until now, YouTube impressions have been staggeringly under-priced relative to their value – most CPMs on YouTube are a miniscule fraction of the corresponding CPMs on Meta. That’s going to change. Google’s changes to the Power Pack will push more brands into YouTube, and their updates to measurement (see below) will - undoubtedly - better highlight the value YouTube is creating for advertisers. The end result? A lot more money will flood into YouTube over the next 6-12 months – which means higher costs.
- It’s a Full-Funnel Battleground: We all must stop referring to YouTube as an "upper-funnel" or "awareness" play. With direct transaction capabilities now embedded into the video player itself, YouTube is finally realizing its potential as a full-funnel platform, where a view can move from discovery to purchase in under 60 seconds. Your measurement framework and team structure must evolve to reflect this - the days of YouTube as the “brand” or “awareness” play while search is categorized as the “performance” channel are over.
YouTube is no longer the second screen. It is the main screen, the retail screen, and the social screen, all fused into one.
Measurement Enhancements + Cheaper Incrementality
One of the other “neat” additions to this year’s GML was a set of updates to Google’s measurement tool set, including a revamp to Google’s open-source MMM (Meridian). There were two things that made me happy here: (1) a 95% reduction in the spend required for an incrementality test (from $100k to $5k) and (2) the fact that Google is even talking about incrementality. While (1) may sound crazy, the math is actually quite simple using a Bayesian approach + Google’s treasure trove of data.
The second major announcement here was the addition of new data manager tools, allowing brands to collect more first-party data from more sources (like HubSpot, Salesforce, etc.) and store it in Google. Once again, no surprise – the best data wins. These updates simply make it easier for more marketers to ingest more data into Google, which, ultimately will have the effect of making Google’s ad products even better.
Final Thoughts: Don’t Abdicate, Orchestrate
Google wants to manage your spend, write your ads, and serve your customers. That’s fine, as long as you trust the system. But systems don’t think - they optimize. If the system isn’t optimizing toward your goal (or your best interest), what initially seems like a dream feature could turn into a nightmare.
Marketers must still lead. Define your customer, design your conversion path, and use Google’s tools as assistants. The brands that win over the next 12 months won’t be the ones that max out automation. They’ll be the ones that wield it with precision.
Enjoy your Memorial Day!
Cheers,
Sam
Loving The Digital Download?
Share this Newsletter with a friend by visiting my public feed.
Follow Me on my Socials