Google Marketing Live 2026: what an AI-native Google actually means for CMOs and investors

Having sat through plenty of Google product announcements over the years, most of them repackaging, a sales pitch, or a nudge toward more spend (‘have you thought about [YouTube/Demand Gen/PMAX]?’), Google Marketing Live 2026 is the most strategically significant version I can recall.

The reason is that if Google achieves its aims, then advertisers are about to lose meaningful control over three things at once: how they bid, how they measure, and where the purchase actually happens. 

The headline-grabbing AI creative tools are not the story. Most teams that I know are already experimenting with generative creative. The story is that Google is using this moment to make its bidding algorithms compulsory for new inventory, to position itself as the unified measurement layer across all digital channels (including Meta), and to pull the transaction itself inside its own ecosystem through Native Checkout and the Universal Commerce Protocol.

These shifts surface first in marketing, but the consequences run through pricing, retention, product distinctiveness and segmentation. The implication for CMOs and the boards behind them is that the competitive advantage shifts away from media execution and toward first-party data quality, commercial signal integrity, and measurement discipline that sits outside the platforms.

I think there are three shifts every CMO and board should be thinking about – and this article sets out what each means for the business, and what to do next.

1. Advertisers will be forced into black-box bidding to access Google’s new inventory

Google is using genuine changes in consumer search behaviour to justify making its automated targeting the price of entry to AI Mode placements.

Query structures are getting longer and traditional search patterns are fragmenting. The statistics Google shared were revealing for that reason. AI Mode queries are now three times longer than traditional searches, and according to Google, those queries generate 27% more conversions. Seventy-five percent of users report feeling more confident in purchase decisions when using AI-powered search experiences.

Our own client data shows the same direction of travel. The proportion of non-brand search queries containing five or more words has roughly doubled for selected B2B advertisers over the last six quarters. The growth is more modest across B2C businesses, but the behavioural shift Google is describing is real and visible in live accounts.

Percentage of queries with 5 or more words, weighted by volume of clicks

Whilst packaged as a roadmap to match genuine changes in consumer behaviour, it is also a deliberate attempt to protect search revenue against the growth of LLM-based discovery.

That commercial framing makes the announcements easier to interpret.

As Google shifts to an AI native experience, they announced that in order to access AI Mode placements for your ads, you will need to be running AI Max or Performance Max .

Whilst this may be being done to make the barrier to entry lower for advertisers, a more cynical view would be that the more Google requires you to trust its own black box in order to access specific inventory, the less control you have as the advertiser.

If Google forces the use of these strategies, advertisers will need to shift. With this shift comes less control and less granular optimisation than advertisers will have been used to.

With that said, across several live AI Max activations we have reviewed, the picture on AI Max specifically is more nuanced than the platform messaging suggests.

Early performance from clients supports a measured view rather than the transformative one Google is positioning. AI Max ROAS sits mid-pack, better than broad match, worse than exact, rather than meaningfully ahead of established match types.

ROAS, comparing match type from two businesses who have deployed AI max

More importantly, only around a quarter of AI Max spend in the most material activation went to queries the campaign had not previously served. The rest was re-routing of existing demand through a new matching mechanism. In a second account, the genuinely net-new queries that AI Max surfaced delivered zero attributed conversion value over the period observed.

Queries AI Max spend is triggering

Whilst the current results from these early tests look moderately appealing at best, these examples are from large, sophisticated B2C businesses, we would advocate any client of ours tests the performance for themselves in their own market, as the product develops. 

If Google is successful and advertisers move this way in numbers, advantage will move away from manual optimisation mechanics towards signal quality, data architecture, commercial interpretation and measurement discipline. The more automated the execution layer becomes, the more dangerously simplistic optimisation becomes, and the less control you have over performance as an advertiser. Further opaqueness is created in that much of the insight to understand performance is not easily accessible within the Google Ads interface. It is only visible to advertisers who pull the underlying query data out of the platform and audit it independently.  

The position we’d advocate for advertisers is measured scepticism: test it rigorously through holdout experiments rather than platform-reported attribution, audit query selection, and resist pressure to scale it without independent evidence of incrementality.

2. Advertisers will lose independent measurement if Google becomes the attribution layer

What makes the picture even less transparent is the second announcement.

Google introduced new ‘unified attribution’ within Google Analytics across all platforms (including Meta), promising to give a view of attribution across all of your digital activity and even forecast future performance.

Google has signalled clearly, and not for the first time, that it is seeking to become the unified measurement layer for increasingly automated marketing systems.

Conceptually, it makes sense. Channel fragmentation and attribution complexity have become difficult for many businesses to manage.

The tension, however, is obvious. The same platform now proposes that as an advertiser you should trust it to distribute spend, optimise delivery, define attribution logic, predict future value and report effectiveness. This combination would be useful if it was not the main media channel itself offering the service.

Would you really trust Google to tell you where the next most profitable auction opportunity sits? The system is unlikely to often conclude that the optimal commercial decision is to spend materially less with Google. We also see many examples with our clients where Google’s bidding models ‘leave money on the table’.

That does not mean the recommendations will always be wrong. In many cases they will improve performance.

The risk here mirrors the first issue. If the highly automated system allows reporting convenience to replace commercial understanding, the divergence between ROI reported by Google and what you see in your own P&L may widen.

In a client example, we compared Google’s attribution to a multi-touch model, utilising journey-level data to build an accurate picture of performance that everyone around the executive team can understand. (We refer to this at Coppett Hill as Rules-Based Attribution which combines journey level data with commercial business rules specific to each client we work with). 

 The results were very typical of this exercise, with Google overweighting the revenue it was attributing towards paid search, and underweighting performance it allocated to other channels.

Google overstates attribution of paid search performance vs Rules-Based Model

To be effective, marketing teams need to adopt their own transparent measurement plans that sit outside of the platforms to measure the impact of performance media spend. 

As they begin experimenting with new channels, teams need to think about incrementality testing and how to prove their worth before trusting that Google’s attribution will tell you the answer as to how it performed.

3. Advertisers will lose the customer relationship if the transaction moves inside Google

The third announcement is Google’s plan to shift towards, or defend themselves from, agentic commerce.

For those who missed the announcement, Google is positioning UCP to become a standard shared machine language across AI agents, merchants and payment tools. The use of UCP will enable Native Checkout: embedding a native ‘buy’ button into Search, Gemini and YouTube.

This move is designed to pull larger parts of the purchase journey inside Google’s own ecosystem. Google have of course tried this in the past with varied success. Whilst booking a restaurant tables or comparing hotels is possible (to differing levels of success) in Google today, they closed their financial services comparison service in 2016.

Historically, many businesses have treated Google as an acquisition layer that handed customers off to their owned platforms, where the commercially important signals live (conversion economics, margin, inventory position and so on).

If Google is successful in moving consumers to buy within the Google platform, the game changes for advertisers.

And if this is the future of e-commerce (which Google and other LLMs believe it will be), what does that mean for businesses with limited brand preference, non-unique product lines, weak retention mechanics or no clear reason for the customer to return to their platform?

For those that do implement UCP and Native Checkout, they will outperform if they can move faster and make more commercially accurate decisions about where the next marketing pound should go. Instead of how well you can execute ads, it becomes a question of the quality of your first-party operational data that you are able (and willing) to feed back into the platform itself.

A strategy that allows the business to optimise towards which products to accelerate, where margin pressure is emerging, where inventory risk exists or where customer quality is deteriorating will become really important signals to control.

This is a glimpse into Google’s view of the future. A world in which more of the shopping journey happens through conversational interfaces. It will be interesting to follow adoption and how advertisers perform as the features roll out across territories.

What should a CMO be doing right now?

Resisting automation entirely is not the right response. For many businesses, manual optimisation will not outperform machine-led execution over time, and even if the performance is not there yet, growing use of automation will be net beneficial in a lot of cases. But as the ecosystem becomes more black-boxed, the controllable layer is increasingly going to become first-party data quality, operational responsiveness and commercial signal integrity. 

On bidding: optimise toward your own economics, not the platform’s proxies.

If Google is going to force you into automated bidding to access AI Mode inventory, the quality of the signals you feed it becomes the differentiator. Many businesses still optimise acquisition around conversion volume, blended ROAS or shallow attribution metrics. As automation expands, your own view of Ideal Customer Profiles, margin, repeat behaviour and contribution economics is what stops the platform optimising toward proxies that look good in the dashboard but don’t move the P&L.

On measurement: build a transparent measurement plan that sits outside the platforms.

If Google is going to mark its own homework, you need an independent answer that exists outside of the platforms. That means incrementality testing as the default for new channels and tactics, with clear learning loops. It also means being prepared for budget and traffic to shift between channels if necessary. Utilising independent attribution reporting at a granular level and structured incrementality testing as part of your trading cadence will allow you to optimise based on your business, not the platform.

On the purchase journey: reshape the team for an AI-native acquisition environment.

If the transaction moves inside Google, the scarce capability becomes commercial interpretation rather than campaign orchestration. There is no established playbook, but the fundamental skills shift toward data literacy, commercial fluency and a willingness to experiment. The capability mix looks different from even twelve months ago, and the teams that adapt fastest will be the ones extracting commercial signal from a system designed to hide it.

What an investor should be asking the executive team

As AI further impacts the consumer journey, the next 12-24 months should see investors prepared for volatility in marketing results. It will be important to take a data-driven approach to testing new channels, and being prepared for teams to do this will be important. 

On loss of control over bidding:

• If Google CPCs rise 15% tomorrow, do we know how we would respond?

• Are we actively tracking metrics that show how the acquisition mix is changing for us even within platforms?

• How well-structured is our first-party data for AI-driven acquisition?

On loss of control over measurement:

• How are we measuring marketing effectiveness in real time to inform trading decisions?

• How are we measuring the impact of AI Overviews and LLMs on our business today?

• If SEO traffic falls 20% due to AI overviews, how would we respond?

On loss of control over the purchase journey:

• How exposed is the business if Google takes the website layer out of the journey?

• Is the product or offering distinct enough to be chosen in a conversational interface?

• Do we have the skill mix, or a plan for the skill mix, for an AI-first consumer journey?

With the channels in marketing shifting quicker than ever, not every business will build competitive advantage through media execution. Some will build it through understanding their own economics better than the platforms optimising around them.

If you would like to discuss how these platform shifts may affect your measurement, acquisition strategy or operational data architecture, please contact us.