Google's AI-powered Shopping ad ecosystem: Where to focus your strategy

The Google Shopping of the past is part of a bygone era of PPC. If you want to maintain high conversion rates in the modern market, you’re now up against AI, with all its intricacies and limitations.

The modern Google Shopping campaign works hand-in–hand with everything that AI has to offer. It’s doing a lot of the heavy lifting; bidding, placements and even creative combinations. Performance Max and Smart Shopping campaigns alike all require marketers to rethink their approach to strategy with AI in mind.

This isn’t about setting up a campaign and letting AI take control; success lies in knowing how to guide the campaign to the best results. In this article, we’ll take a look at how to do just that. Let’s get started.

What AI powers Google Shopping today:

These aren’t entirely new developments; notice how many of these features are simply more advanced versions of elements of PPC that have been guiding our conversations for several years now.

Product feeds and merchant signals
AI is led by its data sets, and in PPC, that task falls to the product data found in your Google Merchant Centre. This is the foundation of your Shopping campaigns. 

AI takes this data and uses structured elements (such as titles, descriptions, GTINs, availability and price) to determine eligibility, relevance and ranking to match users to products in real time.

Beyond AI’s placement, merchant signals like pricing competitiveness, shipping speed, and seller ratings influence the visibility of your ads. So there’s a balance between AI and human-led metrics.

Auction-time bidding
Smart Bidding in Shopping and PMax optimises bids in real time based on predicted conversion value, using AI to guide these bids. Some of the signals considered include: device, location, time of day, user behaviour, and more.

Advertisers can set values (e.g. target ROAS), but bidding decisions are automated at query level. This prevents the need for hours of manual intervention at query level. It’s about saving you time, so you can focus on your wider strategy, knowing that the most detailed elements are taken care of.

Audience intent modelling
Intent modelling is one of the most interesting AI developments. Basically, Google's AI combines user data from Search, YouTube, Gmail, and third-party sites to predict the buying intent of users interacting with your brand.

Audience signals (e.g. in-market segments or customer lists) help guide this model, especially in PMax. The system can identify new high-intent users that you might not manually target.

It’s not a replacement for your own findings, but another option for you to consider, analyse and optimise.

Creative asset generation
One of the most time-saving AI assets is your asset generation. Often found in ad types like PMax, Google Ads dynamically assembles creatives using your provided assets: headlines, images, logos, and videos. It can also create auto-generated assets using feed content and product data. 

You feed in the very best examples that match your tone, marketing goals and style, and Google matches elements together, and finds the best pairings for conversions.

Creative combinations are matched to user behaviour and channel (Search, Display, YouTube, Discover etc.) In short: automation handles targeting, bidding, and creative assembly: human control focuses on input quality.

What you can't control

What you can't control (and why that matters)
If it’s been a minute since you last took command of a PPC campaign, the landscape has changed significantly. While marketers used to use manual controls in Search, Display and Shopping ads to maintain their marketing vision, now huge elements of that management are controlled by AI.

Limitations
There’s no longer any keyword-level targeting: ads are triggered based on product and query match, not explicit keywords. Within PMax campaigns there’s also no placement control: you can’t choose or exclude specific networks (e.g. YouTube Shorts).

Currently, you’ll find limited visibility into channel performance within PMax , although beta updates are emerging at speed, so this is clearly an avenue many marketers are keen to uncover in the coming years.

PMax channel performance reporting (beta update)
On that note, Google is testing new reporting to break down PMax performance by channel (Search, Display, YouTube etc.) Performance Max Search terms are now going beyond insights alone, and are being rolled out to Google Ads users as the latest dynamic addition of paid search. There’s even word of Channel Reporting coming out of BETA in the coming weeks, providing even more reporting features that are as bespoke as the most modern PPC AI options.

This allows advertisers to see where creative excels or underperforms. It’s still in early rollout and not widely available, but a promising step for transparency.

Why this matters
Without full visibility, it’s harder to diagnose performance issues as a PPC manager. Instead of what some marketers are used to (manual control of detailed elements) modern marketers must focus on optimisable levers: feed, signals, structure, and creative.

While this may take some getting used to, it’s the balance of machine learning and the expertise of your human-led knowledge of your brand and audience that will really set you apart from your competitors.

Where to focus your strategy in an AI-driven Setup

So where can you implement your expertise in an AI-driven world of PPC? There are still significant areas of your campaigns that you have direct control over:

Optimise your product feed
Strong feeds improve ad eligibility, relevance, and ranking, so use high-quality, descriptive product titles with relevant keywords.

Get the basics right; so ensure you have accurate Global Trade Item Numbers, product categories, and structured attributes (e.g. colour, size, brand). Build upon this foundation with compelling, clear product images, ideally on white backgrounds for additional clarity. Finally, use supplemental feeds to add missing data without changing your source feed.

Use audience signals wisely in PMax
Audience signals are only as useful as the data sets you provide them with, so upload first-party data: customer match lists, website visitors, past converters to give the most meaningful information to your campaign.

Layer in relevant in-market and affinity segments to guide any machine learning, and use audience exclusions where appropriate to avoid wasting spend on existing customers.

Remember, your audience signals don’t restrict targeting but guide optimisation. The better your signals, the better your reach.

Prioritise high-margin products
Your feed doesn’t just list products: it defines your catalogue strategy. Highlight and bid more aggressively on high-margin or best-selling items to signal to your algorithm that these items are worth the additional budget.

Exclude or down-bid low-converting, low-margin products to prevent your spend being eaten up by less useful clicks. And use custom labels to group products by performance tier or margin. This helps keep things organised and makes it easier when creating reports for analysis at margin-level.

Budget segmentation
Avoid lumping all of your spend into a single PMax campaign without clear structure, especially as a beginner.

Segment your campaigns by goal, product category, or lifecycle stage for intentional management and optimisation that is based on a clear marketing goa

When testing new products or seasonal ranges, use separate budgets. This prevents wasted ad spend and allows you to manage your budget differently depending on the average CPC.

Finally, consider splitting your Standard Shopping and PMax campaigns to retain some manual oversight where needed. At the end of the day, you don’t need to push every single product into one single feed, just because AI is involved.

Measurement in a blended shopping landscape

Flash back before the era of AI, and measuring success and its causes was simple. But with AI taking up a lot of heavy lifting, it’s vital to understand how to measure your results so you understand where AI is helping, or hindering.

Attribution challenges in PMax
Paid search attribution is becoming increasingly pivotal where AI is concerned. PMax blends Shopping with Display, YouTube, Discover and more, so conversions may not be clearly attributed to Shopping versus another channel.

Google uses data-driven attribution (DDA), but a lack of visibility can create reporting blind spots.

Beta reporting offers hope
The answer lies in what Google does next. A new channel reporting beta could help advertisers see how different placements perform. While it is still limited in rollout, it’s clearly a priority for Google, so watch this space. 

Human strategy still powers AI campaigns

At the end of the day, while automation delivers, it’s human strategy that drives performance. As a marketer, you still decide what products to promote, how to structure campaigns, and how to allocate budget. You’re at the control panel.

At the same time, creative assets, feed quality, and audience signals are all manually controlled inputs. This places you at the creative heart of your campaigns.

Beyond the initial setup and feed and asset curation, your manual ongoing testing, optimisation, and strategic restructuring are what will bring in success. AI is simply there to help speed up the process.

In Summary

Automation is here to stay, but so is the personal innovation that comes from human insight and strategy in PPC. If your feed is clean and controlled, your structure is intuitive and smart, and your creative assets are top quality, you’re set to scale with the help of AI.

Ultimately, the quality control lies with you, and that’s where modern strategies can pick up basic campaigns and elevate them to new levels of success. Retain control and strategic overview of your campaigns, and guide your latest AI innovation to new heights.