Quick Answer
AI has fundamentally changed how Google Ads works. Smart Bidding processes millions of real-time signals to set bids more accurately than any human could manually.
Performance Max distributes spend across every Google surface from a single campaign. AI Max for Search expands keyword reach using Gemini’s machine learning.
The result is that properly configured AI-driven campaigns can genuinely outperform manual setups, but only when the inputs are right.
Inaccurate conversion tracking, weak audience signals, and lack of regular account maintenance will result in the algorithm confidently optimising towards the wrong outcome. The AI is only as good as what you feed it.
How Google Ads Became an AI-First Platform
Google Ads three or four years ago looks almost unrecognisable compared to what it is today.
Manual bidding, keyword-level control, and separate campaigns for Search, Shopping, and Display were the standard approach. The advertiser made the decisions; the platform executed them. That model has been systematically dismantled since 2021.
Google has moved almost every meaningful lever: bidding, creative, targeting, audience matching, placement distribution, under the control of its machine learning systems. Enhanced CPC was deprecated in March 2025.
Dynamic Search Ads are being auto-upgraded to AI Max starting September 2026. AI Max attracted hundreds of thousands of advertisers in its first year and is now the fastest-growing AI product in Google’s portfolio.
Source: Google Ads & Commerce Blog, 2026
For e-commerce businesses, this shift is significant. It means the traditional skill of building tightly structured campaigns with exact-match keywords and manual bid adjustments matters less than it used to.
What matters now is the quality of data you give the algorithm, the structure you use to guide it, and the human judgement you apply to interpret what it’s doing.
This isn’t a reason to panic. AI-driven Google Ads campaigns deliver results that manual setups simply can’t match at scale, but only when they’re set up and managed correctly. And, most businesses aren’t doing this.
Smart Bidding Is The Foundation
Smart Bidding is Google’s umbrella term for its AI-powered bid strategies: Target ROAS, Target CPA, Maximise Conversion Value, and Maximise Conversions. If you’re running any form of Google Ads in 2026, you’re almost certainly using one of these.
The reason Smart Bidding outperforms manual bidding isn’t complicated: it processes signals that a human simply can’t act on in real time.
Device type, location, time of day, browser, previous search behaviour, whether the user has visited your site before and what else they’ve searched for in the last 24 hours.
All of these are weighted and combined into a bid decision that happens in the fraction of a second before an auction fires.
What Smart Bidding does well for e-commerce:
Target ROAS is the AI-powered strategy recommended for most e-commerce accounts. You tell Google what return on ad spend you want to achieve, and it adjusts bids across every auction to hit that target across the campaign as a whole.
For businesses selling products with different margins, it means bids are weighted towards the searches most likely to deliver the commercial outcome you’ve defined.
For new accounts: jumping straight to Target ROAS before Google has sufficient conversion data is one of the most common setup mistakes I see. Without enough signal to learn from, the algorithm restricts traffic heavily trying to hit a target it doesn’t yet understand.
I always recommend starting with Maximise Conversion Value and only introducing a Target ROAS once the campaign has accumulated sufficient conversion history.
Source: Triple Whale, 2025
For businesses working with tighter budgets where conversion volume is lower, our guide to running Google Ads on a small budget covers how to structure campaigns to build data efficiently without burning spend.
Accurate Conversion Tracking is Vital:
The AI is only as smart as the data you feed it, and if your tracking is counting the wrong events, firing on the wrong pages, or missing transactions entirely, the algorithm will optimise toward that flawed signal with complete confidence.
For UK and EEA advertisers, Google’s Consent Mode V2 has been mandatory since March 2024 and enforcement tightened significantly in mid-2025. On average, only 31% of users accept tracking cookies. Without Consent Mode V2 implemented correctly alongside Enhanced Conversions, Smart Bidding is learning from a fraction of your actual conversions.
A real-world case reported by PPC.land in April 2026 documented Harvest Digital measured conversions dropped 90% overnight after a Consent Mode misconfiguration, with only 40% of that attribution data recoverable after the fact.
Source: PPC.Land, 2026
In my experience, the best setup for conversion tracking in 2026 means Consent Mode V2 correctly configured, Enhanced Conversions enabled to recover signal lost to privacy restrictions, GA4 events imported cleanly into Google Ads, and server-side tagging for accounts spending past around £10,000 per month.
Before you look at bid strategies, campaign structure, or any AI feature, check your tracking.
The latest development worth knowing is Smart Bidding Exploration:
Originally available only for Search campaigns, it was extended in May 2026 to Performance Max and Shopping.
Rather than loosening your targeting, it allows you to set a ROAS tolerance range, for example, a primary target of 400% with exploration allowed down to 350%.
Within that range, Google’s AI bids on queries it would otherwise ignore: new audience segments, emerging intent signals, search terms outside your existing keywords.
According to Google’s internal data, Search campaigns using Smart Bidding Exploration on average see 27% more unique converting users.
It works best for established accounts with strong conversion history, healthy margins, and a genuine need to find new customers beyond their current audience.
Source: Google Ads, 2026
Performance Max (The Most Misunderstood Campaign Type in E-Commerce)
Performance Max (PMax) is Google’s fully AI-driven campaign type. You provide creative assets, your product feed, audience signals, and a conversion goal. Then Google’s algorithm decides where to show your ads, who to show them to, and how much to bid, across Search, Shopping, YouTube, Display, Gmail, and Maps simultaneously.
It launched in 2021 and was immediately controversial and labelled a “black box”, where advertisers would spend money and hope for the best. Early versions offered almost no transparency into where the budget was being spent or why.
Since then, Google has made meaningful progress on the control side by providing campaign-level negative keywords, channel performance reporting that shows which properties drive conversions, and search theme inputs doubled from 25 to 50 per asset group.
What the data shows:
Google’s official data shows a 12% increase in conversion value when upgrading from Smart Shopping to Performance Max at the same or better ROAS, with over one million advertisers now using Performance Max globally.
Swedish fashion retailer Lindex reported a 27% ROAS improvement after switching to Performance Max, crediting Google’s AI with identifying high-value customers from their membership data in a way their previous campaigns couldn’t.
Source: Think With Google, 2022
What E-Commerce Businesses get wrong with PMax:
The biggest mistake is treating it as “set-and-forget”. The advertisers getting the most out of Performance Max in 2026 aren’t treating it as automation that runs itself. PMax amplifies whatever you feed it.
Several specific issues trip up e-commerce businesses:
Conversion data threshold. Performance Max needs conversion history to learn. If your account has fewer than 30 conversions in the last 30 days, you don’t have enough data for it to optimise effectively. Below this threshold, the algorithm is essentially guessing.
Single ROAS target across different margin products. A single target ROAS applied to a campaign containing both 10% margin products and 60% margin products is a recipe for inefficiency. The algorithm defaults to your bestsellers. If you’re starting out, begin with your highest-margin products only and expand once the algorithm has enough data to optimise effectively.
Audience signals are guidance, not restrictions. When you add an audience signal, you’re teaching Google what your best customers look like so it can find similar people, not restricting who can see your ads.
Customer match is one of the most powerful signals available, it works by uploading your existing customer data (email addresses, phone numbers, or CRM lists) directly into Google Ads, allowing the algorithm to identify and target new users who share the same characteristics as your best buyers.
Google’s own guidance is clear on this, customer match lists help the algorithm identify high-value users faster than starting from scratch.
The hybrid approach is becoming the standard:
Standard Shopping handles your core, known-intent traffic, whereas PMax handles full-funnel discovery across Search, YouTube, Display, Gmail, Discover, and Maps. The hybrid approach is becoming the go-to strategy in 2026.
AI Max for Search: What the Data Actually Shows
AI Max for Search is Google’s newest significant product; launched in 2024, out of beta in 2025, and being rolled out to replace Dynamic Search Ads from September 2026. It’s important to understand what it is and isn’t.
Unlike Performance Max, AI Max is not a separate campaign type. It’s an enhancement layer you apply to existing Search campaigns. It does three things:
- Expands your keyword targeting to queries you haven’t explicitly bid on.
- Dynamically generates ad copy variations.
- Adjusts landing page selection to match query intent.
The infographic below breaks down exactly how AI Max works, what to enable, and whether it’s right for your business before I get into what the data actually shows:
Google's headline claim:
According to Google’s official Help Centre, advertisers who activate AI Max typically see 14% more conversions at a similar CPA or ROAS. With a further 7% uplift available for those enabling all three AI Max features simultaneously.
What independent data shows:
Smarter Ecommerce’s analysis of more than 250 retail campaigns running AI Max found a median revenue lift of 13%, close to Google’s claim. The catch is what came with it: a median CPA (cost-per-acquisition) increase of 16%. The ROAS range across campaigns stretched from +42% to -35%, meaning outcomes are highly variable and the average conceals significant downside risk.
Search Engine Land’s analysis of 23 AI Max tests across 16 mature advertiser accounts found that campaigns enabling all three AI Max features simultaneously saw a 40% higher success rate than those using only the baseline search term matching feature.
Overall, the message isn’t to avoid AI Max. It’s to go in prepared, with tracking that can catch problems early.
Worth flagging for e-commerce:
Google’s official 14% conversion lift benchmark explicitly excludes retail advertisers. The implication is that retail and e-commerce advertisers either do not see the same lift, or performance is variable enough that including retail would significantly lower the reported average.
This doesn’t mean it won’t work for your e-commerce business. It means test it carefully, with a defined measurement window and clear success criteria, before rolling it across your account.
What’s coming: Google announced AI Max for Shopping in closed beta at its April 2026 update, allowing e-commerce advertisers to apply AI Max bidding logic to sponsored product and AI Overview placements. This is worth watching closely as it rolls out.
Dynamic Creative and Asset Optimisation
Responsive Search Ads (RSAs) have been the standard Search ad format since Google retired Expanded Text Ads in 2022. You provide up to 15 headlines and 4 descriptions; Google’s AI tests combinations and serves the most relevant assembly for each individual search query.
The practical implication for e-commerce is significant: rather than writing one static ad, you’re building a creative framework that the algorithm adapts in real time. A user searching “lightweight women’s running shoes” sees a different headline combination than a user searching “women’s trail running shoes size 6”, even if both queries trigger the same ad group.
As of 2026, Google requires a small label on ads containing AI-generated elements. Early testing suggests this label has minimal impact on click-through rates.
However, the quality of AI-generated copy varies, and industries with strict compliance or brand voice requirements should review generated assets carefully.
For e-commerce businesses running Performance Max, Google’s Asset Studio now generates images and videos using Imagen 4 and Veo inside Google Ads.
For stores currently running PMax with only product images and no video assets, this matters, Google’s algorithm favours campaigns with video.
The practical advice here is simple: give the algorithm more to work with, not less. More headline variations, stronger asset diversity, video where possible. But review what the AI generates before it runs. Auto-generated assets have an inconsistent track record, and copy that doesn’t reflect your brand voice, or contains inaccuracies, will go live unchecked if nobody reviews it.
Driving more traffic only delivers results if your site converts it, our guide to ecommerce conversion rate optimisation covers how to make sure it does.
Third-Party AI Tools Worth Knowing About
Beyond Google’s native features, a category of third-party tools has emerged to help advertisers manage, audit, and improve their campaigns:
- Optmyzr is a rule-based automation and reporting platform popular with agencies. It doesn’t replace human strategy but automates repetitive optimisation tasks; bid adjustments, budget pacing and quality score monitoring; freeing up time for the thinking that actually matters.
- AdCreative.ai generates ad imagery and copy variants at scale, useful for e-commerce businesses with large product catalogues that need creative variety without a large production budget. Quality varies by use case; best treated as a starting point for human refinement rather than a final output.
- Smarter Ecommerce (SMEC) specialises in feed-based campaign management and performance analysis for e-commerce.
Keep in mind, several tools make strong claims about autonomous account management that, in practice, still require significant human oversight to perform well. Any tool that promises to run your Google Ads without ongoing human involvement should be approached carefully.
Where AI Still Needs Human Oversight
The algorithm doesn't understand your business
AI doesn't know your brand either
Google’s AI has no understanding of your tone of voice, your unique selling points, or what makes your business different from every other advertiser in your category. It will generate ad copy, match queries, and serve assets based entirely on what the data tells it converts, not what accurately represents who you are.
I’ve seen auto-generated headlines that were factually misleading, copy that could have belonged to any competitor, and landing page selections that made no sense in context. The AI optimises for the click. Whether that click reflects your brand well is your responsibility to check.
It cannot see across your entire account
Google’s AI optimises within the parameters of individual campaigns. It cannot see your full account. It does not know that Campaign A and Campaign B are bidding on the same query, your margin structure, seasonal patterns and which leads have closed.
Campaign cannibalisation (where PMax and AI Max Search campaigns compete for the same queries), is a real and expensive problem that requires human structural oversight to solve.
Negative keywords still matter
The AI will find traffic. Not all of it will be good traffic. Without an active negative keyword strategy, Performance Max and AI Max campaigns will burn budget on irrelevant queries.
One sporting goods retailer saw an immediate 15% cost reduction by adding “free” and “used” as negative keywords, eliminating unprofitable traffic. That kind of hygiene doesn’t happen automatically.
Source: Dataslayer, 2025
Setting aggressive ROAS targets too early kills volume
Setting aggressive ROAS targets before the algorithm has enough data can reduce total conversion volume dramatically, up to 50% in some cases. Give the system time to learn before you start tightening the levers.
Budget guardrails are your responsibility
The Q1 2026 reality check
Smarter Ecommerce’s Q1 2026 benchmark report found that Performance Max conversion growth stalled entirely, dropping from approximately 12% year-on-year in Q4 2025 to 0% in Q1 2026.
AI-driven campaigns are not immune to market conditions, seasonal shifts, or changes in the competitive landscape. Human interpretation of that data and the strategic response to it, remains essential.
Common Mistakes E-Commerce Businesses Make with Google Ads AI
| Mistake | Why it matters |
|---|---|
| Mistake Turning on AI before fixing tracking | Why it matters The AI learns from whatever data you give it, wrong tracking means it optimises confidently toward the wrong outcome. Start with a Google Ads audit before touching any AI feature. |
| Mistake One campaign for everything | Why it matters A single PMax campaign with one ROAS target can’t reflect different product margins or customer values. Segment campaigns to match your actual business economics. |
| Mistake Ignoring the feed | Why it matters Your Merchant Centre feed is your campaign. Inaccurate data limits PMax regardless of budget. Read our guide to optimising Google Shopping campaigns. |
| Mistake Pausing during the learning phase | Why it matters Every significant change re-starts the learning phase. Stability in early weeks outweighs constant tweaking, let the algorithm accumulate data. |
| Mistake Measuring the wrong thing | Why it matters If PMax is capturing branded traffic that would have converted anyway, your ROAS figure is inflated. Separate incremental from baseline performance. See our SEO vs Google Ads guide for more on attribution. |
How Design Box Approaches AI-Powered Google Ads for E-Commerce
We run Google Ads campaigns for e-commerce businesses across the UK, and the shift to AI-first campaign management has changed how we work, but not what we care about.
The fundamentals haven’t changed: conversion tracking has to be right, the account structure has to reflect the business’s actual economics, and every decision has to be justified by data rather than platform defaults.
What’s changed is that we now spend more time on feed quality, audience signal strategy, and interpreting AI behaviour than we do on manual bid adjustments. We use Performance Max where it’s appropriate and test AI Max carefully and methodically.
If you’re an e-commerce business wondering whether your current Google Ads setup is keeping pace with how the platform has evolved, the most useful thing we can do is look at it.
We’ll tell you honestly where the issues are, whether AI features are being used well or just turned on, and what the realistic opportunity looks like.
FAQs
What is AI Max for Search in Google Ads?
AI Max for Search is a suite of AI-powered features that Google layers onto existing Search campaigns, rather than a standalone campaign type.
It expands your keyword targeting using machine learning to match ads to queries you haven’t explicitly bid on, generates dynamic ad copy variations, and adjusts landing page selection to match search intent.
It became widely available in 2025 and is Google’s fastest-growing AI ads product. Dynamic Search Ads are being auto-upgraded to AI Max from September 2026.
Does Performance Max work for e-commerce?
Yes, but results depend heavily on how it’s set up and managed. Google’s official data shows a 12% average increase in conversion value when upgrading from Smart Shopping.
Independent research suggests 10–30% conversion improvements are realistic with proper configuration.
The businesses that get the most from PMax treat it as a data-driven tool that requires active guidance: quality audience signals, a clean Merchant Centre feed, campaign-level negative keywords, and regular performance reviews, rather than automation that runs itself.
Is Smart Bidding better than manual bidding?
For most e-commerce businesses with sufficient conversion volume, yes. Smart Bidding processes real-time contextual signals (device, location, time, browsing behaviour)that manual bidding simply cannot act on at the speed of an auction.
The critical requirement is accurate conversion tracking. Without it, Smart Bidding optimises toward the wrong goal. With it, Target ROAS typically outperforms manual CPC bidding for established accounts.
How much conversion data does Google Ads AI need to work effectively?
The generally accepted threshold is at least 30 conversions per month at the campaign level before AI-powered bid strategies can optimise reliably. Below this, the algorithm doesn’t have enough signal to make accurate predictions.
For new campaigns or lower-volume accounts, Maximise Conversions (without a target) is often a better starting point than Target ROAS, allowing the algorithm to gather data before tighter constraints are applied.
What are the risks of using AI features in Google Ads?
The main risks are: optimising toward the wrong conversion event if tracking is inaccurate; campaign cannibalisation if PMax and AI Max Search campaigns compete for the same queries without structural guardrails; budget waste from insufficient negative keyword management; and inflated ROAS figures if branded traffic is being captured by PMax and attributed as incremental performance.
None of these are reasons to avoid AI features, they’re reasons to implement them with human oversight rather than treating them as self-managing.
Does AI in Google Ads replace the need for an agency or specialist?
No, Google’s AI optimises within campaigns; it doesn’t see your full account, understand your margin structure, or make strategic decisions about where your budget should go.
The shift to AI-first Google Ads has changed the skill required to manage campaigns well, but it hasn’t reduced the value of expert human oversight. If anything, it’s made the gap between well-managed and poorly-managed accounts larger, because the AI amplifies whatever inputs it receives.
How does Google Ads AI interact with AI Overviews and AI Mode search?
This is one of the most significant developments in paid search right now. Google has embedded ads directly into AI Overviews and AI Mode, the conversational search experiences more users now see first.
These aren’t banners next to a result; they surface when relevant to both the user’s query and the AI-generated response itself. Google announced AI Max for Shopping in closed beta in April 2026, allowing e-commerce advertisers to apply AI Max bidding logic specifically to these AI Overview placements.
For e-commerce businesses, this is an emerging opportunity worth monitoring closely. Our guide to AI and automation in marketing covers the broader implications.
