Last updated: April 14, 2026 | Reading time: 24 minutes
There is a moment in the life of every advertising strategy where the data becomes impossible to argue with. For AI-driven advertising, that moment arrived in 2025 when L'Oreal, the world's largest cosmetics company, published what has since become the most cited case study in Google Ads history.
The numbers were striking. A 2x increase in conversion rates. A 31% reduction in cost per conversion. A 27% lift in conversion value. A 20% boost in return on ad spend. All achieved not through a bigger budget or more aggressive bidding, but through a single technology shift: enabling AI Max on their Search campaigns.
But the L'Oreal case study is just one data point in what is now an overwhelming body of evidence. Nielsen's independent analysis of over 50,000 brand campaigns and more than 1 million performance campaigns confirmed what early adopters already knew. AI-powered advertising is not incrementally better than manual management. It is categorically different in what it can achieve. And the gap is widening.
This article breaks down the L'Oreal case study, aggregates the broader evidence base, explores what these results mean for advertisers of every size, and addresses the one critical question that Google's marketing materials conveniently skip over: whose goals is the AI actually optimised for?
L'Oreal did not stumble into AI Max. They were already one of Google's most sophisticated advertising partners, with a track record of early adoption across broad match, Smart Bidding, and Performance Max. When Google launched AI Max for Search campaigns in 2025, L'Oreal was among the first major brands to test it at scale, starting in Chile before rolling the approach out globally.
AI Max for Search is not a new campaign type. It is a suite of AI-powered features that layers on top of existing Search campaigns. L'Oreal activated three core capabilities.
Search term matching expanded their reach beyond their existing keyword lists. Instead of relying solely on the keywords their team had manually selected, AI Max used Google's machine learning to identify relevant queries that L'Oreal's ads should appear for, even if those queries were not in their keyword lists. This is how they started capturing searches like "what is the best cream for facial dark spots," a natural-language, conversational query that no human keyword researcher would have included in a traditional campaign structure.
Text customisation allowed Google's AI to dynamically generate and adapt ad copy based on each user's specific search context. Instead of static headlines and descriptions, L'Oreal's ads adjusted in real time to match the intent behind each query, creating a more relevant experience for the user and a higher click-through rate for L'Oreal.
Final URL expansion directed users to the most relevant page on L'Oreal's website based on their search. If someone searched for a specific product concern, the AI might bypass L'Oreal's generic category page and send the user directly to the product page most likely to convert them. Less friction, faster path to purchase.
The numbers Google published tell a clear story. Click-through rates lifted by 67% across campaigns compared with other match types. Cost per conversion fell by 31%, meaning L'Oreal was paying nearly a third less for each customer action. Glycolic products saw double the conversions, while serums achieved a 70% increase. Overall, the campaign delivered a 27% lift in conversion value and a 20% boost in return on ad spend.
What makes these numbers particularly compelling is that L'Oreal was not starting from a low baseline. This was already one of the most well-optimised advertising operations in the world, managed by a dedicated team of specialists using sophisticated tools and strategies. A 31% improvement in cost per conversion on top of an already elite operation is not marginal. It is the kind of performance shift that restructures how you think about the entire channel.
L'Oreal brought three advantages that amplified AI Max's effectiveness. First, they had massive data volume. Google's machine learning algorithms perform better with more conversion signals, and L'Oreal generates enormous volumes of traffic and sales data across hundreds of products and dozens of markets. The AI had plenty of signal to learn from.
Second, they had deep product content. AI Max's text customisation and URL expansion features work best when there is rich, detailed content across a wide range of landing pages. L'Oreal's product pages, ingredient guides, and skincare concern pages gave the AI a large library of relevant destinations to choose from.
Third, they had a culture of testing. L'Oreal started in a single market (Chile), measured the results rigorously, and only then scaled globally. They did not flip a switch and hope for the best. They ran structured experiments, validated the data, and made informed decisions about expansion.
The L'Oreal case study is powerful, but sceptics could reasonably argue that it is a single brand in a single vertical. What about everyone else?
That is where Nielsen's research becomes critical. In 2025, Nielsen published the results of an independent analysis commissioned by Google, examining more than 50,000 brand campaigns and over 1 million performance campaigns in the United States. The study spanned a two-year period from July 2022 to June 2024 and covered industries including food, beverages, restaurants, home and personal care, retail, apparel, telecommunications, and automotive.
The methodology was Marketing Mix Modelling, a statistical technique that isolates the impact of specific marketing activities on business outcomes while controlling for external factors like seasonality and economic conditions. This is not self-reported survey data or cherry-picked case studies. This is rigorous econometric analysis across a massive sample.
AI-powered Broad Match delivered 15% higher ROAS and 10% higher sales effectiveness compared to other keyword match type strategies. This is the foundation of AI Max's search term matching feature, and the data confirms that letting AI expand beyond manually curated keyword lists consistently outperforms the traditional approach.
Performance Max delivered 8% higher ROAS and 10% higher sales effectiveness compared to Search-only strategies. This validates the cross-channel approach where Google's AI allocates budget and creative across Search, Shopping, YouTube, Display, Gmail, Discover, and Maps based on where conversions are most likely.
AI-powered video campaigns on YouTube delivered 17% higher ROAS compared to manually optimised campaigns. The gap was particularly pronounced when brands combined Video Reach Campaigns with Video View Campaigns, which together produced a 23% increase in sales effectiveness compared to Video Reach alone.
Adding Demand Gen to Search and Performance Max delivered a 10% increase in ROAS and 12% higher sales effectiveness compared to campaigns without Demand Gen. This demonstrates that the AI-driven full-funnel approach, where awareness, consideration, and conversion campaigns work together in an integrated ecosystem, outperforms siloed campaign management.
The conclusion from Nielsen's analysis was unambiguous. AI-powered advertising solutions consistently outperformed manual campaigns across both ROAS and sales effectiveness, and the biggest gains came when brands combined multiple AI solutions strategically.
L'Oreal and Nielsen's aggregate analysis are the most cited sources, but they are far from the only evidence. The case studies keep accumulating.
MyConnect, an Australian utility connection service, was already running Smart Bidding and broad match before enabling AI Max. They still saw a 16% increase in leads at 13% lower cost per action, with 30% of conversions coming from entirely new queries that their existing keyword lists would never have captured.
LG Electronics tested Demand Gen campaigns against their paid social campaigns and achieved a 24% higher conversion rate while reaching high-value customers at 91% lower CPA. That is not a marginal improvement. That is a fundamental shift in channel economics.
KEH Cameras transitioned from Standard Shopping to Performance Max and saw a 76.3% increase in advertising revenue in the first quarter after migration, according to their agency Inflow.
BioRender, a B2B SaaS company, implemented AI Max and achieved a 208% increase in click-through rates, demonstrating that the technology is not limited to consumer brands with massive budgets.
Escentual.com, a UK beauty retailer, tested campaign total budgets (a feature that expanded to Search, PMax, and Shopping in early 2026) and saw a 16% increase in website traffic without exceeding their budget, alongside a 5% improvement above their target ROAS.
The pattern is consistent across industries, geographies, and company sizes. AI-driven advertising produces better results than manual management. The question is no longer whether this is true. The question is what to do about it.
Understanding why the results are so consistent requires understanding what AI can do that human campaign managers fundamentally cannot.
Google's Smart Bidding evaluates thousands of real-time signals for every single auction. Device type, time of day, location, browser, operating system, previous search history, on-site behaviour, demographic signals, and hundreds of contextual factors that no human could possibly assess in the 100 milliseconds between a user's search and the ad auction completing. A human campaign manager makes one static bid decision. The AI makes thousands of dynamic micro-decisions per day, adjusting for variables the human does not even know exist.
L'Oreal sells hundreds of products across dozens of markets in multiple languages. The number of possible combinations of user intent, product relevance, creative messaging, and landing page alignment is staggering. A human team, no matter how talented, can only test a fraction of those combinations manually. AI Max tested all of them simultaneously, discovered that "what is the best cream for facial dark spots" was a high-converting query for a specific L'Oreal product, and adjusted the creative and landing page to match, all without any human intervention.
Human campaign managers check performance dashboards during business hours. They take weekends off. They go on holiday. They have dozens of other responsibilities competing for their attention. AI runs 24 hours a day, 7 days a week, 365 days a year. Every minute of every day, it is evaluating performance, testing new approaches, and adjusting strategies. The compounding effect of continuous optimisation is enormous, and it explains why the performance gap between AI-managed and human-managed campaigns tends to widen over time rather than narrow.
Google shipped more than 15 significant platform updates in the first two months of 2026 alone. AI Max text guidelines, PMax channel-level reporting, Campaign Mix Experiments, Merchant Center product ID enforcement, Direct Offers in AI Mode, Shoppable CTV in Demand Gen, and more. Each of these changes creates new optimisation opportunities and new risks. An AI system can evaluate and respond to every update immediately. A human team is still discussing the implications in their weekly standup while the AI has already implemented the optimal response.
Here is where we need to be honest about something the L'Oreal case study, and Google's marketing materials more broadly, carefully avoid discussing.
When you enable AI Max, Performance Max, Smart Bidding, or any of Google's AI-powered tools, you are handing optimisation control to Google's AI. Google's AI. Not your AI. Not an independent AI. Google's.
Google is an advertising platform. Its revenue comes from advertisers spending money on its platform. Google's AI is brilliant, and the evidence overwhelmingly shows it improves campaign performance. But there is an inherent tension in having the platform that sells advertising also be the one that decides how much you should spend and where.
L'Oreal can manage this tension because they have a dedicated team of specialists monitoring every campaign, cross-referencing Google's data with their own internal analytics, and maintaining independent oversight of how their budget is allocated. They have the resources, the expertise, and the infrastructure to ensure that Google's AI is working toward their goals, not just Google's.
Most businesses do not have this luxury. A mid-market ecommerce brand spending $15,000 per month on Google Ads does not have a team of PPC specialists independently validating every AI decision. They enable Smart Bidding, trust the algorithm, and hope for the best. And while the algorithm genuinely will improve their campaign performance (the Nielsen data is clear on this), there is no independent check on whether the AI is optimising for maximum advertiser profit or maximum Google revenue. These two objectives overlap significantly, but they are not identical.
This is not a conspiracy theory. It is a structural reality of how platforms work. Google wants you to succeed because successful advertisers spend more money on Google. But "succeed enough to keep spending" and "achieve the absolute maximum return on every dollar" are two different things.
The biggest misconception about AI-driven advertising is that it only works at enterprise scale. The data says otherwise.
The Nielsen study covered campaigns across every advertiser size, not just global brands with seven-figure budgets. The performance improvements were consistent whether the brand was spending millions per month or thousands. The principles are the same. More data helps. But you do not need L'Oreal-scale data to see meaningful results.
A regional dental practice that transitioned from manual optimisation to AI Max saw their campaign performance improve dramatically with a modest budget. BioRender, a B2B SaaS company, is not a household name, but they achieved a 208% increase in click-through rates. MyConnect is an Australian utility connection service, not a Fortune 500 company, but they found 30% more conversions from queries they never knew existed.
The key insight from these smaller case studies is that AI-driven optimisation disproportionately benefits advertisers who lack the resources for sophisticated manual management. L'Oreal had a team of specialists optimising their campaigns before AI Max. A small business owner managing their own Google Ads account does not. For that small business owner, the jump from manual management to AI-powered optimisation is not a 31% improvement. It can be transformational.
Let the AI find queries you would never think of. L'Oreal discovered "what is the best cream for facial dark spots" as a high-converting query. A small skincare brand would discover their own equivalent, the long-tail, conversational queries that their customers actually use but that no human keyword researcher would ever include in a campaign. These net-new queries often convert at higher rates because they represent specific, high-intent search behaviour.
Feed the AI good data. L'Oreal's deep product content gave AI Max a large library of landing pages and product information to work with. Even at a smaller scale, the principle holds. Better product descriptions, more detailed landing pages, and accurate conversion tracking all give the AI better signals to optimise against. The quality of your inputs directly determines the quality of the AI's outputs.
Test before you scale. L'Oreal started in Chile. You can start with a single campaign. Run an AI Max experiment against your existing setup, let it run for four to six weeks, and evaluate the data before expanding. This measured approach works whether your budget is $5,000 or $5 million.
Combine AI tools strategically. Nielsen's data showed that the biggest performance gains came from combining multiple AI solutions. Using AI Max with Smart Bidding Exploration, running Performance Max alongside Search, adding Demand Gen for top-of-funnel. The same principle applies at smaller budgets. A $10,000 monthly budget split across AI Max Search, Performance Max, and Demand Gen will typically outperform the same $10,000 concentrated in a single manually managed Search campaign.
This is where the story gets interesting for advertisers who are not L'Oreal.
Google's AI tools genuinely work. The evidence is overwhelming. But they work within constraints that most advertisers are not equipped to manage. Google's AI optimises for Google's platform across Google's properties using Google's data. It does not monitor whether your PMax campaigns are cannibalising your Search traffic. It does not independently verify that your conversion tracking is accurate. It does not ensure that AI Max's query expansion is not wasting budget on irrelevant long-tail traffic. And it does not coordinate your entire campaign ecosystem to ensure that your Demand Gen, Search, Shopping, and Performance Max campaigns are working together toward your specific business objectives rather than Google's platform objectives.
groas was built to close this gap. Its autonomous AI agents operate on top of Google's native AI tools, adding an independent optimisation layer that is aligned exclusively with your goals, not Google's.
When L'Oreal enabled AI Max, they had a team of specialists monitoring the results, adjusting configurations, and ensuring the technology served their objectives. groas provides that same level of oversight autonomously, for every client, at every budget level. It configures AI Max text guidelines based on your brand voice and conversion history. It monitors PMax channel allocation to prevent budget waste. It runs Campaign Mix Experiments to determine the optimal distribution across campaign types. It evaluates Smart Bidding Exploration results and adjusts ROAS targets based on actual margin data rather than Google's default recommendations.
The result is that advertisers using groas get the full benefit of Google's AI innovations, the same technology that produced L'Oreal's results, plus an additional layer of independent optimisation that ensures those innovations are working for their profit rather than Google's revenue.
Analysis across groas clients implementing AI Max shows an average 23% performance improvement across all verticals, with 15% of that coming from groas's own optimisation layer on top of what AI Max delivers natively. That additional 15% is the difference between using Google's AI and having your own AI that works alongside it.
The trajectory is clear and accelerating.
Google Marketing Live on May 21, 2026 will almost certainly announce further expansion of AI-powered features. The Universal Commerce Protocol is enabling AI agents to handle the entire shopping journey from discovery to checkout. Direct Offers inside AI Mode are pulling the conversion moment earlier into Google's own interface. Demand Gen's Shoppable CTV is bringing AI-optimised advertising to connected television screens.
Every one of these developments widens the gap between advertisers who embrace AI-driven management and those who cling to manual optimisation. The L'Oreal case study was a preview. Nielsen's analysis was the confirmation. The brands that are winning in 2026 are the ones that recognised this shift early and built their advertising strategy around it.
But embracing AI does not mean surrendering control. It means deploying AI strategically, with independent oversight, aligned with your business objectives, and continuously adapting to Google's relentless pace of innovation. That is what L'Oreal does with their team of specialists. That is what groas does with autonomous AI agents. And the advertisers who figure this out first will have a compounding advantage that only grows larger over time.
What results did L'Oreal achieve with AI Max?
L'Oreal's AI Max implementation delivered a 2x increase in conversion rates, a 31% reduction in cost per conversion, a 67% lift in click-through rates, a 27% increase in conversion value, and a 20% boost in return on ad spend. Their glycolic products specifically saw double the conversions, while serums achieved a 70% increase. These results were validated by Google's published case study based on L'Oreal's internal data from 2025.
How did L'Oreal use AI Max for Search?
L'Oreal activated three core AI Max features on their existing Search campaigns: search term matching to find new relevant queries beyond their keyword lists, text customisation to dynamically adapt ad copy to each user's search context, and final URL expansion to direct users to the most relevant landing page. They started testing in Chile and scaled globally after validating the results.
What did the Nielsen study find about AI-powered advertising?
Nielsen analysed over 50,000 brand campaigns and more than 1 million performance campaigns in the US from July 2022 to June 2024. They found that AI-powered Broad Match delivered 15% higher ROAS, Performance Max delivered 8% higher ROAS than Search-only strategies, AI-powered YouTube campaigns delivered 17% higher ROAS than manual campaigns, and adding Demand Gen to the mix increased ROAS by 10% and sales effectiveness by 12%. The study used Marketing Mix Modelling to isolate the impact of AI from external factors.
Does AI Max work for small businesses, not just big brands like L'Oreal?
Yes. While L'Oreal's massive data volume amplified AI Max's effectiveness, the underlying technology works across all advertiser sizes. Google reports an average 14% conversion lift for all advertisers enabling AI Max, rising to 27% for accounts previously using mostly exact and phrase match keywords. Case studies from smaller companies like MyConnect (16% more leads at 13% lower CPA) and BioRender (208% increase in click-through rates) confirm that the benefits scale down to smaller budgets. The key requirements are accurate conversion tracking, quality landing page content, and sufficient conversion volume for the algorithm to learn from.
What is the difference between Google's AI and an independent AI like groas?
Google's AI tools like AI Max, Smart Bidding, and Performance Max optimise your campaigns within Google's platform. They are powerful and effective, but they are built by the same company that generates revenue when you spend more. groas adds an independent optimisation layer on top of Google's native AI. Its autonomous agents are aligned exclusively with your business objectives, monitoring for issues like PMax cannibalising Search traffic, ensuring AI Max's query expansion stays relevant, and coordinating your entire campaign ecosystem to maximise your profit rather than Google's revenue. Analysis shows groas clients see an average 15% additional performance gain beyond what Google's native AI delivers alone.
How does AI Max compare to Performance Max?
AI Max and Performance Max serve different but complementary roles. AI Max is a feature suite that enhances your existing Search campaigns, maintaining full search term visibility and control while adding AI-powered optimisation. Performance Max runs across all Google channels from a single campaign with more automation but less transparency. Nielsen's data showed that the highest performance came from using both strategically together, with AI Max handling Search intent and Performance Max handling cross-channel reach.
Is AI-driven advertising actually better than manual management?
The data is now overwhelming. Nielsen's study of over 1 million performance campaigns found that AI-powered solutions consistently outperformed manual campaigns across every metric measured, including ROAS and sales effectiveness. Google reports 14 to 27% conversion lifts from AI Max specifically. The advantages stem from the AI's ability to process thousands of real-time signals per auction, test millions of creative and targeting combinations simultaneously, and optimise continuously without fatigue or downtime. Manual management still has advantages in specific scenarios requiring strict message control or very low-volume accounts, but for the vast majority of advertisers, the performance gap now favours AI.
What budget do I need for AI-driven advertising to work?
There is no hard minimum, but Google's AI tools generally need sufficient conversion data to learn effectively. A common guideline is at least 30 to 50 conversions per month per campaign for Smart Bidding to optimise effectively. For most businesses, this translates to a monthly Google Ads budget of roughly $2,000 or more, though the exact number varies by industry and average order value. Below this threshold, the AI may not have enough signal to outperform thoughtful manual management. Above it, the performance advantage of AI compounds as budget and conversion volume increase.
What should I do first if I want to implement AI-driven advertising?
Start with three foundations. First, ensure your conversion tracking is accurate and comprehensive, including Enhanced Conversions if possible. Poor data leads to poor AI decisions regardless of the tools you use. Second, audit your landing page content. AI Max's text customisation and URL expansion work best when your website has detailed, relevant product or service pages. Third, run an AI Max experiment on a single Search campaign for four to six weeks to establish a baseline comparison. If the results are positive, scale gradually across your account. For advertisers who want the full benefit of AI-driven management without the operational overhead, groas handles all of this autonomously from day one.
