Google Ads Tightens Lookalike Targeting: How to Maintain Control

Do you feel like you are gradually losing control over what you have built? That is precisely the feeling of thousands of Google Ads advertisers worldwide as they realize that one of their most powerful targeting tools is being fundamentally altered.

Starting in March 2026, Google officially transformed how Lookalike segments operate in Demand Gen campaigns. Instead of maintaining a strict targeting mechanism based on similarity thresholds, Google is turning them into "suggestions" for the AI system. This means: even if you choose "narrow" to limit your audience to the top 2.5% most similar users, the system can still cross that boundary if Google's AI thinks it will yield more conversions.

That sounds reasonable in the era of automation, doesn't it? But for advertisers who have spent years building strategies based on strict control, this is a shock. And in this article, we will dissect every aspect of this change – from the old mechanism, what is currently happening, to how you can retain control if you choose to.

The Old Lookalike Mechanism: When Boundaries Were Inviolable

To fully understand the significance of this change, we need to go back to the period before March 2026. At that time, Lookalike segments in Demand Gen acted as a hard barrier – an impassable boundary.

The process was clear: you upload a seed list from your first-party data – which could be a list of past purchasers, website visitors, or people who have interacted with your YouTube channel. Then, you choose one of three reach levels:

Level	Similarity Threshold	Meaning
Narrow	~2.5% of users	Targets users most similar to your seed list
Balanced	~5% of users	Balances reach and similarity
Broad	~10% of users	Expands reach, accepting a lower level of similarity

The crucial point is: this boundary was absolute. If you selected "narrow," Google would only target approximately 2.5% of users in the target geographic area who shared the most characteristics with your seed list. The system could not and was not allowed to exceed that threshold, even if the AI thought other users outside that threshold also had high conversion potential.

For many advertisers, especially those operating in industries with strict target audience requirements – such as finance, healthcare, or specialized B2B – this control was a prerequisite. They needed to ensure their ads only reached genuinely relevant people without wasting budget on unrelated audiences.

Another important detail: the seed list required a minimum of more than 100 matching users for the segment to function. Segments were refreshed automatically every 1-2 days based on existing customer data. And you could combine up to 10 seed lists in a single segment, including customers, website visitors, app users, and YouTube interactions.

The Shift from March 2026: When AI Makes the Decisions

So what has changed? Basically, everything.

Starting in March 2026, Google officially transitioned Lookalike segments to suggestion mode. According to Google's official documentation: "Lookalike segments in Demand Gen will evolve from being restricted by similarity thresholds into audience suggestions."

Simply put: the reach slider (narrow, balanced, broad) is no longer a hard barrier. Instead, it is just a directional signal. Google's AI system will use your seed list and chosen reach level to prioritize users, but it can exceed those parameters if Google's model predicts that expanding beyond the threshold will yield better conversion performance or a lower cost per action (CPA).

Think about this in a real-world context. You selected "narrow" because you wanted to target a small group of people with interests and behaviors similar to your existing customers. You believed that reaching a small but high-quality group would be more effective than spreading your budget thin across a larger, less relevant group.

But now, Google AI might decide: "Hmm, I see these users outside the 2.5% threshold have a high conversion probability. I'm going to target them." And you – the person paying for the campaign – have no veto power.

For Google, this is a logical step in its AI-driven automation and optimization strategy. According to their documentation, this change aims to: optimize performance (AI can optimize for specific goals like conversions or CPA), expand reach and inventory (the system can reach high-quality users that rigid similarity filters previously missed), and improve conversion efficiency compared to the previous version.

However, for advertisers, this is a trade-off between control and potential performance. And not everyone is ready to accept that trade-off.

Why Is Google Doing This?

Behind this change lies a philosophy that has shaped Google's product strategy for years: AI understands humans better than humans do.

Google believes its AI models are capable of predicting user behavior better than any manual filter. When you limit the system to a rigid similarity threshold, you are holding back the AI's potential. You are telling Google: "I know my customers better than you do."

And in some cases, that is true. But in many other cases, Google can prove the opposite.

This philosophy has been clearly demonstrated through products like Performance Max – where Google controls almost the entire ad delivery process, relying solely on conversion goals set by the advertiser. And with Demand Gen, Google is gradually applying the same mindset: AI leads the way, humans set the goals.

However, there is an important difference. Performance Max was designed from the beginning as a fully automated product. Demand Gen, conversely, was promoted as a solution that balances automation with control – where advertisers still have a say in deciding their target audience. Shifting Lookalikes to suggestion mode erodes that promise.

What Can You Do? The Opt-Out Form – A Narrow Door Still Open

The good news is: Google is not forcing all advertisers to accept this change immediately. There is still a way to maintain the previous behavior of Lookalike segments.

However, this method is not easy. To keep Lookalike behavior functioning as a Targeting Constraint, you must submit an opt-out form to Google. According to Google's support documentation: "If you want to keep Lookalike behavior consistent with the previous version, you can opt out of this change by signing up through this form."

The processing time announced by Google is approximately 3 days to 1 week. During that time, you still face the risk that the system has already begun applying the new model.

It is worth noting that Google stated direct opt-out options in the user interface (UI) will only be launched in late 2026. This means for several months, the only path to retaining control is through a form – a manual, slow, and unfriendly process.

And even if you successfully opt out, you still need to monitor closely to ensure the old behavior is genuinely maintained. Because in the world of Google Ads, everything can change.

Strategy for Advertisers: 5 Action Steps Right Now

So what should you do? Here are the practical steps I recommend taking right now:

1. Check if Your Demand Gen Campaigns Are Using Lookalike Segments

This is the first and most important step. Log into Google Ads, check each Demand Gen campaign, and identify if any ad groups are using Lookalike segments. If so, assess how critical strict control is to the performance of that campaign.

2. Identify Campaigns That Need to Keep the Old Behavior

Not all campaigns need to opt out. If you are running a test campaign with the goal of discovering new audiences, letting the AI freely expand might be beneficial. But if you have a campaign optimized for a specific audience group with strict lead quality requirements, keeping control is mandatory.

3. Submit the Opt-Out Form for Necessary Campaigns

For campaigns where you decide to keep the previous behavior, access the opt-out form provided by Google and submit your request immediately. Do not hesitate, as processing time can take up to a week.

4. Monitor CPA and Lead Quality in the First 1-2 Months

After applying any changes, closely track your performance metrics. Compare CPA, conversion rates, and lead quality before and after the change. If you notice a significant drop, reconsider your strategy.

5. Maintain High-Quality Seed Lists

Whether you choose to opt out or accept the new model, the quality of your seed lists remains a key factor. Ensure your customer lists are constantly updated with the highest-value users – those who truly drive revenue and profit for your business.

When Control Becomes a Precious Asset

I have witnessed the digital advertising industry go through many revolutions. From the days when we could control every minute parameter, to the era where AI takes the lead. Each time, the central question remains: Who is actually in control?

The change in Google's Lookalike targeting is not just a technical update. It is a philosophical statement: Google believes its AI understands your customers better than you do. And they are gradually narrowing the space for you to prove otherwise.

But the important thing is: you still have a choice. At least for now. The opt-out form still exists. Control has not been entirely stripped away.

The question is: How will you use that power? Will you accept the change and trust Google's AI? Or will you fight to retain what you have built?

There is no right or wrong answer. There is only the answer that fits your strategy and goals.

Have you ever encountered a situation where Google arbitrarily changed targeting mechanisms without asking your opinion? What do you think about Google's philosophy that "AI knows humans better than humans do" – is it an inevitable progress of automation, or an encroachment on the control you deserve as the person paying for the campaign? If you are an advertiser running Demand Gen, will you choose to opt out to keep control, or accept the new model and let AI lead the way? Share your perspective below, because in this advertising game, whoever understands the rules best wins!