Google Ads attribution models: which one to choose in 2026

The attribution model decides which ad Google credits for a conversion. It sounds like a technical detail for specialists, but in practice it's what determines where you shift budget, which keywords you keep live, and which campaigns you declare "failures." Change the model and the numbers change right in front of you, without a single sale actually being different.

In 2026 the choice is much simpler than it was two years ago, because Google has cut nearly all the options. But "simpler to choose" doesn't mean "easier to understand." In this guide we explain what's left, what's gone, and above all how to read attribution without making the wrong budget calls.

What an attribution model actually is

Picture a customer buying from you. Before reaching the conversion, they clicked a Shopping ad, searched your brand a week later and clicked a Search ad, then saw a Demand Gen ad on YouTube and finally converted. Three or four different touchpoints, one single sale.

The question is: which of those clicks "produced" the sale? The attribution model is the rule Google uses to answer. It's not an objective truth, it's a convention. And different conventions tell you different stories about the exact same purchase journey.

This matters because Google's smart bidding (target CPA, target ROAS, maximize conversions) optimizes based on the attribution model you've set. If the model gives all the credit to the last click, the algorithm will learn to push the final touchpoints (typically brand searches) and neglect the ones that open up discovery. This is where attribution stops being reporting and becomes a strategic lever.

What's changed: only two models remain in 2026

Until 2023, Google Ads offered six models: last click, first click, linear, time decay, position-based, and data-driven. Today only two remain.

Google retired first click, linear, time decay, and position-based between 2023 and 2024, automatically migrating any conversion actions still using them to data-driven. The stated reason: those models were used for less than 3% of conversions. They were rigid, desk-decided rules that ignored the real data in your account.

Here's the current picture.

There's another important change: the minimum data requirement for enabling data-driven has disappeared. It used to take a threshold of conversions and clicks (roughly hundreds per month) for the model to "learn." Today any account can use it, even with low volumes. This has made data-driven the default option and the recommended choice for the vast majority of advertisers.

Last-click: simple, convenient, and often misleading

The last-click model is intuitive: whoever closed the deal wins. It's how most people instinctively think about sales. And that very intuitiveness is what makes it dangerous.

The problem is that last click systematically inflates what sits at the bottom of the funnel and starves what sits at the top. Brand searches, remarketing, ads that catch a user who's already decided: all of it takes 100% of the credit. The campaigns that generate demand, that introduce your product to people who didn't know you, look useless because they "don't convert on the last click."

The practical risk is concrete: you cut discovery campaigns because they look unproductive, and a few weeks later brand conversions collapse too, because you shut off the tap that was filling the top of the funnel. It's a classic trap. Anyone who reasons only in last-click terms ends up rewarding the campaigns that harvest and punishing the ones that plant.

This isn't a flaw exclusive to Google Ads: it's a structural limit of any single-touchpoint reading. We've covered the topic more broadly in marketing attribution models and how to choose between them and in the limits of attribution and the real alternatives.

Data-driven: what it actually does under the hood

The data-driven model (DDA) doesn't apply a fixed rule. It uses Google's machine learning to analyze your account's conversion paths, compare converters against non-converters, and estimate how much each touchpoint actually mattered. Then it distributes credit fractionally.

A concrete example. A €200 conversion might break down like this: 0.4 attributed to the Shopping campaign that drove discovery, 0.25 to Demand Gen on YouTube, 0.35 to the final brand search. No one takes it all; everyone takes what they contributed. The total remains one conversion, but spread across whoever actually built it.

The decision-making upside is huge: you finally see that the discovery campaign, which looked like zero conversions on last-click, was actually responsible for 40% of the value. Your judgment changes, your budget changes, the result changes.

Watch out for DDA's honest limits.

• It's a black box. Google doesn't show you the exact weights or the model's logic: you have to trust the output.
• It only sees what happens inside Google. A touchpoint on Meta, an email, word of mouth: all invisible to Google Ads attribution.
• With very low volumes, the model has less signal to learn from, so its estimates are noisier (even though it's now technically available to everyone).

The messy middle: why the journey is never linear

Google itself coined the term "messy middle" to describe what happens between the moment a person feels a need and the moment they buy. It's not a tidy funnel: it's a messy back-and-forth between exploration (the user widens the options) and evaluation (they narrow them down), repeated dozens of times, across different channels, over days or weeks.

In this chaos, expecting a single interaction to explain the sale is unrealistic. A B2B customer may touch 8-12 points of contact before converting. That's why data-driven, imperfect as it is, is structurally better suited than last-click to represent a multi-touch journey: it distributes credit instead of handing it all to one designated winner.

If you want to understand how Google models this behavior and how to use it in your campaigns, we've dedicated a guide to the messy middle and the real purchase journey. It's also the same reason attribution should always be read alongside the rest of the system: the CRM knows things Google can't see, and vice versa. More on that shortly.

Data-driven or last-click: which to choose

In almost every case the answer is data-driven. It's the default, it's free, it requires less data than before, and it represents reality better. But there are situations worth thinking through.

A common mistake is switching models and panicking because "conversions have disappeared." They haven't disappeared, they've been redistributed. Moving from last-click to data-driven, conversions on brand campaigns drop on paper while conversions on discovery campaigns rise. That's exactly the correct behavior. The total stays roughly the same.

Before switching, though, make sure your tracking is solid. A sophisticated model on dirty data produces sophisticated but wrong decisions. If you haven't sorted out the basics yet, start with why conversion tracking is the real foundation and look into enhanced conversions to recover data that cookies lose. In 2026 first-party data is central to the conversation: also read our first-party data strategy for Google Ads.