Google Ads AI Bidding in 2026: Smart Bidding Strategies That Actually Work

Smart bidding is no longer optional in Google Ads. With manual CPC bidding being progressively deprecated for most campaign types and Performance Max campaigns now the default recommendation for new advertisers, understanding how Google's AI bidding algorithms actually work — and how to work with them rather than against them — has become a core competency for anyone managing paid search.

How Google's Smart Bidding Actually Works

Smart bidding is a subset of automated bidding strategies that uses machine learning to optimise for conversions or conversion value at every auction. Unlike rule-based bidding, which applies fixed logic, smart bidding evaluates dozens of contextual signals — device, location, time, audience, search query, browser, and more — to predict the likelihood of conversion for each individual auction and adjust bids accordingly.

The algorithm requires conversion data to function well. This is the point most advertisers misunderstand: smart bidding is not a set-and-forget solution. It is a data-hungry system that requires a minimum volume of conversion signals to make accurate predictions. Google recommends at least 30–50 conversions per month per campaign as a baseline, though more competitive and valuable conversion types require significantly more data.

The Four Smart Bidding Strategies and When to Use Each

Target CPA (tCPA)

Target CPA optimises bids to achieve as many conversions as possible at or below your specified cost per acquisition. This is the right strategy when you have a clear, stable value for a conversion — a lead, a sign-up, a sale — and your primary goal is volume at a controlled cost. It works well for lead generation campaigns with consistent offer economics. Pitfall: setting a tCPA that is too low too quickly will starve the algorithm of bid opportunities and actually reduce conversion volume.

Target ROAS (tROAS)

Target ROAS optimises toward a target return on ad spend, weighing not just conversion probability but expected conversion value. This is the right strategy for e-commerce campaigns where different products have different margins and values. It requires robust conversion value data — ideally dynamic values pulled from your actual transaction data. Avoid tROAS if your conversion values are largely static or estimated; in that case tCPA is more appropriate.

Maximise Conversions

Maximise Conversions simply tries to get as many conversions as possible within your budget, with no target constraint. This is useful in the early stages of a campaign when you need to build conversion history. Once you have sufficient data (typically 30+ conversions), transitioning to tCPA gives you more control over efficiency.

Maximise Conversion Value

Similar to Maximise Conversions but weighted toward higher-value conversions. Useful when conversion values vary significantly. Like Maximise Conversions, this is typically a transitional strategy to build data before moving to tROAS.

Performance Max: The Opportunity and the Risk

Performance Max (PMax) campaigns are Google's most automated campaign type, serving ads across Search, Shopping, Display, YouTube, Gmail, and Maps from a single campaign. Google heavily promotes PMax, and the data shows it can outperform manual campaigns — but only when fed with high-quality inputs: creative assets, audience signals, conversion data, and well-defined goals.

The common failure mode with PMax is treating it as truly autonomous. Accounts that simply launch PMax with minimal asset input, no audience signals, and poor conversion tracking frequently see budget consumed on low-intent placements. The algorithm is only as good as the data and creative it receives.

• Provide at least 15 images, 5 videos, and multiple headline/description combinations
• Upload customer match audiences as audience signals to guide the algorithm
• Set up conversion tracking with actual revenue values, not just binary conversion events
• Use brand exclusions to prevent PMax from cannibalising branded search traffic
• Monitor search term reports (where available) to identify wasted spend

The Data Quality Problem

The single biggest factor determining smart bidding performance is conversion data quality. With iOS privacy changes, increased use of ad blockers, and Google's own cookie deprecation timeline, measured conversions in-platform increasingly undercount actual conversions. This means the algorithm is optimising toward a partial signal.

The solutions are Enhanced Conversions (which match hashed customer data to Google accounts for better attribution), offline conversion imports (for businesses where sales close offline), and Server-Side Tagging (which significantly improves first-party data capture by moving tracking from the browser to the server). These are not optional enhancements in 2026 — they are prerequisites for competitive smart bidding performance.

Working With the Algorithm, Not Against It

The mindset shift required for smart bidding success is moving from control-based management to signal-based management. Manual bidding gave account managers granular control over individual keyword bids. Smart bidding removes that control but offers a far more powerful optimisation surface if fed the right signals. Your job as an account manager is not to override the algorithm — it is to give it the best possible data, creative, and constraints to work within.