Automating Ad Bids on Google and Meta: Everything About Smart Bidding and AI

Smart Bidding is Google Ads' automated bidding system that calculates an optimal bid in real time at every individual ad auction, using signals such as device, location, time of day and search intent. Meta Ads has a comparable system: the Advantage+ bidding engine also optimises per auction, determining who sees your ad and what you pay for it. Advertisers who want to automate ad bids have everything they need with these two platforms to advertise structurally more efficiently than manual bidding ever allowed.

Key takeaways

• Smart Bidding and Meta's Advantage+ bidding optimise every bid in real time using dozens of signals simultaneously, something manual bidding cannot replicate.
• The two most used strategies are Target CPA for lead generation and Target ROAS for e-commerce.
• Automated bidding requires reliable conversion tracking as its foundation: without good data, the algorithm optimises towards the wrong goals.
• Smart Bidding's learning phase typically lasts two to four weeks; intervening during this period disrupts the learning process.
• AdBrains' own AI layer adds an extra optimisation cycle on top of platform algorithms, with daily tCPA/tROAS adjustments and server-side signal enrichment.

What exactly is Smart Bidding and how does it work?

Smart Bidding is the collective term for automated bidding strategies within Google Ads that rely on machine learning to maximise conversions or conversion value. Each time a user searches or visits a page where your ad might show, an auction runs. For every single one of those auctions, AI bid optimization in Google Ads recalculates what the optimal bid should be, drawing on far more signals than any human manager could realistically handle.

Google Ads Help (2026) lists the signals processed per auction: device type, operating system, location, day of the week, time of day, the actual search term and whether the user belongs to a remarketing audience. Put those together and you get something like this: for the same keyword, a user who already browsed your site can receive a meaningfully higher bid than someone who has never heard of you.

Where this departs from manual bidding is not just speed. Manual bidding reacts to historical averages; Smart Bidding works from real-time context. A warm Saturday in June will produce a different bid for an electric folding bike webshop like E-4motion.com than a Tuesday morning in November, even when the keyword is identical. That gap between average and contextual makes automation structurally more precise, not merely more convenient.

The four main Smart Bidding strategies on Google

Google offers four Smart Bidding strategies, each built around a different primary goal. Which one fits depends on your business type and how much conversion data you have to work with.

• Target CPA (tCPA): The algorithm aims for an average cost per conversion that you define. This works well for lead generation where leads carry roughly comparable value. Clima-Active.nl, which handles quote requests for air conditioning and heat pump installations, is a good example: every quote request has a similar commercial weight, so tCPA is a natural fit.
• Target ROAS (tROAS): Here the algorithm tries to maximise conversion value relative to spend, based on a return target you specify. It suits e-commerce businesses where order values differ from one transaction to the next. ToetsJeKennis.nl sells online exams and courses at an average order value of €50, and tROAS helps steer budget towards the higher-value course packages.
• Maximize Conversions: Gets you as many conversions as possible within a fixed budget, with no cost-per-conversion or ROAS target attached. Often the right choice during a campaign's start-up phase or when conversion data is still thin.
• Maximize Conversion Value: Works on the same principle as Maximize Conversions, but focuses on total conversion value rather than volume. Useful when order values vary a lot and you prefer not to commit to a fixed tROAS target.

A fifth option exists: Enhanced CPC (eCPC), which adjusts manual bids up or down based on estimated conversion probability. Google has been pulling back from eCPC for some time now, pushing advertisers toward fully automated strategies like tCPA and tROAS instead. This shift is reflected in the Google Ads Help product documentation (2026).

Automated bidding on Meta: how does Advantage+ bidding work?

Meta Ads has its own automated bidding system, functionally comparable to Smart Bidding but more deeply tied to the platform's placement and audience optimisation. When you set up a campaign on Meta, you pick an objective, such as sales, leads or reach, and Meta then applies Advantage+ placements and bidding by default unless you step in and override it manually.

There are two main bidding options. Cost per Result Goal works much like tCPA: if you set it to €15 per lead, the algorithm spreads your budget so the average across the campaign period stays close to that figure. ROAS Goal works along the same lines as tROAS. If you don't set an explicit goal at all, Meta falls back to Lowest Cost, which simply pushes spend toward wherever the cost per result happens to be lowest at any given moment.

Where Google Smart Bidding and Meta's system actually diverge is in the type of signal each one relies on. Google Ads steers primarily on search intent, what a person is typing right now. Meta steers on interest and behavioural signals: who someone is, what they've looked at, what they've bought before. That's a meaningful difference in practice. Both systems do perform best when they're fed rich, reliable conversion data through server-side tracking.

When does Smart Bidding underperform?

Smart Bidding delivers poor results in a number of specific situations. Recognising those situations is just as valuable as knowing when it does work well.

• Too little conversion data: Google recommends at least 30 conversions per campaign per month for tCPA, and preferably 50 or more for tROAS (Google Ads Help, 2026). Below that threshold, the algorithm has too little data to learn and bidding behaviour becomes erratic.
• Incorrect or incomplete conversion tracking: If the measured conversions do not match actual business results, Smart Bidding optimises towards the wrong goal. This is the most common pitfall we encounter in practice.
• Intervening too aggressively during the learning phase: Smart Bidding needs two to four weeks to calibrate. Changing the tCPA or tROAS target every week during that period repeatedly disrupts the learning process.
• Strong external fluctuations without seasonality adjustments: During promotions, sales or external events, you can use the seasonality adjustment feature in Google Ads to temporarily steer. Without it, the algorithm cannot explain the conversion spike.

Bidding strategies overview by campaign type and goal

The table shows that data minimums differ slightly between platforms and campaign types. Meta recommends higher weekly volumes than Google for stable algorithm performance, meaning smaller advertisers on Meta reach the limits of automated bidding faster.

How AdBrains AI takes Smart Bidding and Meta bidding further

Smart Bidding and Meta's bidding systems are genuinely powerful, but they are built to serve millions of advertisers at once. They have no visibility into your margins, your definition of a quality lead, or the fact that a campaign is temporarily thrown off by something outside your control. That is the gap AdBrains' AI technology addresses.

AdBrains built its own automatic tCPA/tROAS optimisation system that evaluates bidding strategies daily, per client. It compares the realised CPA or ROAS over the past seven days against the agreed target and adjusts settings when the deviation moves outside a predefined bandwidth. The distinction from what Google does natively matters here: Google optimises within the targets you have set, whereas AdBrains AI decides when those targets themselves need to move, taking into account seasonal patterns, the current campaign phase and the client's margin objectives.

Running in parallel, server-side signal enrichment adds depth to the conversion signals sent to both Google and Meta. Through AdBrains' own server-side Google Tag Manager infrastructure, every conversion is sent with first-party data attached, including order value, product category and customer segment. The result is that Smart Bidding and Meta's algorithm receive richer input than a standard implementation would ever provide, which feeds directly into the quality of the automated bid.

The multi-agent verification system sits upstream of any actual change. Before a tCPA or tROAS target is adjusted, four independent AI agents each review the proposed change. A temporary conversion spike from a one-off promotion, for instance, will not quietly translate into structurally wrong bidding settings because of this check.

The strategy-switch system handles a different risk: campaigns that are about to fall out of the learning phase because conversion volumes have dipped too low. Rather than letting the algorithm thrash on thin data, the system automatically pauses the campaign and reactivates it once the signal is reliable again. For clients such as LeroyBrouwer.nl and E-4motion.com, this means bidding strategies are reviewed not once a quarter but continuously, driven by daily data.

Conversion tracking as the foundation of automated bidding

No automated bidding system, on Google or Meta, performs better than the quality of the conversion data it receives. Browser-based conversion tracking via standard GTM implementations misses a growing share of conversions due to ad blockers, ITP and cookie restrictions. That means Smart Bidding is steering on incomplete data.

Enhanced Conversions and server-side tracking are the solutions Google and Meta themselves recommend. Enhanced Conversions matches first-party data (such as email addresses) with the user's Google account to attribute missed conversions. Combined with server-side tracking via sGTM, this provides a significantly more complete picture of conversion volume, which directly improves the quality of Smart Bidding's output.

FAQ: Smart Bidding and automated bidding

What is the difference between Target CPA and Target ROAS?

Target CPA optimises for a fixed cost per conversion, regardless of the value of that conversion. Target ROAS optimises for return on ad spend: the algorithm tries to maximise conversion value relative to budget spent. For businesses with a fixed lead value (such as Clima-Active.nl), tCPA makes more sense; for webshops with varying order values (such as E-4motion.com), tROAS provides better margin control.

How many conversions do I need before Smart Bidding works well?

Google Ads Help (2026) recommends at least 30 conversions per campaign per month for tCPA, and preferably 50 or more for tROAS. Below that threshold, consider running Maximize Conversions without a target first to build up data before switching to a target-based strategy.

Can I combine Smart Bidding with manual bid adjustments?

Only partially. Google has ignored or restricted most manual bid adjustments (such as device or location multipliers) for Smart Bidding campaigns, because the algorithm already processes those signals automatically. Setting manual multipliers on top of Smart Bidding gives the algorithm contradictory instructions and typically leads to worse results.

How do I know if my Smart Bidding strategy is performing well?

Compare the actual CPA or ROAS over at least four weeks against the set target. Pay attention to the Status column in Google Ads: campaigns with a "Limited" status (due to budget or low Quality Score) rarely hit their target regardless of the bidding strategy. Use the Bid Strategy Report in Google Ads for a detailed breakdown of how the algorithm is pursuing the target over time. Structurally large deviations typically indicate a data or structure problem rather than a fault in the bidding strategy itself.