In Google Ads, automation does not end with Smart Bidding. According to Google documentation and materials, the system can optimize bids in real time, but advertiser control still comes down to setting goals, budgets and rules for excluding queries or keywords.
In practice it's best to treat AI as an execution layer, not a strategic decision-maker. That means the algorithm can manage the bid for each auction individually, while a human defines when a campaign should stop buying traffic, which phrases should be excluded and how aggressively the budget may be shifted.
According to Google, Smart Bidding uses machine learning and analyzes signals available at auction time, not only historical cost per click. That distinguishes it from manual CPC, where the advertiser enters their own bid, and from enhanced CPC, which only adjusts manual bids up or down.
If stability is the objective, separate goal setting from execution. Industry practice recommends changing tCPA and tROAS gradually rather than in large jumps, and leaving a safety margin relative to recent campaign results. Google Blog's new tools reinforce this logic by allowing tests of different budgets and ROI targets in a single A/B test for Search.
The most useful automation for protecting budget concerns exclusions. According to Google Ads automation materials, scripts or rules can detect queries without conversions, add exclusions, pause weak keywords and react to a rise in cost per click above a set threshold.
Articles about AI Max emphasize that with broader match types and automatically generated ad assets, regular checks of the search terms report, brand exclusions and conversion data quality become more important. According to Google Blog and industry materials, AI Max should be tested on a control campaign with brand and geographic constraints maintained.
The safest model is a layered rollout: first native automation in Google Ads, then scripts, and only later broader rules or external AI logic. Automation sources recommend enabling low-risk actions first, such as detecting losing queries and anomaly alerts, before allowing the system to adjust bids and budget.
In practice this also means manually approving actions that could change account direction: new keywords, raising budget, changing bidding strategy or disabling valuable exclusions. This division of roles aligns with Google's approach to testing AI Max and with Smart Bidding materials, where the algorithm optimizes the auction but does not replace quality control processes.
Lub System helps B2B companies implement AI, automation and IT solutions end-to-end - from strategy to deployment. See our services or get in touch to discuss your case.
Source: https://blog.google/products/ads-commerce/ai-max-testing-planning-tools/