Google Ads has been incorporating machine learning into its products for years. Smart Bidding has been available since at least 2017. Responsive Search Ads replaced expanded text ads as the default format. Performance Max campaigns arrived as a fully automated campaign type that distributes spend across all of Google's channels simultaneously. These changes are real, not marketing language. But the way they work, what they require to perform, and where they fall short are frequently misunderstood by small business owners who encounter them for the first time.
This article focuses on what automated features within Google Ads actually do technically, what inputs they need to work, and where a human (or external tool) still needs to make decisions that the algorithm cannot.
Smart Bidding: what it optimises and what it needs
Smart Bidding is a family of automated bid strategies: Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value, and Enhanced CPC. Each one uses Google's signals to set bids on your behalf at auction time, adjusting for factors like device, location, time of day, query context, and user behaviour patterns that manual bidding cannot process at the same granularity.
The critical requirement: Smart Bidding needs conversion data to function. Without a clearly defined conversion action tracking real business outcomes (form submissions, calls, purchases), Smart Bidding has no target to optimise toward. In accounts where conversion tracking is absent or misconfigured, Smart Bidding defaults to maximising clicks, which is not the same as maximising business outcomes. It spends your budget efficiently on volume, not on quality.
The second requirement is data volume. Target CPA, for example, performs best with at least 30-50 conversions per 30-day period in the bidding strategy. Below that threshold, the model does not have enough signal to distinguish patterns from noise. New campaigns with small budgets frequently do not meet this threshold, which is why the recommendation for new accounts is often to start with manual CPC bidding and transition to automated strategies only after sufficient conversion history exists.
Responsive Search Ads: what the testing actually looks like
RSAs contain up to 15 headlines and 4 descriptions. Google's system tests different combinations to identify which pairings produce the highest click-through rates in each auction context. Over time, the combinations with better performance data show more frequently. This is automated copy testing at a granularity that would be impractical to manage manually.
The common misconception is that Google is generating your ad copy. It is not. Google is selecting from the options you provide. The quality of your RSA outputs is bounded by the quality of your inputs. An RSA with 15 generic headlines will produce worse results than a well-crafted set of 8-10 specific headlines, because the optimisation is working with a weaker input set.
The "pinning" feature in RSAs lets you fix a specific headline to a specific position. Use this sparingly. When you pin headlines, you reduce the number of combinations the system can test, which limits the automation's ability to find the best-performing variants. The right use of pinning is for legal or regulatory requirements (disclosures that must appear in a specific position), not for aesthetic control over headline ordering.
Performance Max: the most automated and the least transparent
Performance Max (PMax) is the furthest extreme of Google's automation direction. You provide a set of assets (headlines, descriptions, images, videos, sitelinks), set a budget and a Smart Bidding strategy, and Google distributes your ads across Search, Display, YouTube, Gmail, and Maps automatically. The algorithm decides which inventory to buy, which audiences to target, and which creative combinations to serve in real time.
PMax can produce strong results for businesses with high conversion volumes and a broad enough target audience. For local service businesses on small budgets, the evidence is more mixed. Several specific issues arise. First, PMax does not allow you to exclude Display or YouTube placements directly, which means spend can flow to channels where your target audience is not actively looking for your service. Second, the search term transparency in PMax is lower than in standard search campaigns: you cannot see the full list of queries that triggered your ads. Third, because the campaign type automatically enters all available inventory, it can compete with your own search campaigns, distributing budget across channels without your ability to steer it toward the highest-converting ones.
The most defensible use case for PMax in a small local service business is as a supplement to existing, well-performing search campaigns, not as a replacement for them. If you understand which keywords convert well and have standard search campaigns working on those terms, adding a PMax campaign on top can expand reach. Running PMax as your only campaign type means losing most of the controllability that makes Google Ads manageable for a small budget.
Where automation genuinely helps and where it does not
Automation in Google Ads is genuinely useful for bid adjustment at auction time. No human can calculate the right bid for a specific auction considering hundreds of contextual signals simultaneously. For that task, the algorithm is better than manual management.
Automation does not help with strategy decisions. It cannot tell you which product or service to focus on, what your offer should be, how to price your services, or whether your landing page is converting well. It also cannot account for business context you have not provided. If you know that bookings on Friday afternoons convert at half the rate of weekday bookings because you tend to be less responsive on Friday afternoons, the algorithm does not know this unless you encode it through bid adjustments or campaign scheduling. It only knows what the data tells it.
The practical implication: business owners who treat Google's automation as a set-and-forget solution tend to be disappointed. The automation works on the inputs it receives. Improving those inputs, through better conversion tracking, cleaner keyword structure, more relevant landing pages, and clearer business logic in the campaign settings, is work that happens outside the algorithm and cannot be delegated to it.
What an external management tool adds to this
Tools that manage Google Ads on your behalf operate by connecting to the Google Ads API, reading account data, and making changes through that API in response to performance signals. The decisions being made and the rules being applied differ by tool. Some tools manage bid adjustments. Some manage keyword additions from search term data. Some generate performance reports and flag anomalies.
The thing no tool can fully replace is your judgment about what outcomes matter for your business. A tool can optimise toward minimising cost per conversion. It cannot tell you whether those conversions represent the type of work you actually want, at the margin you can sustain. That judgment sits with you. What a well-configured tool can do is take the mechanical execution of that judgment, the checking of search terms, the reviewing of bid adjustments, the monitoring of budget pacing, and handle it more consistently than a business owner who checks their campaigns twice a week between everything else they have to do.
Stop guessing. Start your 14-day free trial and let AI manage your Google Ads while you run your business.