Google Ads recommendations: how to use them without losing control of your campaigns

Google Ads recommendations are suggestions created by the platform to help improve campaign performance. They appear in your account automatically, based on your campaign data and patterns from similar advertisers, and they cover everything from keyword expansion to bid strategy shifts to conversion tracking setup.

Every account has them. Most advertisers have come to terms that at least a few without fully understanding what they do. Some have regret about it.

This article explains what Google Ads recommendations actually are, how the system works, what the new AI-powered suggestions mean for your account, and how to build a review process that improves performance without handing over control to an algorithm you did not ask to manage your budget.

Whether you run your own campaigns or manage advertising for clients, getting this right has a direct impact on your return on investment. The recommendations in the Google Ads interface are not neutral. Some will help you achieve your goals. Others will lead you away from them. Knowing how to spot the difference is the skill worth developing.

If you have tried applying recommendations without a structured process, you will recognise the experience. Metrics shift in ways that are hard to attribute. Changes you did not deliberately make show up in the account. Automated suggestions that looked reasonable at first turn out to have moved spend in directions you did not intend.

If you run Google Shopping or Performance Max campaigns, this is especially relevant. The recommendation engine is increasingly aggressive in those campaign types, and the auto-apply settings are on by default for a growing number of accounts.

‍

What are Google Ads recommendations?

Google Ads recommendations are personalised suggestions that appear inside your account and are designed to help you improve campaign performance. They live in the Recommendations tab, and they also now surface inside other areas of the platform, including the Conversion Summary dashboard.

Each recommendation is based on your account history, your campaign settings, and data from advertisers running similar campaigns. Google uses past performance signals, current bid strategy behaviour, and auction trends to create suggestions tailored to your specific goals. Every datum it collects feeds into a model that continues to refine what it surfaces to you over time.

The system is built to support advertisers in improving their ROI. That framing is accurate in the narrow sense that some recommendations do improve results. It is incomplete in the sense that it does not tell you which ones, or under what conditions.

The stated aim is to increase conversions, improve your return on investment, and make better use of your ad spend. In some cases that is true. In other cases, the recommendations are designed to increase your spend rather than improve your results.

The recommendations panel gives each account an optimisation score. According to Google's official documentation on the optimisation score, this is a number between 0 and 100 that represents how closely your account settings match Google's suggested configuration. Allowing all recommendations would push your score to 100. That does not mean your campaigns would perform better. It means you would be running the account the way Google wants you to run it.

Understanding that distinction is the starting point for using recommendations well.

‍

How the recommendation engine works

Google creates recommendations by analysing your account data and comparing it against patterns in similar campaigns. The system looks at your keyword list, your bid strategy, your budget, your audience targeting, your conversion tracking setup, and your ad types.

It then identifies gaps between your current setup and the configuration that, according to its models, tends to create better results across the user base it is drawing from.

The problem is that the aggregate data used to make these suggestions does not always match your specific business context. A recommendation to switch to broad match keywords might improve average click volume across thousands of accounts. It might also burn through your budget on irrelevant search queries within your account. The system cannot tell the difference without you reviewing it.

There are two ways recommendations are applied. Manual apply means you review each suggestion and choose to allow or dismiss it. Auto-apply means Google automatically implements adjustments without asking. Auto-apply is available for a wide range of recommendation types and is enabled by default in some accounts.

If you are not sure whether auto-apply is active in your account, check it now. Go to Recommendations, then Auto-apply, and review what is switched on. This is one of the most important things to audit when you take on a fresh client account or review an existing one.

‍

The new AI-powered recommendations in the Conversion Summary dashboard

Google Ads is now surfacing AI-powered conversion recommendations directly inside the Conversion Summary dashboard. These are not the same as the standard recommendations in the Recommendations tab. They are appearing in a fresh location and they are labelled specifically as "Recommended by Google AI."

The two examples currently visible are:

Create a Page view conversion for your website using Google Analytics. The suggestion tells you that your retail business will see improved results by measuring view item conversions in GA4.

Create an Engagement conversion for your website using Google Analytics. The same framing is used, with a focus on measuring engagement via Google Analytics 4.

Both recommendations push you toward setting up additional conversion tracking actions and connecting them to your campaigns.

This matters for several reasons.

First, adding new conversion actions shifts what your bid strategy optimises toward. If you add a Page view conversion and your Smart Bidding strategy starts counting page views as conversions, your cost per conversion will drop significantly but your actual sales performance may not improve at all. The improvement in the numbers is cosmetic.

Second, these recommendations are appearing in a part of the platform that many advertisers and account managers check regularly. The placement is deliberate. By surfacing suggestions in the Conversion Summary dashboard rather than just the Recommendations tab, Google is increasing the chance that users will see and accept them without a full review.

Third, the "Recommended by Google AI" label adds authority to the suggestion. It implies that an intelligent system has reviewed your specific data and identified a genuine gap. That may or may not be true. The label is a design choice as much as it is a factual statement.

This does not mean the recommendations are bad. Adding proper conversion tracking, including view item and engagement events via GA4, is genuinely useful for many accounts. The issue is whether you are adding them because they serve your measurement goals or because a prompt appeared in your dashboard and you clicked Set up conversion.

‍

Which recommendations are worth accepting

Not all Google Ads recommendations are low quality. Some are straightforwardly good and worth applying. The challenge is knowing which is which.

The following types of recommendations tend to be safe to accept in most accounts.

Fixing conversion tracking issues. If Google flags that your conversion tracking is broken, misconfigured, or recording zero conversions, that is worth investigating immediately. Accurate conversion data is the foundation of everything else. Without it, your bid strategy is working blind.

Adding audience segments to observation mode. Adding audiences in observation mode does not shift how your ads are targeted. It only gives you data on how different audience types perform within your existing targeting. That information is useful and the risk is low.

Updating disapproved ads. If ads are disapproved due to policy violations or outdated content, fixing them is correct. Do not leave disapproved ads in your account.

Adding ad extensions that are missing. Sitelinks, callouts, and structured snippets add information to your ads and can improve click-through rate. If your campaign is missing basic extensions, adding them is reasonable.

The following types of recommendations require careful review before accepting.

Switching to broad match keywords. This is the recommendation Google pushes most aggressively. Broad match allows your ads to show for a much wider range of search queries. For accounts with strong Smart Bidding data and well-structured negative keyword lists, broad match can work. For accounts that are still building conversion history or have limited budgets, it often results in wasted spend on irrelevant queries.

Raising your budget. Google frequently recommends increasing your budget when campaigns are hitting their daily cap. Whether this is a good idea depends entirely on your current return on ad spend and whether the additional spend would be profitable. The recommendation does not tell you that. You have to calculate it yourself.

Adding new keywords. Suggested keywords are created based on your existing content and search trends. Some will be relevant. Many will not. Review each one individually before adding it to your campaign.

Switching bid strategies. Recommendations to move from manual CPC to Target CPA or Target ROAS can improve performance in mature campaigns. In new campaigns or campaigns with limited conversion data, they can cause instability. The general rule is that Smart Bidding needs at least 30 to 50 conversions per month to perform well. If you are below that, be cautious.

‍

The optimisation score: what it measures and what it does not

The optimisation score is a number Google uses to tell you how fully you have implemented its recommendations. A score of 100 means you have accepted everything. A score of 60 means you have dismissed or ignored 40 percent of the suggestions.

The score is not a measure of campaign performance. It is a measure of recommendation adoption. Google's own Help Centre notes that each recommendation shows the estimated score uplift you would receive from applying it — reinforcing that the score tracks adoption, not results.

Google has positioned the optimisation score as a proxy for account health, and many agency and management tools surface it prominently. Some customers ask about it. Some performance reviews include it. The framing encourages you to treat it as something to improve.

The risk is that optimising your score by accepting recommendations is not the same as optimising your campaigns for return on investment. The two can move in the same direction, but they often do not.

A more useful number to track is your actual conversion performance relative to your target CPA or ROAS. If your campaigns are hitting their goals and your ad spend is creating profitable returns, a low optimisation score is not a problem.

If you click dismiss on a recommendation, Google gives you the option to provide a reason. Using this feature is good practice. It helps you maintain a record of decisions you have made deliberately, and in accounts with multiple users, it signals to other team members that a suggestion was reviewed rather than overlooked.

‍

How to build a recommendation review process

If you manage one account, reviewing recommendations manually each week is straightforward. If you manage multiple customer accounts, you need a process that is consistent and scalable.

Here is a simple framework for reviewing Google Ads recommendations without either ignoring them or accepting them uncritically.

1. Turn off auto-apply for high-risk recommendation types. The categories that should never be on it include broad match keyword expansion, bid strategy changes, budget increases, and anything that affects conversion tracking. Keep automatic apply only for low-risk updates like fixing ad policy issues or updating outdated assets.
2. Set a weekly review cadence. Go to the Recommendations tab once a week and review new suggestions. Do not review them daily because changes need time to affect the data before you can evaluate them. Do not let them accumulate for longer than a week because time-sensitive recommendations, like flagged tracking issues, need a faster response.
3. Categorise each recommendation before deciding. Ask three questions. Does this align with the current goal for this campaign? Do I have enough data to judge whether this change would improve results? Is there a risk that accepting this changes something I have deliberately configured?
4. Dismiss rather than ignore. If you are not going to accept a recommendation, dismiss it and click on a reason. This keeps your recommendations panel clean and gives you a decision log over time.
5. Test before you apply recommendations across all campaigns. If a suggestion looks plausible but you are uncertain, try it on one campaign first and measure the result over two to four weeks before rolling it out more broadly. Broad match keyword expansion is a good example of something worth testing on a contained budget before applying across your account.

A useful post-test question to ask: did the people clicking these ads convert at the same rate as before, or did the shift bring in more volume with worse outcomes? Volume and performance are not the same thing.

‍

Google Ads recommendations and Performance Max

Performance Max campaigns offer their own set of recommendations, and they deserve specific attention.

PMax is already a campaign type that gives Google significant control over targeting, bidding, and placement. The recommendations for PMax campaigns often push you to increase your asset variety, add audience signals, or raise your budget. Some of these are useful.

The asset quality recommendations are generally worth following. More high-quality creative assets give the algorithm more to work with, and that tends to improve performance over time. Adding audience signals helps the system start optimising faster, particularly in new campaigns.

Budget increase recommendations in PMax require the same scrutiny as in any other campaign type. Do not accept them without confirming that your current ROAS or CPA is hitting its target. If the campaign is already profitable at its current budget, scaling spend is worth considering. If it is not, increasing budget will only double more of the same unprofitable performance.

One of the most effective ways to reduce this kind of waste before it starts is to segment your catalog using custom labels in Google Merchant Center. Assigning profitability tiers to your products before PMax launches means the algorithm works from structured data rather than spending its learning budget across your full catalog indiscriminately.

The conversion tracking recommendations appearing in the Conversion Summary dashboard are particularly relevant to PMax. Performance Max optimises heavily toward conversions, so adding low-value conversion actions, like page visitors or interaction events, to a PMax campaign can distort its optimisation target significantly. If you add these events, make sure they are set as secondary conversions rather than primary ones, or exclude them from Smart Bidding optimisation entirely.

‍

Personalised recommendations and privacy

Google generates personalised recommendations based on data from your account. That data includes your conversion history, your search term reports, your audience performance, and signals from users who interact with your ads.

It also draws on aggregated data from other advertisers. The system uses p4atterns across a large population of accounts to inform the suggestions it surfaces to you.

This raises a question that is worth being aware of: the recommendations are not neutral. They are generated by a system that has its own goals, which include increasing engagement with the platform and growing overall ad spend. That does not make every recommendation wrong. It does mean you should evaluate and include each one on its own merit rather than treating the label "Recommended by Google AI" as a sufficient reason to accept it.

From a privacy standpoint, the data used to generate recommendations is governed by Google's privacy policies and the terms of your Google Ads account. If your clients have questions about what data is used to generate account-level recommendations, the information is available in Google's documentation, though it is not always easy to surface quickly.

What to tell clients about recommendations

If you manage Google Ads on behalf of customers, recommendations are a topic that comes up regularly. Clients sometimes see the Recommendations tab, notice the optimisation score, and ask why it is not 100.

The most useful thing you can tell them is that the optimisation score measures recommendation adoption, not campaign performance. A score of 70 with campaigns hitting their ROAS target is better than a score of 95 with campaigns missing it.

You can also explain that some recommendations are accepted as part of routine account management, some are dismissed because they conflict with the current strategy, and some are tested before a decision is made. That framing positions you as someone who reviews and makes decisions rather than someone who either ignores or follows the algorithm.

If a client wants to review recommendations themselves, walk them through the difference between manual apply and auto-apply, and make sure auto-apply is configured appropriately before giving them access to the account. If they are also evaluating tools to help manage this layer of the workflow, our overview of ProductHero alternatives covers several options that add a structured profitability layer on top of native Google Ads management.

‍

A note on the Conversion Summary dashboard recommendations

The new placement of AI-powered recommendations inside the Conversion Summary dashboard is a design change worth monitoring. Google has a pattern of introducing suggestions in new locations across the platform over time, and each placement increases the chance that users will interact with them without a full review.

The two recommendations visible in the current rollout, Page view conversions and engagement conversions via GA4, are not inherently bad suggestions. For an e-commerce account that is only tracking purchase conversions, adding view item and engagement tracking can provide useful signal for audience building and reporting.

The key is to add them correctly. Set them as secondary conversion actions. Do not let your primary bid strategy optimise toward page visitors if your goal is revenue. Use them for insight, not for Smart Bidding fuel.

If you are setting up a new account or reviewing an existing one, the Conversion Summary dashboard is now a place to check as part of your standard audit. Look at what recommendations are being surfaced there, assess whether they are appropriate for the account's current goals, and make a deliberate decision about each one rather than accepting or ignoring them by default.

‍

Frequently asked questions

What is auto-apply in Google Ads? It is a setting that allows Google to implement recommendations in your account automatically, without you manually reviewing or approving each change. It is available for a wide range of recommendation types and is enabled by default in some accounts. You can review and turn it off by going to the Recommendations tab and clicking it.

How does the optimisation score differ from performance metrics? The optimisation score measures how closely your account settings match Google's recommended configuration. It has no direct relationship to campaign performance. An account with a score of 65 that is hitting its ROAS target is performing better than an account with a score of 95 that is not. Track your actual conversion data and cost-per-acquisition against your goals, not the optimisation score.

Are AI-powered recommendations more reliable than standard ones? Not necessarily. The "Recommended by Google AI" label means the suggestion was generated or surfaced using Google's machine learning systems. It does not mean the recommendation is appropriate for your specific account. Apply the same review process to AI-powered recommendations as you would to any other suggestion.

Should I always dismiss budget increase recommendations? Not always. If your campaign is already hitting its ROAS or CPA target and is limited by budget, scaling spend can be profitable. The problem is that Google recommends budget increases regardless of current performance. Before accepting, calculate whether your current return on spend justifies the additional investment.

What happens if I dismiss a recommendation? Dismissed recommendations are removed from your active panel. Google prompts you to click on a reason for dismissing, which is worth doing. It creates a record of deliberate decisions, and in multi-user accounts it signals to colleagues that the suggestion was reviewed rather than missed.

‍

Final thoughts

Google Ads recommendations are a useful feature when you treat them as suggestions to review rather than instructions to follow. The system surfaces real patterns from account data and can flag genuine issues. It also pushes changes that serve Google's business model, and it uses increasingly prominent placement to increase the chance you will accept them without full consideration.

The optimisation score is not a performance metric. Auto-apply is not a management strategy. And "Recommended by Google AI" is not a reason on its own to set up a new conversion action.

Build a review process, understand what each recommendation changes, and make decisions based on your campaign goals and your client's definition of success. That is how you use recommendations in Google Ads well.

‍