Google Recommendations: The Complete Guide for Advertisers in 2026

Google's Recommendations tab is designed to help Google. Sometimes it helps you too. The hard part is knowing which is which before you click "Apply."

Every Google Ads account has an optimization score. Every advertiser sees a list of suggestions. The score goes up when you accept recommendations and down when you reject them. Simple. Except the math behind that score isn't built around your cost per acquisition, your lead quality, or your pipeline. It's built around whether you accepted the changes Google wants you to make.

That's the distinction nobody in Google's interface will say out loud. And it's the reason I audit accounts with 90%+ optimization scores and regularly find 20–30% of their spend is being wasted.

This post is the complete operator view on Google Recommendations: what they are, how the score actually works, which categories of recommendation to accept, which to reject, and why a score in the low 80s often means the account is healthier than one at 95%.

What Google Recommendations actually are

Google Recommendations is Google's automated suggestion engine inside the Ads interface. It scans your account daily and surfaces changes Google's systems think would improve performance. Each suggestion is tagged with an estimated impact — impressions, clicks, conversions, or "optimization score improvement."

There are dozens of recommendation types. They fall into broad buckets:

• Bidding — switch to Target CPA, switch to Maximize Conversions, adjust bid targets
• Budget — raise daily budget, remove campaign spend caps
• Keywords — add new keywords, change match types, remove underperforming keywords
• Ads and assets — add responsive search ad variations, add sitelinks, add callouts
• Targeting — add audience signals, expand locations, add similar audiences
• Campaign structure — create Performance Max campaigns, consolidate campaigns, restructure ad groups
• Auto-apply — a switch that lets Google implement future recommendations without asking

Each of these exists for a reason. Some are genuinely good for your account. Most are designed to expand spend.

How the optimization score is actually calculated

The score runs from 0% to 100%. Each active recommendation in your account is weighted — some move the needle a lot, some barely at all. Google doesn't publish the exact weights, and they change over time.

What's consistent: accepting a recommendation raises the score by the weight of that recommendation. Dismissing it also raises the score (you've "dealt with" it), but by less. Ignoring it leaves the score where it is.

Sounds fair. Until you notice the pattern in what gets weighted most heavily.

The highest-weighted recommendations are almost always the ones that expand your spend: match type broadening, budget increases, Performance Max creation, audience expansion, auto-apply activation. The lowest-weighted are things like "add a sitelink" that don't affect spend much either way.

A 100% optimization score is mathematically only achievable by accepting every spend-expanding recommendation Google offers you. That's not a coincidence. It's the business model.

The three categories of recommendations

After auditing a few hundred accounts, I sort every recommendation Google offers into one of three buckets.

Spend-expanding (the majority)

These recommendations make Google more money and sometimes make you more money. The correlation between the two is much weaker than Google's UI suggests.

Examples:

• Switch this keyword from phrase to broad match
• Raise daily budget by $X
• Create a Performance Max campaign
• Add these 47 new keywords (most of which are tangential)
• Switch from Manual CPC to Maximize Conversions or Target CPA
• Add similar audiences to targeting
• Enable auto-applied recommendations

Each of these changes, individually, might improve your account. What they do reliably is increase the amount you spend. Whether the increased spend is efficient is a separate question Google doesn't answer for you.

Truly efficient (the minority)

A smaller set of recommendations actually improves your account's efficiency without just pushing you to spend more.

Examples:

• Pause this ad with 0 conversions over 90 days
• Add this high-performing search term as an exact match keyword
• Remove these keywords that have accumulated 500 clicks and 0 conversions
• Add negative keywords for these irrelevant searches

These are the ones worth accepting. They're also the ones with the lowest optimization score weighting, because accepting them doesn't grow Google's revenue.

Neutral (the in-between)

A third group is genuinely context-dependent. Add a sitelink. Add a callout. Add an RSA variation. Whether these help depends on your account. They're safe to accept in most cases and don't meaningfully change spend.

The auto-apply trap

The most dangerous feature in the Google Ads interface is the auto-apply toggle. Once enabled, Google implements future recommendations automatically, without your review.

I have not, in twenty years of managing paid media, seen a vertical where auto-apply actually works for the advertiser. Every account I've taken over that had auto-apply enabled had problems directly traceable to changes Google made without asking.

What auto-apply typically does:

• Adds broad match keywords to existing ad groups
• Raises daily budgets when the system predicts you'd spend more
• Applies match type changes to existing keywords
• Adds "similar" audiences and locations
• Auto-generates and applies ad variants

Any one of these, individually, might be okay. Together, in an account nobody's watching, they spiral. Broad match keywords added to a tightly-structured campaign pull in traffic the account wasn't designed to handle. Budget increases pair with that expanded traffic and suddenly you're spending 40% more with a 20% higher cost per acquisition.

If you take one thing from this post: turn off every auto-apply setting in every account you run. I do this within the first five minutes of taking over any new engagement.

The tracking prerequisite Google doesn't talk about

Every Google recommendation — good, bad, or neutral — is evaluated against the conversion data in your account. That's how Google decides what "better" means when it suggests a change.

If your conversion data is wrong, Google's recommendations are wrong. Not because the algorithm is broken, but because it's optimizing against the wrong target.

The common failure modes:

• Every page view counted as a conversion. I see this more than I should. Someone set up the conversion tag to fire on page load of the contact page instead of on form submission. Google now thinks you have a 100% conversion rate. Every recommendation to raise budgets, broaden match types, and expand audiences looks justified because on paper you're converting everyone. In reality you're setting fire to money.
• Duplicate conversions firing. The same form submission fires the conversion tag twice because someone installed it in two places. Your reported conversion volume is 2x reality. Google optimizes aggressively toward the inflated number and your actual CPA is double what the dashboard says.
• Missing conversions. Thank-you page doesn't have the tag. Phone calls aren't tracked. Offline sales from lead forms never get imported back to Google Ads. Google sees a fraction of your real conversions and starves the campaigns that are actually working.
• Conversion actions misconfigured. Primary vs secondary conversion assignment is wrong. Google optimizes toward the wrong outcome — maybe the newsletter signup instead of the demo request.

Any one of these creates a false signal. Google's recommendations — and more importantly, your own judgment about whether an account is healthy — get calibrated to the false signal. You end up accepting recommendations that look like they're working and rejecting ones that look like they're not, when the underlying data was lying the whole time.

Before you touch the recommendations tab, verify:

1. Every conversion action in your account is firing on the correct event (form submit, purchase complete, phone call, etc.) — not page load, not page view
2. Each conversion fires once per event, not multiple times
3. Your reported conversion counts in Google Ads match your CRM or backend within a reasonable margin
4. Primary vs secondary conversion assignments reflect what actually matters to the business

Tracking work is unglamorous. It's also the only work that makes every other paid media decision defensible. This is marketing operations territory — the plumbing under the paid media work. Both posts are about the same underlying principle: if the infrastructure is broken, every decision on top of it is wrong.

Field notes: eighteen months in senior care

The clearest version of this story came from a former employer who asked me to take over paid media for a 65+ primary care network. When I started, they were running 33 clinics across Florida. When I left eighteen months later, we were running 136 clinics across five states. Monthly ad spend grew from $75K to $600K over that period.

The account had an average optimization score of 75% when I inherited it. Low traffic, low conversions, not enough lead volume to feed all the clinics. The previous team had been selectively accepting recommendations but was missing the strategic picture.

First moves:

• Match type triangulation — I stacked broad, phrase, and exact variants of priority keywords in the same ad groups to force Google's auction to serve the lowest-CPC qualifying match. I'll write about this technique in detail in a follow-up post.
• Turned off every auto-apply setting — none of the recommendations Google was silently applying had improved the account. Several had hurt it.
• Killed Performance Max — after $20K in testing, Performance Max was producing leads at roughly $150 each, which looked reasonable on the dashboard. The problem was downstream. The lead-to-appointment rate from PMax was under 0.5%, compared to 8–12% for properly structured Search campaigns. On a cost-per-actual-patient basis, PMax was the worst-performing channel in the account.
• Blocked geographies ruthlessly — we excluded every country except the US, every state we weren't operating in, and every county inside our active states where we had no clinics. A startling amount of budget had been flowing to clicks from countries that had nothing to do with the business.
• Restructured campaigns around clinic geography — one campaign per clinic with a 15-mile radius, allowing overlap between adjacent clinics, plus "state main bucket" campaigns covering every location in a given state.
• Banner ads paused, Display network disabled — both were bleeding money. We later added Search Partners slowly, one at a time, after validating they weren't dragging performance down.

The optimization score we actually optimized for

Google kept pushing us toward a 90%+ optimization score. We ignored it. When we did chase higher scores, our cost per acquisition went up and our daily budgets ran out by mid-afternoon — well before the evening rush when our 65+ audience was most likely to convert.

Our sweet spot was in the low 80s. Google's interface complained about it constantly. We ignored the warnings. The numbers agreed with us.

What the numbers did

Over eighteen months, with spend scaling 8x:

• Lead-to-appointment rate moved from 2–3% to roughly 15%
• Cost per lead moved from $200–$250 down to $100–$125
• Cost per actual patient moved from nearly $1,500 down to below $300

The CPL improvement was significant. The cost-per-patient improvement — the number that actually matters to the business — was dramatic because the quality of each lead improved along with the unit cost.

None of that happens in an account running at a 95% optimization score. The score and the business performance were actively pulling in opposite directions.

Which recommendations to accept

A short list. These are worth reviewing and usually accepting:

• Negative keyword suggestions from your own search term data — Google surfaces queries that are wasting budget. Skim the list weekly, add them as negatives.
• Pausing ads or keywords that have accumulated meaningful volume with no conversions — usually a real efficiency gain.
• Adding ad extensions (sitelinks, callouts, structured snippets) that match your business — low-effort, helps CTR, doesn't materially change spend.
• Adding an exact-match version of a search term that's already converting — moves traffic from broader matches to the more efficient exact variant.
• Bid adjustments for specific devices or demographics when the data clearly supports it — review the underlying data first.

Accept these. Move on.

Which recommendations to reject

The longer list. These are the ones that expand spend without a corresponding improvement in efficiency:

• Broad match expansion on existing keywords
• Budget increase recommendations (almost always)
• Auto-apply in any form, for any setting
• "Switch to Performance Max" for B2B, considered-purchase, or high-consideration categories
• "Add similar audiences" without a clear audience strategy
• Bid strategy changes from Manual to automated without a clean conversion dataset feeding the automation
• "Add new keywords" lists that expand beyond your core intent

For a deeper look at specific recommendation categories and why each one backfires, see Why Google's Recommendations Are Mostly Traps. That post covers the individual recommendation types in more detail.

Why a lower optimization score often means a healthier account

Put this next to each other:

• Account A: 95% optimization score. Cost per acquisition up 20% year over year. Leads arriving from outside the target geography despite geo-blocks. Performance Max eating 40% of budget and returning 0.5% lead-to-appointment. Budget running out by 3pm daily because every recommendation Google offered to raise it got accepted.
• Account B: 82% optimization score. Cost per acquisition down 40% year over year. Triangulation structuring in place. Auto-apply off. PMax off. Campaigns structured around actual business geography. Budget lasts through the full day.

I've seen both. Account B is healthier. Account A looks better in Google's UI.

If you're running a Google Ads account and your optimization score is above 90%, you're almost certainly accepting recommendations that are growing spend without a matching growth in pipeline. If it's in the 60–85% range, you're probably exercising judgment. If it's below 60%, you may have drifted into neglect — there are some recommendations worth accepting, and below a certain threshold the score is signaling missed opportunities rather than operator discipline.

The goal isn't a high score. The goal is to understand what each recommendation does and make the call consciously, one at a time.

The multi-platform blind spot

Even a perfectly run Google Ads account has a limit. Google recommendations — even the good ones — only tell you how to optimize within Google's platform. They tell you nothing about whether your next dollar should go to Google at all, versus Meta, LinkedIn, Microsoft, or anywhere else.

That multi-platform allocation question is structurally invisible to every platform's native recommendation engine. Meta doesn't know what your Google CPA is. Google doesn't know your LinkedIn cost per opportunity. Each platform optimizes locally, and nobody optimizes globally.

This is the gap that made me start building Campaign Budget Optimizer — a cross-platform budget allocation tool that connects to Google Ads, Meta, Microsoft Ads, LinkedIn, TikTok, and Google Analytics, and surfaces where the next marketing dollar should actually go. It launches in May 2026.

For now, the manual version of the same thinking is in Where should the next dollar go?.

What to do in your account this week

If you read this far and want to act on it, four moves to make in the next sitting:

1. Verify your conversion tracking first. Before you touch a single recommendation, confirm your conversions are firing correctly — on the right event, exactly once per event, and matching what your CRM reports. If tracking is wrong, every decision you make about recommendations is based on bad data. See the tracking prerequisite section above.
2. Open your Google Ads account. Go to Settings → Account Settings → Auto-apply recommendations. Turn every auto-apply setting off. This is the single highest-leverage five-minute action you can take once tracking is verified.
3. Go to the Recommendations tab. Filter by category. Look at what Google is suggesting for your account, grouped by type. Count how many fall into "spend-expanding" versus "truly efficient."
4. Check your current optimization score. If it's above 90%, plan an audit to figure out what you accepted to get there. If it's below 60%, plan an audit to figure out what opportunities you've been missing.

---

If you want a senior operator to look at your account, SEM consulting and marketing audits are both services I run. Engagement pricing is on the pricing page. First conversation is always free.