Google Ads Optimization Score: What to Apply, What to Ignore, and Why Auto-Apply Can Quietly Cost You

Every Google Ads account has a small orange or green circle sitting near the top of the screen with a number inside it, and a lightbulb icon next to a running count of “recommendations” waiting for a decision. For a lot of advertisers, that number becomes a low-grade source of guilt. You log in to check yesterday’s spend, and the interface quietly reminds you that your account is only performing at “68% of its potential.” It feels like a report card, sitting right next to the metrics that actually matter, styled the same way, colored the same way, asking for the same kind of attention.

It is not a report card. Treating it like one is how a lot of accounts end up with broader match types, higher budgets, and looser bid targets than the person running them ever actually decided on — not because those changes were wrong, necessarily, but because they were accepted for the wrong reason. This is a practical look at what Google Ads Optimization Score actually measures, why the score and your account’s actual performance are two different things that sometimes move in opposite directions, which recommendation categories are worth your time, and why the “auto-apply” setting sitting quietly in your account settings deserves a lot more scrutiny than most advertisers give it.

What Optimization Score Actually Measures

Google’s own documentation is fairly direct about this, even if the product design sends a different message. Optimization score is an estimate of how well an account is set up to perform, expressed as a percentage from 0% to 100%, where 100% means the account can perform at its “full potential.” It’s recalculated in real time based on your account’s statistics, settings, campaign status, the recommendations currently available to you, and your recent history of applying or dismissing recommendations.

The mechanism behind the number is a weighted sum. Each available recommendation carries an estimated impact value, shown to you as a percentage-point gain (“+2.4% if applied”), and those individual impacts roll up into the overall score. Dismiss a recommendation and its weight drops out of the calculation; apply one and the score moves up by roughly that amount. This is the detail that matters most: the score is a measurement of how many of Google’s suggestions you’ve acted on, not a measurement of how your campaigns are actually performing. An account with flat conversion volume and a rising cost per lead can still carry a high optimization score, because the score has no opinion on your actual conversion data. It only knows whether you’ve applied what it suggested.

That distinction gets lost easily because the presentation is built to look like a health score, complete with a colored ring and a completion percentage, sitting in the same visual language as things that genuinely are performance indicators, like Quality Score or impression share. If you haven’t already, it’s worth reading how Quality Score is calculated and what it actually reflects, because the two metrics get confused constantly and they measure genuinely different things — one estimates ad and landing page relevance against what people are searching for, the other estimates recommendation adoption.

The Blind Spots Built Into the Score

A score built from “did you apply this suggestion” has a few structural blind spots worth naming directly, because they explain most of the cases where a high score and a struggling account coexist.

The first is conversion tracking accuracy. Optimization score has no way to check whether the conversions feeding your bid strategy are actually the ones that matter to your business. If a form-fill is tracked twice, if a phone call gets logged as a conversion regardless of call length, or if a purchase confirmation page fires on page refresh, the score doesn’t know and doesn’t care — it will happily recommend leaning harder into a bid strategy that’s optimizing toward broken data. That’s a separate problem from optimization score entirely, but it’s worth ruling out before accepting any bidding or budget recommendation, since a bid strategy change built on top of conversion numbers that don’t actually add up tends to make a tracking problem more expensive rather than less.

The second blind spot is account maturity. A brand-new account with three weeks of data gets recommendations calibrated the same way as a five-year-old account with a long, clean performance history, even though the newer account has far less evidence to justify aggressive automation. Smart Bidding strategies in particular need a reasonable volume of recent conversions to calibrate against, and a recommendation to switch to Target ROAS doesn’t pause to check whether you actually have that volume yet.

The third is business context. The score has no concept of margin, seasonality, inventory constraints, sales capacity, or the fact that you deliberately want to stay small in a geographic area because your delivery radius is limited. It treats “more volume” as inherently good, because more volume is usually what pushes the score up. A recommendation engine that can’t see your P&L is not in a position to tell you your budget is too conservative.

Seasonality is a good example of how this plays out in practice. A recommendation to raise budgets or loosen a Target ROAS generated during your slow season looks identical to the same recommendation generated during your peak season, because the algorithm is reading recent performance trends, not your calendar. A landscaping company scaling back deliberately in December doesn’t need a nudge to spend more just because a competitor’s seasonal surge is pushing average auction dynamics around; it needs someone who knows the business is intentionally quiet right now, which is exactly the kind of context the score was never built to hold.

The Recommendations Feed Isn’t Neutral

Recommendations arrive sorted into a handful of categories, and they are not equally trustworthy. It helps to think about them in three rough buckets.

The first bucket is genuine repairs: conflicting negative keywords that are blocking your own ads, a broken sitelink, an ad group with zero active ads, a disapproved asset, a tracking tag that stopped firing. These are close to free wins. They don’t ask you to spend more or loosen targeting; they just point out something that’s obviously broken. Apply these without much hesitation — they’re the closest thing in the recommendations feed to an actual audit finding.

The second bucket is structural or creative suggestions: adding responsive search ad assets, adding sitelinks and callouts, consolidating ad groups, enabling automatically created assets. Most of these are low-risk and often genuinely useful, but they deserve a quick look before applying, because “automatically created assets” in particular means Google is allowed to generate and test headlines and descriptions using your existing site content and ad copy without you writing or approving them first. That’s a reasonable trade for a lot of accounts. It’s a bad trade for anyone in a regulated industry, anyone with strict brand voice requirements, or anyone whose landing pages have thin or outdated copy that would make a poor source for auto-generated ad text.

The third bucket is where the incentive conflict is sharpest: bidding, budgets, and keyword expansion. This is the category where a recommendation that raises your optimization score can very plausibly lower your account’s actual efficiency. A “increase budget” recommendation is generated whenever a campaign is losing impression share to budget limitations — which is true information, but it says nothing about whether that lost impression share is worth chasing at your current cost per conversion. If you want to check whether a budget recommendation is actually warranted before you accept it, it’s worth first confirming whether the impression share you’re losing is a budget problem or a rank problem, because those two causes call for completely different fixes, and only one of them is solved by spending more.

Bid strategy recommendations sit in the same bucket. Google will suggest switching from Maximize Conversions to Target CPA, or lowering a Target ROAS, or moving off manual bidding entirely, and the suggestion is often reasonable — but only if your account has the conversion volume and tracking accuracy to support the automation, and only if the target itself is one you’d choose on your own terms rather than one Google is nudging you toward. If you’re not sure which bidding approach actually fits your account’s volume and goals, that decision deserves more thought than a one-click “Apply,” and it’s worth working through how to choose between Target CPA, Target ROAS, and Maximize Conversions on its own merits before letting a recommendation make the call for you.

Keyword recommendations carry a similar problem. “Add these keywords” or “broaden this match type” suggestions are generated by looking at search query patterns and existing performance, and they can genuinely surface relevant terms you missed. They can also add volume that looks fine in aggregate and is quietly wasteful once you dig into it. The best gut-check for any keyword expansion recommendation is the same skill used for regular account maintenance: actually reading what people are searching before you decide the new terms are worth bidding on, the same way you’d work through a search term report to separate the terms worth keeping from the ones worth excluding.

Newer, more automated campaign types tend to generate the most recommendations of this kind, simply because they’re built around ceding more control to the algorithm in the first place. If you’re running AI Max for Search campaigns, expect the recommendations feed to lean heavily toward broadening whatever inputs you’ve kept control over — text customization, URL expansion, final URL flexibility — because that’s the direction that raises the score fastest for that campaign type. Knowing what you still actually control in that setup, and what’s already been handed over by design, is what makes it possible to tell the difference between a recommendation that’s expanding your reach and one that’s just removing your last remaining constraints.

A Worked Example: Reading a Budget Recommendation Correctly

Say a campaign shows a recommendation to raise its daily budget by 20%, with a projected optimization score gain of 3%. Before touching it, the questions worth answering are all findable in the account, not in the recommendation card itself.

Start with impression share lost to budget on that specific campaign, over the last 30 days, not just the last few days — budget-limited campaigns often show spiky daily patterns rather than a consistent shortfall, and a single bad week can trigger the recommendation even if the campaign is usually running its full budget comfortably. If the loss is real and consistent, look at the campaign’s current cost per conversion against your actual target, not against Google’s suggested one. A campaign converting at a profitable cost per lead that’s also losing impression share to budget is a reasonably strong case for raising the cap, because you have direct evidence the extra spend converts at a rate you can live with.

A campaign that’s already sitting above your target cost per conversion, losing impression share to budget, tells a different story. In that case the honest read is that the campaign wants to spend more at a price you’ve already decided is too expensive, and a budget increase mostly buys more of the same marginal cost, not more efficient volume. The optimization score has no way to make that distinction — it sees “impression share lost to budget” and treats it as unclaimed opportunity regardless of what that opportunity costs. The account owner is the only one positioned to check both sides of that trade before clicking Apply.

The same kind of check applies to a keyword or audience expansion recommendation, just with a different set of numbers to pull first. Before adding a batch of suggested keywords, look at how the existing, closely related keywords in that ad group are actually converting, and check the match type they’re running on. If tightly matched terms are converting well and the suggestion is to add broader variants of the same theme, that’s a reasonable test to run with a capped budget and a short review window. If the suggestion is to add terms that only loosely relate to what the ad group is actually selling — a common outcome once match types get wide enough — the search term data will usually show that mismatch within a week or two, well before it shows up as a real dent in overall efficiency. Either way, the decision comes from reading the account’s own data, not from the size of the projected score gain sitting next to the suggestion.

Why the Target Isn’t 100%

Because the score rewards adoption rather than results, chasing it to 100% usually means accepting every budget increase, every broad match expansion, and every automated bid strategy change the system offers — regardless of whether any individual one makes sense for your account. Advertisers and agencies who track this over time tend to land in a similar place: a score somewhere in the 70–85% range is a reasonable steady state for an account that’s being actively managed, not neglected. Anything close to 100% is a signal worth checking, not celebrating, because it usually means every available lever has been pulled, including the ones that trade efficiency for scale.

The reverse is also true. A very low score isn’t automatically bad either. If you’ve deliberately dismissed a string of recommendations because they didn’t fit your account — turning down a broad match suggestion because your match type discipline is intentional, or declining a bid strategy switch because you don’t yet have the conversion volume to support it — your score will reflect those “no” decisions as missed opportunity, even though each one was the right call. The score has no way to distinguish a deliberate, informed dismissal from neglect. That’s a real limitation of the metric, not a flaw in your account.

The Results Tab: A Slightly Better Signal

Google has been rolling out a “Results” view alongside the recommendations feed that tries to close part of this gap by tracking what actually happened after a recommendation was applied, rather than just crediting the score bump. Coverage of it so far shows it focused on a couple of categories to start — budget recommendations and target recommendations — showing the before-and-after performance for campaigns where you accepted the suggestion.

This is a genuinely useful addition if you use it as a feedback loop rather than more decoration. If you accepted three budget increase recommendations last quarter, the Results tab is where you go to check whether those campaigns actually held their cost per conversion afterward, not the optimization score, which never revisits a decision once it’s counted. Treat it as a way to audit your own past “Apply” clicks, and use what you find there to get more selective about which categories of recommendation you accept going forward — if budget increases have consistently held efficiency for your account, that’s useful evidence; if they’ve consistently pushed cost per conversion up, that’s useful evidence too, and it should change how quickly you click Apply next time one shows up.

It’s still worth being clear-eyed about the limits of that view. Two categories of tracked outcome is a start, not full coverage — keyword expansions, ad creative changes, and audience adjustments don’t get the same before-and-after treatment yet, so the same manual scrutiny still applies to everything the Results tab doesn’t cover.

Auto-Apply: The Setting Doing the Most Damage Quietly

Optimization score itself is mostly harmless if you treat it as a suggestion box rather than a scoreboard. The setting that causes actual damage is auto-apply, and it’s worth understanding exactly how it works because the interface makes it easy to turn on more of it than you meant to.

Auto-apply recommendations let Google implement certain categories of suggestions automatically, without you reviewing or approving each one. They’re managed from the Recommendations page under an “Auto-apply” tab, where each recommendation type — new keywords, expanded audiences, budget changes, bid strategy switches, ad rotation settings, and more — has its own individual toggle. There is no single “turn it all off” switch. Each category has to be reviewed and disabled on its own, which means the default state of a lot of accounts is some patchwork of auto-apply settings that got turned on at some point, by whoever set up the account, and never revisited since.

The categories worth being most careful with are the ones that touch spend and targeting directly: automatic budget increases, automatic bid strategy or target changes, and automatic keyword or audience expansion. Turn any of these on and Google will make the change itself the next time the recommendation appears eligible, with no approval step and no notification that draws your attention the way a manual review would. You’ll see the change after the fact, buried in a change history log most people don’t check daily. If your account runs on tight margins, or if you report performance to a client or a boss on a monthly cycle, discovering three weeks in that your target ROAS was auto-lowered without a conversation is not a good place to be.

Checking your current exposure only takes a few minutes: open Recommendations, go to the Auto-apply tab, and go through each toggle one at a time. For any category touching budgets, targets, or keyword and audience breadth, the safer default is off, with a note to review that category manually on whatever schedule you’ve settled on for the rest of the account. For genuinely mechanical categories — things like removing disapproved ads or fixing broken final URLs — leaving auto-apply on is a reasonable, low-risk choice, since the alternative is usually just a broken ad sitting live for longer than it needs to.

Auto-apply isn’t universally wrong. For an account that’s genuinely under-managed — nobody logging in for weeks at a stretch, no one available to review recommendations regularly — letting Google handle basic repairs and low-risk asset suggestions automatically is probably better than leaving obvious problems to sit. The distinction is between auto-applying fixes (broken sitelinks, disapproved ads, conflicting negatives) versus auto-applying decisions (budget levels, bid targets, match type breadth, audience scope). The first category is largely mechanical. The second category is strategy, and strategy shouldn’t run on autopilot without someone checking the direction it’s heading.

A Practical Way to Triage Recommendations

Rather than clicking “Apply all” or ignoring the tab entirely, a workable middle path is a short recurring review, treated the same way you’d treat any other piece of account maintenance. The right frequency depends on account size and spend level, and it’s worth thinking about that cadence deliberately rather than by accident — how often you actually need to check a Google Ads account is a genuinely different answer for a $2,000/month local services account than for a seven-figure e-commerce account, and the same logic applies to how often the recommendations tab needs a look.

When you do sit down with the list, a few questions make the triage faster:

• Is this a repair or a decision? Broken assets, conflicting negatives, and disapproved ads are repairs — apply them. Budget, bid target, and audience changes are decisions — they need your judgment, not Google’s default.
• What’s the actual impact estimate, and is it believable? Google shows a projected score gain, and sometimes a projected conversion or click estimate. Treat the score gain as irrelevant and look only at the performance projection, and even then, treat it as a rough directional guess rather than a guarantee.
• Does this recommendation increase spend, broaden targeting, or reduce control? If yes to any of those, it deserves a manual look at recent performance data before you apply it, not a reflex click.
• Would I make this exact change if the recommendation panel didn’t exist? This is the simplest filter. If a colleague suggested raising your budget 20% or switching bid strategy, you wouldn’t do it without checking recent trends first. A recommendation card shouldn’t get an exemption from that same scrutiny just because it’s pre-packaged with a one-click button.

For campaigns running on Performance Max, this gets harder because the transparency is already limited — you’re trusting an automated system with less visibility into why it’s making the choices it makes, and the recommendations feed for these campaigns tends to lean even further toward “give the algorithm more room to work.” If PMax is a meaningful part of your account, it’s worth pairing recommendation review with a proper look at how to audit a Performance Max campaign when you can’t see inside it, so you’re evaluating those recommendations against real placement and asset-group data rather than the optimization score’s word for it.

The Risk Looks Different Depending on Who’s Running the Account

How much this all matters in practice depends a lot on who’s actually watching the account. A solo advertiser managing their own campaigns alongside running the rest of the business is exactly the profile most likely to lean on auto-apply out of necessity — there simply isn’t time to review every recommendation weekly — which makes it worth being deliberate about which categories get that trust and which don’t, rather than leaving the defaults as they were configured on day one.

An agency managing a portfolio of client accounts has the opposite problem: enough accounts that a five-minute recommendations review per account, done properly every week, adds up to real hours, and the temptation to batch-approve or lean on auto-apply scales with the number of accounts on the roster. That’s precisely where a change made without a client conversation — a budget increase, a bid target shift — becomes a reporting headache a month later, when the client asks why cost per lead moved and nobody has an easy answer because nobody remembers approving the change.

Where Ongoing Monitoring Fits

The pattern underneath most of this is straightforward: a weekly or monthly glance at the recommendations tab catches the recommendations, but not necessarily the consequences of the ones you already accepted, or the ones an auto-apply setting quietly implemented a few days after your last login. An account that’s watched daily catches a budget cap change or an auto-added keyword within a day or two, while an account checked once a month can carry an unnoticed change for weeks before anyone connects it to a shift in cost per conversion.

This is the specific gap that Growera’s continuous account monitoring is built to close. Instead of relying on a periodic manual check of the recommendations tab and the change history log, it watches the account daily and flags meaningful shifts — in spend, in bid targets, in structural changes — close to when they happen rather than at the end of a reporting cycle, so a recommendation that turned into a real change doesn’t sit unnoticed until the next scheduled review.

Summary

Optimization score measures how many of Google’s suggestions you’ve applied, not how well your account is actually performing, and the two can genuinely move in opposite directions. The score also can’t see conversion tracking quality, account maturity, or business context like margin and capacity, all of which matter more to a real decision than the score ever will. Recommendations split into repairs, which are close to free wins, and decisions about spend, targeting, and bid strategy, which deserve the same scrutiny you’d apply to any change you made on your own initiative. A score in the 70–85% range is a reasonable, healthy place for an actively managed account to sit — 100% usually means every lever got pulled, including the ones trading efficiency for scale. Auto-apply is the setting most worth auditing directly: it has no master off switch, only individual toggles, and the categories touching budget, bid targets, and keyword or audience expansion are the ones worth turning off and reviewing by hand. Treat the recommendations tab as an input to your own decisions, on whatever cadence fits your account’s size, and check in on what actually changed rather than letting the score tell you the job is done.