- Optimization score does not measure how your account performs. It measures whether you have dealt with the recommendations sitting in the queue.
- Dismissing a recommendation moves the score, because the score is computed from the recommendations still available. Applying is not the only way up.
- Across 17,380 accounts, the high scorers were not the obedient ones. Only 5.5% had accepted Google’s recommendations at all.
- Auto-apply can change your bids, targets, keywords and ad rotation. It cannot raise your budget.
- In September 2026, campaigns still using text customization get upgraded to AI Max whether you agreed or not.
Every Google Ads account opens on a nag. There is a percentage at the top of the Recommendations tab, and it is almost never 100%. Next to each card, Google prints how much your score would rise if you said yes.
Most advice about this tab picks a side. Google’s documentation treats the score as account health. A large slice of the PPC internet treats every card as a trap. Both readings assume the score is telling you something about performance.
It is not. Once you know what the number is computed from, the tab becomes useful and the weekly decision gets simpler.
Quick answers:
- Should you apply Google Ads recommendations?
- Does dismissing a recommendation hurt your score?
- Is a perfect optimization score worth chasing?
- Can auto-apply recommendations raise your budget?
- How do you see what Google auto-applied?
- Does optimization score affect your Quality Score or Ad Rank?
What the Recommendations tab is, and what optimization score actually measures
Start with Google’s own definition, because the giveaway is in it. Optimization score “is an estimate of how well your Google Ads account is set to perform,” on a 0 to 100% scale (Google Ads Help). Read the verb. Set to perform, not performing.
The calculation makes the distinction concrete. Google says the score is “calculated in real-time, based on the statistics, settings, and the status of your account and campaigns, the relevant impact of available recommendations, and recent recommendations history” (Google Ads Help).
That phrase, available recommendations, is the whole mechanism. Your score is dragged down by cards that are still sitting there. A card stops being available in two ways. You apply it, or you dismiss it. Google confirms both: “Applying or dismissing these recommendations changes the overall optimization score of your account” (Google Ads Help).
So the score tracks queue hygiene. Jyll Saskin Gales spent six years at Google before becoming an independent Google Ads coach. She puts it flatly. The score “is not a measure of how well your account is performing; it is a measure of whether you are reviewing your recommendations.” Dismissing a card, she writes, delivers “the exact same score uplift as applying it” (Search Engine Land, December 2025).
That single fact removes the pressure the interface is designed to create. You can hold a 100% score and take almost none of Google’s advice.
Key Numbers – Optimization score runs 0 to 100%, and one of its inputs is the relevant impact of the recommendations still available to you (Google Ads Help, read August 2026). – Across 17,380 accounts spending $500 to $1M a month, accounts scoring 90 to 100 beat sub-70 accounts by 186% on ROAS. Only 5.5% of accounts had accepted Google’s recommendations, so acceptance is not what separates the bands (Optmyzr, August 2024). – 19% of accounts reached a 90-plus score having accepted no Google recommendations at all, and 32% scored under 70 (Optmyzr, August 2024). – Text guidelines cap at 25 excluded words or phrases and 40 messaging restrictions per campaign (PPC Land, September 2025). – Starting September 2026, campaigns using text customization are automatically upgraded to AI Max (Google Ads Help, read August 2026).
The 17,380-account study that complicates “never apply recommendations”
There is real evidence that high-scoring accounts perform better, and it deserves an honest reading. Optmyzr analysed 17,380 Google Ads accounts running at least 90 days and spending between $500 and $1M a month. Accounts in the 90 to 100 band beat accounts under 70 by 186% on ROAS, and carried the lowest CPAs in the study (Optmyzr, August 2024).
Taken alone, that looks like an argument for doing what Google says.
The acceptance numbers in the same study say otherwise. Only 5.5% of those accounts had accepted Google’s recommendations, while 32% scored under 70 and 19% reached a 90-plus score without accepting a single one (Optmyzr, August 2024).
Do the arithmetic on that. If only one account in twenty accepted Google’s advice, then acceptance cannot be what separates the top band from the bottom one. Whatever drives the 186% gap is something both bands mostly share an absence of. Spend level, account maturity and whether anyone is actively managing the account are all likelier explanations, and the study does not isolate them.
Two more findings keep the reading honest. ROAS was flat between scores of 70 and 90, and the study found no correlation between optimization score and either CPC or CTR (Optmyzr, August 2024). The score is not a smooth dial you can tune, and this is correlational work rather than a controlled test.
So the conclusion is not “ignore the tab.” It is that the tab is a review queue.
What each recommendation family actually changes in your account
Judging 30-plus individual cards under a nag banner is the wrong unit of work. Judge families instead, and the weekly review takes ten minutes. The families below are grouped from the recommendation types Google names in its own auto-apply documentation (Google Ads Help).
Read the verdict column as a starting default, not a rule. Most of those positions come from one agency publishing its own defaults, which is a stated preference rather than proof. The middle columns are the mechanics, and those are what should decide your call.
| Recommendation family | What it actually changes | What it cannot change | Accept, decline or test | Why, and the source |
|---|---|---|---|---|
| Measurement (Upgrade your conversion tracking) | How conversions are counted and attributed | Your budget | Accept, after you read the change | Measurement is upstream of every bid decision. Gales sorts this into the “ROI and hygiene” group rather than the reach group (Search Engine Land) |
| Bidding targets (Set a target CPA, Adjust your CPA targets, Set a bidding strategy target) | The target value your Smart Bidding strategy chases | Your budget | Test, never auto | Grow My Ads declines blanket smart-bidding suggestions because the strategy “may not deliver desired optimization results” for the account’s actual goal (Grow My Ads) |
| Keyword removal hygiene (Remove non-serving keywords, Remove conflicting negative keywords) | Deletes keywords and resolves negative conflicts | Your budget | Accept conflicts, review removals | A conflicting negative is a real fault. But Grow My Ads declines redundant-keyword removal, since “even slight keyword variations target different search intents” (Grow My Ads) |
| Keyword and targeting expansion (Add broad match keywords, Add keywords, Use targeting expansion) | Adds match types and widens who sees the ad | Your budget | Decline on auto, test manually | Gales files broader targeting under “reach and spend” (Search Engine Land); Grow My Ads notes not all suggested keywords “align with campaign business goals” (Grow My Ads) |
| Reach expansion (Expand your reach with Google search partners, Use Display expansion) | Serves your ads on surfaces outside Google Search | Your budget | Decline | Grow My Ads declines search partners because ads “may underperform on partner sites versus Google directly” (Grow My Ads) |
| Ads and assets (Improve your Responsive Search Ads, Use optimized ad rotation) | Rewrites ad text and changes which ad shows | Your budget | Decline on auto | Grow My Ads declines ad suggestions because automated copy “may conflict with brand voice” (Grow My Ads) |
| Budget raises | Your daily budget, but only when you apply it yourself | Nothing, because auto-apply never touches this row | Decide manually | Google states auto-apply “won’t increase your budget,” so every budget change stays a human action (Google Ads Help) |
A worked example of the one family worth testing rather than refusing. Take the “Adjust your CPA targets” card. It proposes a new number for an existing Target CPA strategy. Before you touch it, note three things: the strategy name on the campaign, the conversion action it optimises to, and your own target.
Say the campaign runs Target CPA at $85, optimising to a conversion action named Qualified Call, and the card proposes $60. That is not a small tune. A lower target tells the system to buy cheaper conversions, which usually means less competitive auctions and thinner intent. The condition for accepting it is narrow. Take it only if your conversion action counts the outcome you actually sell. And only in a campaign with enough conversion volume for the change to be measured. Otherwise set your own target and dismiss the card. Your score rises either way.
Re-checking those strategy settings after Google’s renames is a separate job. We covered it in auditing your Target CPA and Target ROAS settings after the rename and how Google changed target-based bidding in 2026.
Auto-apply: the full list, and the one thing it cannot touch
Auto-apply is the same recommendation engine with your review step removed. Google lets you hand over whole groups of changes, and it publishes the list. The named types fall into four groups (Google Ads Help).
| Auto-apply group | Named recommendations inside it | Reversible without data loss? | Blast radius if left on |
|---|---|---|---|
| Measurement | Upgrade your conversion tracking | No, historical measurement does not re-run | Changes how every future bid decision is informed |
| Bidding | Set a target CPA, Set a target ROAS, Adjust your CPA targets, Adjust your ROAS targets, Bid more efficiently with Maximize clicks, Maximize conversions, Target CPA, Target impression share or Target ROAS, Maximize conversion value, Set a bidding strategy target | Settings yes, spend no | Silently changes what your budget is chasing |
| Keywords and targeting | Add keywords, Add keywords (Smart bidding), Add broad match keywords, Add audiences, Add audience segments for reporting, Remove redundant keywords, Remove non-serving keywords, Remove conflicting negative keywords, Use targeting expansion, Use Display expansion, Expand your reach with Google search partners, Add dynamic search ads | Settings yes, spend no | Widens who sees your ads without a review step |
| Ads and assets | Improve your Responsive Search Ads, Add dynamic search ads, Use optimized ad rotation | Yes | Changes live ad text and which ad serves |
The one hard limit is often overstated, so know it precisely. Google is explicit: “Auto-applying recommendations won’t increase your budget” (Google Ads Help).
So auto-apply cannot raise your daily budget. It can change what that budget buys, by moving a bid target or adding broad match keywords. Spend control and spend direction are different things, and only the first is protected.
Audit it rather than assume it. The setting is opt-in, and Google’s own instructions are short: “Go to the Recommendations page within the Campaigns menu. Select Auto-apply settings. Uncheck the recommendation you want to turn off. Select Save” (Google Ads Help). Anyone with account access can have ticked those boxes, including a previous agency.
One reading aid from the same piece: blue and yellow notifications are recommendations and are safe to dismiss, while red and purple flag real problems. The colour tells you whether you are being sold something or warned about something. Where that line sits generally is covered in which marketing tasks are safe to automate and which quietly spend budget.
Set your text guardrails before you switch anything on
Not every piece of Google automation arrives as a card you can dismiss. Some of it is a campaign setting that is simply on, and text customization is the one worth knowing before it changes state. It is the feature “formerly known as automatically created assets,” and it is also a campaign-level setting available through AI Max for Search (Google Ads Help). It generates headlines and descriptions using “a combination of ‘light-touch’ extractive techniques (for example, taking snippets from landing page titles, descriptions, and meta tags) and generative AI technology that are grounded in your landing page content” (Google Ads Help).
Google also states these assets “will only serve if they are predicted to perform better than the headlines and description lines you uploaded.”
The controls that hold are configuration, not copy. Google’s text guidelines let you block up to 25 specific words or phrases from generated text, matched as case-insensitive exact terms. You can also add up to 40 messaging restrictions per campaign, written as plain-language directives. Text customization must be set to ON for either to work (PPC Land, September 2025).
Here is what that looks like typed into the fields rather than described.
Term exclusions (25 slots, exact term, case-insensitive):
free, guaranteed, cheapest, instant approval, no cost, lowest price, discount, sale, risk-free, bestMessaging restrictions (40 slots, written as directives). PPC Land documents the format with “Don’t imply our products are discounted” and “Never mention any city names in my ads.” Written out for a residential HVAC campaign, a set looks like this:
Don't imply our products are discounted Never mention any city names in my ads Do not state or imply a price Do not imply we service commercial buildingsNote what these bind. Guidelines constrain what the generator may write. They do not rewrite the headlines and descriptions you uploaded yourself, so a restriction and your own pinned asset are separate controls that need to agree with each other.
Then write the boundary into a headline you pin yourself. A generated headline optimises for matching the query. A written one can carry a limit, and a limit is what stops the wrong click:
Headline 1: AC Repair, Not Installs Headline 2: Residential AC Service Description 1: Licensed AC repair for homeowners. Same-day appointments. No new system quotes.That first headline is doing unpaid work. It sends the homeowner shopping for a full system replacement away before they cost you a call. Note that PPC Land also flags the failure mode in the other direction: overly restrictive guidelines can strip out assets that would have performed. The asset-level controls that pair with this live in the asset-level controls on your responsive search ads.
The recommendations that stop being optional in September 2026
One of these is on a timer. Google’s documentation states that “starting in September 2026, campaigns using text customization, formerly known as automatically created assets, will automatically be upgraded to AI Max” (Google Ads Help).
Read what that changes. Today, text customization is a campaign-level checkbox you can leave off. After the upgrade lands, the campaigns that still had it on are AI Max campaigns. The decision does not get taken away from you, but the default moves, and defaults decide most accounts.
The setting to find, and the state to record. Open the campaign, then Settings, and locate text customization. Write down whether it is on or off today, per campaign. That is your before state, and after September you will want it.
The two dates worth putting in the calendar:
- January 26, 2026. The “Add responsive search ads” recommendation “will no longer automatically suggest or apply new responsive search ads” (Google Ads Help). Read that card name carefully, because Improve your Responsive Search Ads is a separate recommendation and remains available to auto-apply.
- September 2026. Campaigns still running text customization become AI Max campaigns (Google Ads Help).
Automation is not a one-way ratchet. It is a set of individual defaults, each of which can move independently, which is why the audit is a recurring job rather than a one-off.
The wider AI Max timetable sits in the AI Max September deadline and what it changes for broad match, and the Shopping side in how AI Max works in Shopping campaigns.
What the Recommendations tab does not do
It does not report on your results. Optimization score “is not used by your Quality Score,” and Google shows it only for active Search, Display, Video Action, App, Performance Max, Demand Gen and Shopping campaigns (Google Ads Help). It is not an auction input.
It does not know what you sell. The engine optimises toward whatever conversion action you configured. If that action fires on a form view, every card that follows helps you buy more of the wrong thing efficiently.
It does not read your margins, your capacity or your deliberate constraints. A budget cap you set on purpose looks identical to one you forgot about.
And it does not settle anything. The Optmyzr correlation is not a controlled test, Gales is one expert with a clear position, and Grow My Ads is an agency stating its defaults. What holds across all three is procedural: open the queue on a schedule, decide every card deliberately, and let the score follow.
Frequently Asked Questions
Should you apply Google Ads recommendations?
Some of them, deliberately, and never on autopilot. Treat measurement recommendations such as conversion-tracking upgrades as the highest-value group, since everything downstream depends on counting the right thing. Treat reach and targeting expansion as the group to decline by default, because it widens who sees your ads without telling you what that traffic is worth. Bidding target changes sit in between and are worth testing manually. Declining costs you nothing on the score: dismissing a card moves it the same way applying does (Search Engine Land).
Does dismissing a recommendation hurt your score?
No. Google states that “applying or dismissing these recommendations changes the overall optimization score of your account” (Google Ads Help). The reason is in the calculation. The score is computed in part from the relevant impact of the recommendations still available to you. Once a card leaves the queue, its drag on the score leaves with it. Jyll Saskin Gales, a former Google employee of six years, describes the uplift from dismissing as identical to the uplift from applying (Search Engine Land). Dismissing is the correct action for any recommendation you have genuinely evaluated and rejected.
Is a perfect optimization score worth chasing?
Only as a by-product. Optmyzr’s analysis of 17,380 accounts found that accounts scoring 90 to 100 beat sub-70 accounts by 186% on ROAS. But only 5.5% of accounts had accepted Google’s recommendations, and 19% hit 90-plus having accepted none (Optmyzr, August 2024). The score is tracking whether somebody works the account, not whether they obey it. The same study found ROAS flat between 70 and 90, and no correlation with CPC or CTR. Grinding from 85 to 95 for its own sake is not a performance lever.
Can auto-apply recommendations raise your budget?
No. Google’s documentation is explicit that “auto-applying recommendations won’t increase your budget” (Google Ads Help). Budget increases are surfaced as cards you can choose to apply, never applied for you. That protection is narrower than it sounds. Auto-apply can still change your bid strategy targets, add broad match keywords, switch on targeting or Display expansion, and alter your ad rotation. Those change what your existing budget buys, and they can move your effective cost per outcome substantially without any change to the daily cap.
How do you see what Google auto-applied?
Use the History tab, and audit the settings alongside it. Google’s instructions are: go to the Recommendations page within the Campaigns menu, select Auto-apply settings, uncheck the recommendation you want to turn off, then select Save (Google Ads Help). Do this even on accounts you believe never enabled it, since any user with access can have ticked the boxes, including a previous agency. Record the state before you change it, so you know what the account was doing rather than what you assumed.
Does optimization score affect your Quality Score or Ad Rank?
No. Google states plainly that optimization score “is not used by your Quality Score” (Google Ads Help). It is an account-management prompt, not an auction input, so it does not influence what you pay or where you rank. Google also limits where it appears: active Search, Display, Video Action, App, Performance Max, Demand Gen and Shopping campaigns only. A low score does not throttle your ads. It means there are unread cards in a queue, which is a workflow fact rather than a performance one.
