Auto Recommendations Firewall: Free Google Ads Recommendations Checklist

What you get

One file, plus the full 22-verdict table on this page. Nothing on the page needs a sign-up to read.

The Auto Recommendations Firewall sheet (JLM Agency Format). The same reference we use when we take over a Google Ads account: every recommendation type scored by impact, segment, and verdict, with the reasoning written next to it. Use the verdict column as your auto-apply settings, and the sheet as the review log for everything Google suggests after that. Copy the Google Sheet

This resource sits alongside our other worksheet templates and pairs with the Customer Avatar Builder, which defines who your spend should point at before you decide which recommendations to trust.

The firewall: 22 verdicts

The split is not balanced, and that is the point: 2 Always, 7 Sometimes, 13 Never. Most recommendation types change bidding, keywords, or targeting. Those are strategy decisions, and strategy decisions do not belong on autopilot.

Two notes on freshness. Google states that from January 26, 2026, the "Add responsive search ads" recommendation no longer automatically suggests or applies new responsive search ads; the verdict stays in the sheet because the type still appears in older change histories. And Google changes which recommendations can be auto-applied over time, so we re-check the list against the official auto-apply documentation quarterly and version the sheet.

Want it as your own working copy? Copy the Google Sheet and filter it by segment or verdict.

Why recommendations need a firewall

Google Ads recommendations are not a scam, and this page is not a conspiracy theory. Some of them fix real problems. The issue is narrower: recommendations optimize for conversions inside Google's system, and they get applied to your account automatically if auto-apply is on. Nobody in that loop is checking your margins, your match type logic, or why a negative keyword exists.

Auto-apply settings sit in the top bar of the Recommendations section, at account level, and Google offers two bundles there: "Maintain your ads" and "Grow your business." One click on a bundle can switch on a group of recommendation types at once, including bidding and keyword changes. That is how accounts end up with broad match keywords and moved ROAS targets that nobody on the team remembers approving.

The optimization score sits on top of this as the pressure mechanism. It is Google's 0 to 100% estimate of how well your account is set to perform, shown at campaign, account, and manager level. Two things in Google's own documentation matter more than the number itself. First, the score is not used by Quality Score, so it has no direct effect on your auctions. Second, applying or dismissing a recommendation both change the score. You can hold a high score by reviewing the queue and dismissing what does not fit, without letting a single automated change through.

The best independent data agrees with that reading. Optmyzr's August 2024 study of 17,380 Google Ads accounts found that higher-scoring accounts did tend to perform better, and also that 19% of accounts held scores of 90 or above without accepting Google's recommendations at all. Their conclusion, and ours: the score works as a diagnostic inventory of what Google wants to change, not as a KPI to chase.

What this looks like on real accounts

On Mossy Oak Graphics, a US vehicle wraps brand, part of my job was deciding whether Google's suggestions, like enabling AI Max, were actually useful for the account or a source of unnecessary expansion. The account had a granular structure that was earning its keep, so before touching broader automation I mapped the risk of query expansion, campaign overlap, and loss of control. Some suggestions made sense. Others would have unwound months of structural work in one click. That evaluation step is exactly what auto-apply deletes.

On ThatBlanket.co, a US personalized photo blanket brand, the negative keyword framework was built deliberately and cautiously, because words like photo, custom, and gift can be good or bad depending on the phrase. The account's 125-day export ending February 26, 2026 shows what those negatives were protecting against: kids, sports, and discount-intent terms spent about $1,621 and returned about $403 in conversion value. Two recommendation types on the Never list, "Remove conflicting negative keywords" and "Add broad match keywords," would reopen exactly that kind of spend if they ran unsupervised. One honesty note: those figures describe that specific export window, not ongoing performance, and the rebuild is a work in progress.

This review-before-apply discipline is the default on every account we manage. It is a core part of how we run Google Ads for ecommerce brands.

How to set your firewall in about 10 minutes

1. Step 1

   Open the auto-apply settings. In Google Ads, go to Recommendations and open the auto-apply controls from the top bar. They live at account level.

2. Step 2

   Record the current state. Screenshot the toggles and check the History view for who opted in and when. If changes already went through, note them before you touch anything.

3. Step 3

   Turn off both bundles. Maintain your ads and Grow your business enable groups of recommendation types in bulk. Switch them off and work from individual toggles only.

4. Step 4

   Re-enable the two Always items. Optimized ad rotation, and the conversion tracking upgrade only if your conversion setup is clean and deduplicated. Everything else stays off.

5. Step 5

   Set a weekly queue review. Apply Sometimes items only when your own data clears the bar in the sheet, dismiss what does not fit, and re-check the settings page monthly, because Google adds new recommendation types without asking.

What is a good optimization score

A good optimization score is whatever an honestly reviewed queue produces, and on well-run accounts that is often 80 to 100% through dismissals rather than blind applies. Treat the number as a to-review list. The moment it becomes a target, you are optimizing for Google's queue instead of your P&L, and the Optmyzr data above shows a fifth of high-scoring accounts got there without accepting a single recommendation.

The score matters in one indirect way: Google reps and automated emails use it as the opening argument for changes. A documented verdict sheet gives whoever manages your account a standing answer, which is worth more than the score itself.

Run it yourself, or bring us your change history

If you spend a few hundred dollars a month, copy the sheet, set the toggles once, and you are done. Nothing on this page is held back, and you do not need us.

At $10,000 or more per month, recommendation control stops being a settings exercise and becomes governance: who reviews the queue, what the change history shows, whether targets still map to margin, and what an applied recommendation quietly changed three weeks ago. That is part of what we check in a Just Free Audit. We pull your change history, list what auto-apply and rep-driven changes actually did to the account, and tell you what to revert, whether or not you hire us.

Get a Just Free Audit

What happens next: you share read-only access, we review the account against this checklist and your change history, and you get the findings on a call within a few business days. No obligation. Most useful for stores spending $10,000+/month in the US or £5,000+/month in the UK on Google Ads.