Stop Following Google Ads Recommendations (UK Guide 2026)

The blunt answer: turn auto-apply off, treat the Recommendations tab as a list of suggestions from someone whose job is to sell you more inventory, and apply maybe 1 in 10 recommendations after manual review. Google's recommendations are optimised for Google's revenue and "optimisation score" - not your CPA, your ROAS, or your business. Below, Ben breaks down which recommendations are genuinely dangerous, which are occasionally useful, and how to lock auto-apply down in 60 seconds.

The short answer

Three rules: (1) turn off every auto-apply toggle today, (2) treat the Recommendations tab as a checklist to review, never to blindly accept, and (3) ignore optimisation score - it doesn't correlate with revenue. The recommendations engine is built to maximise inventory consumption inside Google's network. Some recommendations help; most expand keyword match types, add audience signals, raise budgets and convert manual campaigns to automated ones - all of which benefit Google more than they benefit your CPA.

Why most Google Ads recommendations are bad

Three structural problems:

1. Conflict of interest. Google's recommendations are generated by Google. Almost every accepted recommendation either expands spend, broadens targeting, or moves you to a more automated product where Google takes more control.
2. Account-blind. The engine doesn't know your margins, your lead-to-sale rate, your CRM data, or that "form_submit" includes spam. It optimises for whatever conversion action is set as Primary - even if that conversion action is unreliable.
3. Forecasts are best-case. The "expected lift" numbers shown next to each recommendation are projections under ideal conditions. Real-world delivery is consistently lower, sometimes negative.

Optimisation score is a marketing metric, not a performance metric

Optimisation score is the percentage at the top of the Recommendations tab. It's not correlated with revenue, ROAS, profit, lead quality, or anything that matters. Google publicly states it's a guide; in practice it's used by sales reps to push you to apply more recommendations.

Ben has audited UK accounts with optimisation scores below 50% delivering 6-8x ROAS, and accounts at 100% delivering 1.5x. Score is a tactic, not a target. Ignore it.

The worst recommendations to ignore

1. "Add broad match keywords"

Broad match in Smart Bidding is a different beast to broad match a decade ago, but it still routinely matches queries that have nothing to do with your product. Apply this and your search terms report fills with junk within days. Decision: ignore by default. Test broad match deliberately in a separate experiment, never via auto-apply.

2. "Increase your budget by X%"

Frequently recommended at +30%, +50% or even +100% in one go. As covered in Ben's article on budget increases, anything above ~20% triggers a Smart Bidding learning phase and tanks CPA for 2-3 weeks. Decision: ignore. Scale at 20% in your own time.

3. "Apply auto-apply for all suggestions"

Google sometimes pre-toggles this when you create a new account or add a new user. Decision: turn it off the moment you see it.

4. "Upgrade to Performance Max"

Recommended for almost every Search and Shopping campaign. PMax is right for some accounts and wrong for many others - and once you upgrade, downgrading is messy. Decision: never accept this from a recommendation. Decide separately based on the account.

5. "Add new responsive search ad assets"

Google sometimes auto-generates RSA headlines and descriptions using AI. Quality is wildly inconsistent - factually wrong claims, broken brand voice, bad grammar. Decision: ignore auto-generated copy; write your own.

6. "Switch to Maximise Conversions / Maximise Conversion Value"

Recommended on tCPA and tROAS campaigns. Removing the target hands Google a blank cheque to spend whatever it wants per conversion. Decision: ignore unless deliberately moving to uncapped bidding for a reason.

7. "Apply audience signals to all campaigns"

Often suggests broad audiences (in-market, affinity) that don't match your real customer profile. Decision: review case by case; never auto-apply.

8. "Remove non-serving keywords"

Google flags keywords with low impressions for removal. Often these are exact-match high-intent terms with seasonal demand. Removing them permanently kills future opportunity. Decision: ignore.

The sometimes-useful ones

A small minority of recommendations are worth reviewing manually:

• "Add sitelinks / callouts / structured snippets" - if you genuinely don't have them, this is a free CTR boost. Apply manually after writing your own copy.
• "Use enhanced conversions" - genuinely useful for recovering 5-15% of lost UK conversions. Apply, but configure properly via GA4 or gtag rather than the one-click flow.
• "Add conversion tracking" - if a campaign genuinely has no tracking, fix it. Use the recommendation as a prompt, not the implementation.
• "Fix disapproved ads / extensions" - useful as a notification. The fix itself is usually a manual edit.
• "Add negative keywords" from search-terms recommendations - sometimes Google catches obvious junk you missed. Review the list, apply selectively.

How to turn off auto-apply (60 seconds)

1. Open the account in Google Ads.
2. Go to Recommendations in the left nav (or Goals → Recommendations).
3. Click "Auto-apply" in the top right of the page (sometimes labelled "Apply automatically").
4. Untick every box in both the "Maintain your ads" and "Grow your business" sections.
5. Save.

For MCC accounts, repeat this on every linked client account - auto-apply settings live at the account level, not the MCC.

Verify by returning to Recommendations a week later and confirming nothing has been applied without your input.

How to evaluate any recommendation

Before applying, ask four questions:

1. Whose interest does this serve? Spend goes up = good for Google, neutral or negative for you.
2. What's the rollback path? Some changes (PMax upgrade, removing keywords) are hard to reverse cleanly.
3. Does the data behind it match my CRM? If conversions in Google Ads and your CRM disagree, the recommendation is built on bad inputs.
4. Could I run this as an experiment instead? Most meaningful changes deserve a Drafts and Experiments split, not a one-click apply.

Real damage Ben has seen

Three quick UK examples from accounts Ben rescued:

• £18,000 in three weeks. Auto-applied "increase budget by 50%" on a Performance Max campaign during a quiet sales month. Daily spend tripled, CPA went from £42 to £138, ROAS halved.
• Lost 80% of organic-style search terms. Auto-applied "add broad match" expansion converted carefully scoped phrase-match keywords to broad. Within a week the search terms report was 70% irrelevant queries.
• Killed brand campaign. Auto-applied "remove non-serving keywords" deleted a brand keyword variant during a quiet weekend. Branded paid clicks dropped to zero on Monday morning - took 3 days to spot.

None of these were detectable without weekly manual auditing. Auto-apply means you find out when the spend report comes in.