Google's optimization score measures how much of Google's available ad inventory your campaigns access, not how well those campaigns serve your business. The three most common recommendations - AI Max, broad match expansion, and Performance Max - primarily expand the inventory Google can sell you. For B2B companies, these recommendations often increase spend without increasing qualified pipeline.
You log into Google Ads and there's a number staring at you. 62%. Maybe 74%.
Below it, a list of suggestions. Each one promises a specific percentage uplift. There's a bright blue "Apply" button next to each. And a small badge that says: "This recommendation is an AI Essential."
It feels like a to-do list from a system that knows more than you do. The percentage feels scientific. The badge feels urgent. So you click Apply.
Your costs go up. Your pipeline doesn't.
One business owner described what happened after accepting Google's optimization recommendations: his ad spend more than doubled in the first two months, quadrupled by month three, and hit ten times his normal budget by month five. "Meanwhile," he wrote, "there's zero significant change to business leads."
This isn't a bug in the system. It's the system working exactly as designed.
Before we go further: this blog is not a case against Google Ads. For capturing people who are actively searching for what you sell, there's no better channel. Google Ads works. But the platform's recommendations are designed around Google's interests, not yours. The problem isn't the car. It's who's driving it.
When you hand the steering wheel to Google, it drives toward what success looks like for Google - more of your budget spent, across more of Google's ad space. Your job is to set guardrails that keep it driving toward your destination. The optimization score doesn't help you do that. It helps Google.
What the score actually measures
Google's own documentation defines the optimization score as "an estimate of how well your Google Ads account is set to perform." That sounds like a health check. It isn't.
The score is calculated from one thing: how many of Google's recommendations you've dealt with. Each recommendation carries a weighted percentage. When you accept one, your score goes up by that amount. When you dismiss one, your score goes up too - but by less. The score isn't tracking whether your campaigns make money. It's tracking whether you've reviewed Google's suggestions.
Search Engine Land put it bluntly: the optimization score "is not a measure of how well your account is performing; it is a measure of whether you are reviewing your recommendations."
Think about what this means. The score goes up when you do things that expand Google's ability to show your ads to more people, in more places, more often. Broader keywords. More campaign types. Higher daily budgets. Automated bidding. It goes down when you restrict - when you use only precise keywords, set manual bids, or limit your geography.
It never penalizes you for spending more. It only penalizes you for spending less, or spending more carefully.
Now, in fairness: Google's Ads Liaison has clarified that the optimization score has "no influence on the auction." It doesn't affect whether your ads show up or where they rank. That's true, and worth knowing. But the score's power was never about the auction. It's about the psychology. A number below 80% feels like you're doing something wrong. The blue "Apply" button feels like the fix.
Optmyzr analyzed 17,380 Google Ads accounts to test Google's claim that raising your score by 10 points leads to 15% better results. What they found: only 5.5% of accounts actually accepted Google's recommendations. 19% of accounts achieved scores above 90% without implementing a single recommendation - just by dismissing them. And the correlation between high scores and good performance? It existed, but mostly because well-managed accounts are tidy accounts. The score wasn't causing better results. Good managers just happened to keep things clean.
Optmyzr's conclusion: "Optimization Score is not and should never be a KPI."
The real question isn't what the score measures. It's what the recommendations do.
The three recommendations that cost you the most
Three suggestions appear in nearly every Google Ads account. They're flagged as "AI Essentials." Each one promises a specific uplift. And each one does the same structural thing: it expands the surface area of Google's ad inventory that your budget can be spent on.
"Turn on AI Max" - the +25% promise
This isn't a diagnostic. It's an upsell.
AI Max is a bundle of automations you switch on inside a regular search campaign. Once enabled, it does three things. First, it expands your keyword matching - your ads can now appear for searches you never bid on, based on what Google's AI considers relevant. Second, it can rewrite your ad copy using AI, generating headlines and descriptions you didn't write. Third, it can send people to different pages on your website than the ones you specified.
Ex-Googler Jyll Saskin Gales describes it as "Search-only PMax - like DSA and broad match on top of your existing keywords." Google claims it delivers 14% more conversions at a similar cost. For campaigns that were previously on precise keywords only, the claimed uplift is 27%.
Here's where it gets complicated.
If you sell shoes online, AI Max can probably figure out that someone searching "white running shoes size 10" wants what you have. Product categories are legible to algorithms. But if you sell AI-based payroll software for mid-market companies, the intent behind a search is layered. "Payroll automation" means something different when a CFO searches it versus when a student writes a term paper. AI Max doesn't parse that. It pattern-matches to volume.
The +25% claim is an average across all kinds of advertisers - including online stores where a "conversion" might be a page view. It's not a prediction for your account.
And if your website has a blog, AI Max's URL expansion can start routing paid traffic to your "What Is Payroll?" article instead of your product page. Unless you explicitly block those URLs, you're paying for clicks that land on content designed for awareness, not for buying.
Google Ads API engineer Darshan Modi advises against enabling AI Max in mission-critical campaigns without testing first. That's Google's own engineer saying: be careful.
"Add broad match keywords" - the +13.3% promise
Your restraint isn't a mistake. It's a strategy.
If you've deliberately chosen 15 precise keywords because those are the exact searches your buyers use, Google sees that as a problem. Not because your keywords aren't working - but because precise keywords limit how many auctions Google can enter you into.
Broad match removes that limit. Instead of showing your ad only when someone searches your exact keyword, Google's AI decides which related searches should also trigger your ad. It considers the searcher's history, your landing page content, other keywords in your account, and whatever patterns it sees across billions of searches.
The recommendation language is telling: "Your campaigns are missing some broad match keywords." You didn't forget them. You chose not to include them. Google calling that a gap is like a car dealership telling you your sedan is "missing" a roof rack.
Here's the critical thing most people don't know: broad match is designed to work only with automated bidding. Google's own documentation says this explicitly. And Search Engine Land confirms: "Running broad match without Smart Bidding is no longer how the product is designed to work." Broad match opens the floodgates. Automated bidding is supposed to be the filter. Without enough data to train that filter, you just get the flood.
How much data is "enough"? Google's soft minimum is 15 conversions per month. Most practitioners say you need 30 to 50 quality conversions per month before broad match has enough signal to be useful - and some lead-generation specialists push that to 80 or 100.
B2B SaaS specialist Scott Gelber calls Google's broad match advice "a disaster for B2B SaaS companies," leading to "a ton of money wasted on irrelevant search terms and unqualified leads."
One SaaS PPC professional on Reddit put it this way: "I would avoid broad match like the plague for SaaS... Google will not be able to understand (nor care) about the nuances of the software you're promoting."
Broad match can work - but as a scaling tool, not a starting tool. It amplifies whatever signal already exists in your account. If you have strong conversion data and proven keywords, it can find more buyers like the ones you already have. If you don't, it amplifies noise.
"Create a Performance Max campaign" - the +9.9% promise
Great dashboard numbers. But who were those 'conversions' really searching for?
Performance Max runs your ads everywhere Google sells ad space - Search, YouTube, Display, Discover, Gmail, Maps - simultaneously. You give it themes and creative assets. It decides where, when, and to whom.
For online stores with product catalogs, PMax can work well. Google can understand "blue Nike Air Max, size 11" and match it to a buyer. One ecommerce case study showed a 76% increase in ad revenue after switching to PMax. When there's a concrete product with a clear category, the algorithm has something real to optimize against.
But there's a structural problem that affects nearly every PMax account, and it's more damaging than most advertisers realize.
PMax cannibalizes your branded search.
Here's the mechanism. You set up PMax with themes related to your product. PMax shows ads across various placements - including search. Some of those search placements are for people typing your company name. Those people convert at high rates, because they were already looking for you. PMax learns from those conversions and shows more ads to more people searching your name. The dashboard looks impressive - great click-through rates, strong cost per acquisition. But underneath, Google is now charging you for showing ads to people who were going to find you anyway. Your organic traffic puts on the mask of paid performance.
This isn't speculation. Optmyzr analyzed 503 accounts and found that 91% had exact keyword overlap between their Search and PMax campaigns. 97% had overlapping search terms. And where performance differed meaningfully, Search campaigns were more likely to outperform PMax on click-through rate and conversion rate.
Didgeheads, an independent PPC agency, documented a test: exclude your brand name from PMax and see if your total branded conversions drop. In many cases, they don't. The customers were coming regardless. PMax was just taking credit.
The attribution problem goes deeper. A study of 280,000+ items by Smarter Ecommerce found that PMax frequently receives fractional conversion credit - as low as 1% of a conversion - just for briefly showing an ad somewhere in a customer's journey. When those micro-credits get rolled into a "conversions" column, PMax looks more impactful than it is.
For B2B, the picture is worse. Search Engine Land's review noted that PMax's AI is "much more impactful for ecommerce" and that "we don't see a lot of lead gen quality from Performance Max campaigns" in B2B. There's no product feed equivalent for "AI-powered payroll for mid-market companies." PMax can't categorize that. So it guesses. And B2B purchases aren't impulsive - a buyer needs to evaluate, demo, get approval, negotiate. PMax is optimized for impulse journeys, not considered ones.
Like broad match, PMax can work as a scaling mechanism - but only with strict guardrails: brand exclusions, dedicated branded search campaigns running alongside it, strong conversion tracking tied to real business outcomes, and constant monitoring. Without those, it's an expensive way to pay for traffic you were already getting.
The thing that makes all three dangerous
Every recommendation above shares one vulnerability. They all optimize for whatever you've told Google counts as a "conversion."
If your conversion action is a page view, Google's AI will find people who view pages. If it's a button click, it'll find clickers. If it's a form submission with no qualification, it'll find form-fillers. The AI is very good at optimizing. The question is what it's optimizing for.
Most accounts have soft conversion definitions as their primary signal - newsletter signups, contact page visits, time on site. Google's AI treats these with the same mechanical weight as a signed contract. The optimization score then rewards you for getting more of them.
The fix isn't a tool change. It's a definition change. Your conversion actions should map to outcomes your business actually cares about - demo requests booked, qualified forms submitted, phone calls over 60 seconds. If Google says you got 40 conversions this month but your CRM says you got 3 qualified leads, your CRM is the source of truth.
Why this isn't a bug
Google Ads is not a tool. It's a marketplace. And in any marketplace, the platform makes money when transactions increase - not when your transactions get better.
Every recommendation in the optimization score expands Google's addressable inventory for your account. That's not a conspiracy. It's the business model. Broader keywords mean more auctions to enter. PMax means more placements to fill. Higher budgets mean more room to spend. Each one increases what Google can sell you.
Internal Google documents surfaced during the DOJ antitrust case showed that the company chose to "rank ads sub-optimally in exchange for more revenue." That's not an allegation. It's from Google's own files.
The optimization score creates a pressure loop. Score is low. Google recommends expansion. You apply. Spend increases. Score goes up. You feel better. But then new recommendations appear - because there's always more inventory to access. The loop never terminates.
This pressure isn't just on business owners. Google's Partner program now requires agencies to maintain optimization scores above 70% to keep their certification badge. Which means even agencies that know better are sometimes pressured to follow Google's playbook. It's not always incompetence when your agency accepts a recommendation you didn't ask for. Sometimes it's structural compliance.
The learning phase makes it worse for smaller budgets. Google's automated bidding needs roughly 30 to 50 conversions per month to calibrate properly. Most accounts spending $3,000 to $5,000 a month don't generate that volume. So the AI enters a "learning phase" where it experiments with your budget - and in one documented B2B case, costs per click spiked from $0.95 to $19.87 in 48 hours. The recommended fix? Increase your budget so the AI can learn faster. The solution to underperformance is always more spend.
What control actually looks like
The opposite of the optimization score trap isn't abandoning Google Ads. It's making your own scorecard.
A 62% optimization score with a healthy pipeline beats a 95% score with inflated conversions. Every time. Plenty of experienced practitioners now treat the score as a counter-indicator - not a target. As one put it: "I ignore this metric 100% of the time. I dismiss the recommendations which brings it up to almost 100% - which conveniently also lets you know just how much that score actually means."
If you want to take back control, here are four things you can do right now - most take less than five minutes:
Turn off auto-apply recommendations. This is the single most important setting in your account. When enabled, Google can change your bids, add keywords, rewrite your ads, and adjust targeting - without asking you first. Go to Admin → Account Settings → Auto-Apply, and uncheck everything. There are roughly 22 options in there. Uncheck all of them. You can always apply individual recommendations manually after reviewing them - but nothing should change in your account without your approval.
Check your other default settings while you're in there. New Google Ads campaigns come with several defaults that work against you: Search Partners Network is usually on (shows your ads on low-quality partner sites), Display expansion on Search campaigns is often enabled (mixes your search ads into the display network), and automatically created assets can rewrite your headlines without telling you. Turn all of these off unless you've deliberately tested them and seen results.
Redefine your conversion actions. Check what Google is currently counting as a "conversion" in your account. If it's a page view, a scroll, or a generic form submission with no qualification - that's what every recommendation is optimizing for. Change it to something your business actually values: a demo booked, a qualified form submission, a phone call over 60 seconds.
Choose keywords based on how your buyers describe their problem. You know the language they use when they're actually ready to buy. Google's AI doesn't. That knowledge is your competitive advantage - don't hand it over.
And if reading this makes you think "maybe I should just switch to Meta" - that instinct is worth examining too. Switching channels doesn't solve a strategy problem. The optimization score trap is a symptom of not having a channel-independent strategy for how your buyer finds, evaluates, and chooses you. Fix that, and Google Ads becomes a powerful tool. Skip it, and every channel will have the same problem with different labels.
Frequently Asked Questions
- Does Google's optimization score affect ad auctions?
- No. Google's Ads Liaison confirmed the optimization score has no influence on whether your ads show or where they rank. Its power is psychological - a score below 80% feels like a problem, and the blue 'Apply' button feels like the fix. But the score measures whether you've reviewed recommendations, not whether your campaigns make money.
- Should I use broad match keywords in Google Ads?
- Only if you have 30-50+ conversions per month and are using automated bidding. Broad match is designed to work only with Smart Bidding - without enough data to train the algorithm, you just get noise. For B2B SaaS with nuanced buyer intent, broad match often leads to wasted spend on irrelevant search terms.
- Is Performance Max good for B2B lead generation?
- Often not. PMax cannibalizes branded search - Optmyzr found 91% of accounts have keyword overlap between Search and PMax campaigns. You end up paying for people who would have found you organically. Search Engine Land noted PMax is 'much more impactful for ecommerce' with little lead gen quality in B2B.
- What is AI Max in Google Ads?
- AI Max is a bundle of automations for search campaigns that expands keyword matching, rewrites ad copy using AI, and can route traffic to different landing pages than you specified. Google claims +14-27% more conversions, but for B2B with nuanced buyer intent, it often pattern-matches to volume rather than qualification.
- Should I turn off auto-apply recommendations in Google Ads?
- Yes. Auto-apply allows Google to change your bids, add keywords, rewrite ads, and adjust targeting without asking. Go to Admin → Account Settings → Auto-Apply and uncheck all 22 options. You can always apply recommendations manually after reviewing them.
- Why does my Google Ads spend keep increasing without more leads?
- Every optimization score recommendation expands Google's addressable inventory for your account - broader keywords mean more auctions, PMax means more placements, higher budgets mean more room to spend. The recommendations optimize for whatever you've defined as a 'conversion' - if that's page views or generic form fills, Google will find more of those, not more qualified leads.