Optmyzr released a study back in April examining around 20,000 Google Ads accounts, and one finding has been making the rounds: Ad Strength doesn’t predict performance. Ads rated “Average” had the best CPA ($12.43) and the highest conversion rate (12.65%). “Excellent” ads had the worst CPA in the dataset ($28.68) and the lowest conversion rate (4.97%). “Poor” ads had the best ROAS of any group. It also lines up with what they found in 2024, so this isn’t a fluke.
The headline itself isn’t new. Ginny Marvin, Google’s Ads Liaison, has been saying for years that Ad Strength isn’t used in the auction; it’s a diagnostic tool, not a ranking input, and Google’s own Help Center says the same thing. What the Optmyzr study adds is performance data alongside that statement.
I don’t want ya’ll to read this as direction to treat ad strength as useless and something to ignore, because that’s not it either. The score does its job; we, as advertisers, have misunderstood what it actually does.
What the score actually measures
Ad Strength is a completeness check. It tells you whether you’ve given the RSA format enough to work with: enough headlines, enough description coverage, enough keyword variation across your assets. Google built it to stop advertisers from under-resourcing the format, and for that, it works fine. Google’s own documentation calls it an asset-coverage indicator, which is what it is.
What the score was never built to do is tell you whether any of those assets will speak to the right buyer. A 28-character headline scores better than a 14-character one. However, the performance data runs the other way: headlines under 20 characters delivered a CPA of $9.35, compared to $18.27 for longer ones, with higher CTR and conversion rates to top it off. The interface rewards length; the conversions reward directness. Those are two different things, and the gap between them is where accounts quietly drift.
Where creative drift comes in
When you optimize toward Ad Strength, you feed the system more: more headlines, more variety, more copy stuffed into the character count. You’ve handed it more raw material and more freedom to rotate. So it does what it’s built to do and finds the combinations that pull the most clicks, CTR climbs, the score climbs and the headline that said something specific to a serious buyer gets rotated out in favour of the one that pulled a cheaper, broader click.
The dashboard reads like everything is great, but then your sales team starts telling you the leads feel different.
That’s creative drift; your ads are getting better at attracting the wrong people while every number on the screen looks like progress. It’s the reason “Average” and “Poor” ads convert better in this data. They have fewer assets and less room to wander, so the system stays closer to what you meant. A tighter creative signal is harder to drift.
The pinning finding
The study also found that partial pinning outperformed both full pinning and no pinning; the best CPA was $13.68, compared to $32.57 for fully pinned ads, with better CTR and ROAS. That maps almost exactly onto how I think about steering versus blocking. You pin the one thing that can’t be left to chance, usually your core offer or your qualifying message, and you leave the rest open so the system can do the part it’s good at.
However, when Optmyzr narrowed its focus to the 268 accounts that used all three approaches, the picture shifted: fully pinned ads delivered the best ROAS in that group, even with a $61.11 CPA and a 4.48% conversion rate. So “partial pinning wins” is the directional read, not something you can bank on in your account. The principle underneath is what you need to take back to your account: pin what has to be said, and let the machine optimize what’s safe to hand over.
For lead gen specifically, position one is usually where your qualification language lives. Leave it unpinned, and the system will test it against a broader hook, and it’ll bench your qualifier the moment the broad line pulls more clicks. The leads keep coming, but they just show up looking a little less like your buyer and a little more like everyone.
Short headlines, sentence case, and the pattern underneath
Two more findings that stood out to me.
1. Headlines under 20 characters converted better than longer ones, which directly contradicts what the interface nudges you toward. You cannot chase maximum character count and short, direct copy at the same time; they’re opposite instructions, so pick the one the data supports. Descriptions had a sweet spot around 61 to 70 characters, enough room to be persuasive without paying for length you don’t need.
- Sentence-case headlines posted a CPA of $7.46, compared to $27.47 for title-case. The controlled sample for headline casing was only 31 accounts, which is thin, so test accordingly. The direction still points hard at sentence case, and the reason is the same reason short headlines win. Sentence case reads like a person talking to you. Title Case Reads Like An Ad, And You’ve Spent Years Learning To Scroll Right Past It.
Every one of these findings points in the same direction. The choices that look most optimized by the platform’s logic, maxed-out headlines, title case, and a perfect score, are the ones underperforming the choices that read like a human wrote them for a specific person.
What to do with this
Use Ad Strength as a floor. Get yourself out of “Poor,” confirm you haven’t under-resourced the ad, and then stop treating the score as a target. Once you’re past that floor, the decisions that drive conversion quality live elsewhere in the Google Ads interface.
The questions worth asking are input-layer questions, which is where control in this platform has always lived. Which asset combinations are serving most of your impressions? Are your highest-serving combinations the ones carrying your qualification language, or have they been displaced by a higher-CTR line that speaks to a broader audience? Is your position one headline pinned to your core offer, or is it in rotation? Control what the system can reach, and it will optimize inside those walls. Leave the walls wide open, and it will optimize toward whatever’s cheapest to win.
Side note: Optmyzr’s data skews toward experienced advertisers, and it’s observational rather than a controlled test, so treat it as directional and worth testing against your own accounts rather than as a settled fact. The full study is here.
The system does exactly what you tell it to. Point it at Ad Strength, and that’s the instruction it will follow.
If you want to see what your own RSAs are optimizing toward, that’s the audit worth running before you touch another Ad Strength score. We created a checklist to help you walk through it. Download it here.
ps. if you’ve ever rebuilt a perfectly good, well-converting ad just to nudge an “Average” into an “Excellent,” no judgment. I’ve done it myself which is why I know the pain so deeply 😅