Quality Score is Theatre — What Actually Moves Your Google Ads Money

1. What Quality Score actually is — and what it isn't

Google says: Quality Score is a 1-10 diagnostic showing the quality of your ads compared to other advertisers. It's made of expected CTR, ad relevance, and landing page experience.

Translation: the number in the column is a smoothed 90-day post-mortem on exact-match queries. You're optimizing a photo of the deceased.

Here's the trap. The QS you see in the UI is NOT the number that actually gets used in the auction. Ad Rank is calculated in real time, for every single impression, factoring in the query, time of day, device, location, audience, user history. The column number ignores all of those signals. It's a directional indicator, not a bidding input — a speedometer, not the gas pedal. You can watch it to understand direction, but the auction doesn't use it as a single number.

Second thing almost nobody says out loud: QS 10 in one auction doesn't save you from QS 3 in another. You see one number, and the system operates on a hundred. I've seen accounts with QS 4 and pennies for CPC. I've seen accounts with QS 9 and CPCs in the tens of dollars. Position, competition, time, audience matter more than the column.

Third — QS only gains any weight once there's statistical volume behind it. A new keyword shows "—" in the column, not a number. That's not "you have QS 0," that's "Google doesn't know yet." Until roughly 200+ exact-match impressions have accumulated, the column stays empty, and that's normal. I've seen it dozens of times in chats — "help, we've got a dash, what do we do!" — and nobody explains that a dash isn't a bad grade, it's the absence of a grade. Do nothing. Wait for traffic.

Fourth — QS is not a direct input into Smart Bidding. Google confirmed this in 2022. If you're on Target CPA or Max Conversions, the algorithm looks at its own signals, which are far richer than the three QS components. Meaning for Smart Bidding, the QS column is basically decoration.

2. Why 9 out of 10 pieces of QS advice are a waste of your time

Google Help says: to improve ad relevance, put the keyword in the ad copy and the final URL.

Translation: build SKAGs (Single Keyword Ad Groups). Advice from 2015. By 2019 it was already questionable. In 2024, with RSAs and Broad match, it's actively harmful.

I tested SKAGs against grouped ad groups on a legal-niche client in 2019. SKAG QS was 8-9, grouped was 6-7. SKAG CPA was consistently higher. Because bid optimization has less data per group, the learning phase never completes, and bids bounce around. You won the beauty contest in the QS column and lost the profit contest.

Tested the exact same thing in 2024 on an e-commerce client — same result. Groups of 10-15 intent-close keywords give a lower QS in the interface and a lower CPA in reality. Didn't work in 2019, didn't work in 2024.

Another piece of advice I hate: "expand negatives to lift QS." Technically true — you cut irrelevant traffic, CTR climbs, QS climbs. But in most cases I've seen, the owner cuts way too aggressively, throws out low-intent queries that were actually converting for cheap on warmed-up remarketing audiences. You lifted QS from 5 to 7 and lost a whole layer of cheap conversions. Congratulations, you made Google happy.

And a favorite: "write new ads every week so QS grows." No. RSAs don't like being changed — you reset the learning phase every time you add a new asset. You draw a conclusion about an ad variant on 40 impressions and treat it as truth. You are the noise in your own system. The best thing you can do for QS on most accounts is leave it alone for a month.

General pattern: if a piece of advice has been around since 2015 and every agency parrots it, there's a high probability it stopped working long ago. Google's algorithms change twice a year. Half of what you were taught in 2020 is toxic in 2026.

3. Three things that actually move money: CTR, relative CTR, conversion rate

Right here, in three numbers, lives the entire economics of your account. Not in the QS column. In these three.

Absolute CTR. This is the cheapest quality signal for Google. If your ad gets a click from 8% of users while the niche average is 3%, Google concludes two things: the ad is relevant, and showing it more often means Google earns more. Google will lower your CPC so you run more actively. This isn't charity, it's business model. You feed Google — Google gives you a discount.

But CTR doesn't rise from "keyword in the headline" or from stuffing the keyword into the URL. CTR rises from four things: (1) you offer exactly what the person is searching for right now; (2) you have a hook that catches; (3) you have a competitive offer; (4) you're in a position that gets seen. Copy is skin over a real offer. "Sale 20%" loses to "Sale 50%" every time, and the keyword count in the headline is irrelevant.

Relative CTR. This is what Google Ads doesn't show directly in the UI, but it lives inside Ad Rank. Your CTR versus competitors' CTR in the same auction at the same position. The absolute number means nothing. I've seen cases where a 2% CTR ad had a high relative CTR in a B2B niche with expensive intent where almost nobody clicks. And I've seen 8% CTR that was below average in a niche where the norm is 12%.

How to measure. Auction Insights gives you overlap, outranking share, position above rate. Direct competitor CTR isn't there, but there's a proxy: your CTR on top positions should be a multiple of your average across all positions. If you're on positions 1-2 with a 3% CTR — that's a red flag. Copying competitor headlines is suicide. You become the median. The median pays maximum CPC.

Conversion rate on the site. This is the real boss. Because it determines how much you can pay per click without going into the red. It feeds Smart Bidding — no conversions, Google runs a random walk. And it's the one parameter competitors can't copy in an hour. Your ad headline I'll rewrite in 5 minutes. Your LP with proper UX and social proof I'll be copying for three months, and still not get it right.

The formula I keep in my head: Profit = Impressions × CTR × CR × (AOV − COGS) − Impressions × CTR × CPC. QS isn't in this formula. All three levers — CTR, CR, and indirectly CPC through relative CTR — are. Optimize what's in the formula.

4. Landing Page Experience — legalized extortion by Google

Google says: Landing Page Experience is a measure of how relevant and useful your website's landing page is to people who click your ad.

Translation: your site had better load fast on 4G in Mumbai, or we'll bump your CPC by roughly 20-30%.

This is the most honest of the three QS components — and the most expensive to improve. Because "fix the landing page" means either rewrite the site or hire a developer. And Google's holding all the cards here: if your LP is slow, you'll pay more per click than a competitor with a fast LP at the same bid. That's extortion, legalized through the technical wrapper of "user experience."

Google wants you to upgrade your hosting. Google wants you to buy Cloudflare Pro. Google wants you to hire a developer for Core Web Vitals. Because the better the internet, the more people click, the more Google earns. It's not a conspiracy, it's just aligned incentives: Google sells the internet as a medium and wants that medium to be fast. But it costs you.

First thing I do on any new client — run PageSpeed Insights on every landing URL being advertised. If LCP > 3.5s, I don't touch Google Ads at all until we fix it. Any bid optimization will whistle straight into a hole. You can rewrite ads 50 times, burn half a thousand dollars on bid tests — Google still penalizes you on LP, and the same CPC multiplier eats every gain.

Another trap almost nobody writes about: LP Experience for ads is evaluated by AdsBot-Google, not the regular Googlebot. It's a separate crawler that walks the URLs from your ads. And here's the gotcha: if your `robots.txt` has `Disallow: /` without an explicit `User-agent: AdsBot-Google\nAllow: /` — AdsBot is blocked. Or if Cloudflare or another bot shield shows a challenge page to AdsBot-Google — you automatically get Below Average. I've seen this maybe ten times across clients: LP Exp tanked, nobody could figure out why, and it was either robots.txt blocking or a WAF throwing CAPTCHAs at AdsBot.

Check it from the terminal:

```

curl -A "AdsBot-Google" -I https://your-landing.com/

```

Should return 200. If you get 403, 503, or a redirect to a challenge — you found it. Also check `https://your-landing.com/robots.txt` — should be either nothing that blocks AdsBot, or an explicit `User-agent: AdsBot-Google` with `Allow: /`. And run URL Inspection in Search Console just in case.

Practical: swap the hero image for WebP. Drop the third pixel tracker nobody reads. Drop Intercom if nobody messages it. Win 1-2 seconds of LCP — and you just recovered a noticeable slice of your ad budget without changing a single thing in Google Ads. And don't run ads to your homepage — the homepage does a hundred things, an LP has to do one.

5. Decision tree: you have QS 5, where to start

OK, you looked at the QS column. You saw 5. What do you actually do?

First diagnosis, not optimization. Questions in order:

1. How much money does this keyword bring in? If ROAS > 3 and LTV covers it — QS 5 and let it be. Don't touch what's working.
2. Are there any conversions at all? If 0 conversions in 30 days on 50+ clicks — the problem isn't QS, it's the offer, the LP, or the audience.
3. Which one of the three components is actually "Below Average"?

Then — the tree. Walk it top to bottom, don't skip. In 8 out of 10 audits I've done, the right answer isn't the one you thought first.

DECISION_TREE_SLOT

Key idea: QS is the last thing to fix. First the site's conversion rate. Then LP speed. Then campaign structure and matching. And only then — once the first three are healthy — do you look at ad relevance. In most cases you don't even get to the last step: the problem was in the first two, and QS climbed a point or two on its own within a couple of months after the real causes were fixed. Remember, QS is a roughly 90-day rolling metric — don't expect a sharp jump in a week, even if you did everything right.

The most common real cause of QS 5 I see on audits is bad match-type routing. You're targeting broad "dental clinic Chicago" and showing up for "dental clinic for dogs." Few clicks, no conversions, CTR sinks, QS sinks. The fix isn't rewriting ads — it's adding sharp negatives or moving to phrase match. I've seen the same pattern many times: a serious search-terms cleanup and 30-50 precise negatives (not 400 in a panic, but a few dozen precise ones) lifts QS by 1-2 points over the next month or two. Keywords didn't change. Ads didn't change. You just stopped showing up for garbage.

One more thing not in the tree: if you have QS 5 on brand queries — that's either a Google-side bug or you're a new advertiser with no history. Do nothing about it for the first month. Just run. History accumulates, QS jumps up a few notches automatically.

6. 5 bullshit QS tips you already paid for in hours of your life

Top 5 pieces of advice I've watched for 15 years that haven't worked for 15 years. Each one cost somebody weeks of work.

1. "Build SKAGs — one keyword per group." Already covered above. Dead since 2020. Smart Bidding has no data, learning phase never completes, CPA rises. Group 10-15 intent-close keywords into one ad group with RSAs inside.

2. "Pause keywords with QS < 5 to lift your account average." Classic. You pause a keyword that was bringing in a layer of conversions because it's "dragging the average down." The average gets prettier. The money gets smaller. I've seen this move 30+ times — it's the most common self-inflicted wound. Google evaluates QS at the keyword level in the auction. "Average account QS" is a metric for your eyes, not for the algorithm.

3. "DKI is dead — it just produces spam like 'Buy Cheap plumber denver Now'." Not really. DKI (Dynamic Keyword Insertion) is still a real lever if you set it up with care — I actively encourage using it. The "Buy Cheap plumber denver Now" nightmare isn't DKI's fault; it's lazy setup: broad match feeding raw lowercase queries into a CTA headline. Do it right instead: DKI in the H2 or H3 slot (never in the primary CTA), TitleCase capitalization, a fallback text that reads well on its own, and a tightly scoped keyword set that can't produce ugly permutations. Done right, DKI boosts relevance and CTR — exactly what you want. Encourage it. Just don't fire it into a broad-match campaign and hope.

4. "Raise your bid to hit position 1 — that raises CTR, and CTR raises QS." Recursion on your dime. You pay more to pay less. Tried this in 2019 with a logistics client — CPA rose in a month, went back to previous bids. Tried it in 2024 with an education client — same outcome: ROAS caved in 6 weeks. The idea doesn't just not work now — it never worked.

5. "Order a Quality Score audit for $500." People sell "audits" where they open your account, screenshot the QS column, tell you "40% of your keywords have QS < 6," and hand you an invoice. An audit with no tie to CPA/ROAS is a presentation of the obvious. If the auditor didn't show you how much money you're losing because of this and how specifically to get it back — that's not an audit, that's cosplay.

Common thread across all five: reverse-engineering a metric without understanding the formula. You're optimizing a number and ignoring the business.

7. Owner's monthly checklist: 10 minutes, once a month

If you run the ads yourself, without an agency — don't look at QS daily. Don't rewrite ads weekly. Do one pass, once a month, roughly 10 minutes, off this list. The 5th of the month, after morning coffee. No more often than that — the metrics need time to stabilize.

1. ROAS and CPA over the last 30 days vs the previous 30. Dropped more than 15% without an explanation — figure out why. Not "what's up with QS," but "what's up with ROAS."

2. Search Terms Report. Walk through your top 10 queries by spend. All relevant? Find 3-5 irrelevant ones → add as negatives. Find 3-5 that are already converting but aren't yet exact — add as exact match. Three minutes of work.

3. Impression Share Lost. If Lost IS (budget) > 20% on a converting campaign — you need either more budget or narrower geo/dayparting. If Lost IS (rank) is large and ROAS is positive — definitely raise bids on the profitable ones.

4. Auction Insights. Who's new on the list? Whose Impression Share is growing at your expense? Those are the competitors you'll be trading bids with soon. Knowing their faces is half the job.

5. Conversion rate by device. If mobile converts at half of desktop — either there's a UX problem or the form is too long. Fixing a mobile form often gives a layer of conversion lift to the whole campaign.

6. PageSpeed Insights on your top 3 LPs. Mobile version, not desktop. LCP < 2.5s? If not — it's a task for the developer this week. Not for the PPC guy, not for the copywriter. The developer.

7. Bidding strategy on each campaign. Manual CPC on a campaign with 50+ conversions/month — move to Max Conv or Target CPA. Manual bidding in 2026 is a relic, like a fax machine at a notary's office.

8. LTV / CAC. How much did the acquired customer cost (CPA) and how much do they bring over their lifetime. If LTV/CAC < 3 — you're earning too little. Either lift CR on the LP, or lower CPA, or lift LTV. This is the main number that sits outside Google Ads.

9. One RSA per ad group with maximally distinct headlines + descriptions — that's enough. Since 2022 Google officially recommends exactly this, and my tests on clients confirm it: two or three RSAs in one group split the traffic, none of them finishes the learning phase, and bids wander. One RSA with 15 truly different headlines (not variations on one, but actually different hooks) and 4 descriptions gives the system more combinations to optimize against than three half-baked ads. If you have duplicates — pause the weaker ones and feed the main one new assets. You don't need 3-4 copies.

10. One test per month. One. Not ten. One campaign — one hypothesis — 14 days minimum — decision. That's it.

QS isn't on this checklist. Deliberately. It doesn't belong on an owner's decision list. It's a diagnostic number for the operator, like tire pressure for the driver: worth knowing, but you don't drive to work just to check the pressure.