What the table saysquoted in guide after guide
1 − 5/10 = 50% off your click
Every row fits: 44.4 at nine, 37.5 at eight, 28.6 at seven. The five is the average score the whole thing is normalised to. Anybody can derive it in a line, which is why it spreads and why nobody cites it.
What it assumestrue only while nothing moves
the Ad Rank you must beat stays put
A better score often moves you up a position, where the advertiser above you is harder to beat. One guide of nine says it: you could pay more than before. Watch cost per conversion, not cost per click.
The discount table everybody quotes is not published by Google and is not folklore. It is Ad Rank rearranged, with your reported score put in the quality slot and the competitor held still. The first of those two is the one that costs you.
Where the numbers come from
Every row is one minus five over your score.
It comes out of the Ad Rank arithmetic. If Ad Rank is your bid multiplied by a quality term, then to beat the same competitor you need bid × quality to stay constant, so doubling quality halves the bid you need. Hold the competitor still, put your reported Quality Score in the quality slot, normalise to 5, and the table falls out. Anybody can derive it in a line, which is why it spreads.
The two things it assumes
The arithmetic is sound. What it rests on is not stated anywhere the table is quoted.
The first line is the one that costs money. Better quality can raise your cost per click and lower your cost per conversion in the same week, and only one of those two numbers is in the table everybody quotes.
Google says both things
The easy version of this article blames the guides. That version does not survive contact with Google's own help centre, which makes the cost claim itself: "Higher ad quality generally leads to better performance, including better ad positions and lower cost." The same page, a few lines down, says "these scores are not inputs in the ad auction."
So the nine guides are not ignoring the documentation. They are repeating its shorthand and dropping its caveat.
What I checked
I opened eleven pages across two searches on 7 September 2026. Americaneagle blocks automated fetches, so ten were read, and one of those is Google's own marketing article rather than a guide. That leaves nine.
All nine connect Quality Score to cost or position. One of the nine says it is not a key performance indicator, quoting Google. One draws the live-versus-published distinction without using Google's phrase, calling the dashboard number "a historical snapshot not a live score". One page prints the discount table in full, and none of the nine derives it.
That last count is the one that matters. The table is the most quoted number in the subject and the arithmetic behind it fits in a sentence, and no page in the sample spends that sentence.
What to do instead
Read it keyword by keyword, never as an average
Low score plus high volume points at one ad group worth fixing. An account average of 7.2 is a mean of diagnostics, weighted by nothing.
Expect the position effect
When quality improves, watch cost per conversion rather than cost per click, because moving up a position can raise the click price while the economics get better.
Work the three component columns
Expected click-through rate, ad relevance and landing page experience each carry their own above or below average status, and those tell you which part is dragging.
Where this sits in our own tool
Google Ads Vitals shows no Quality Score anywhere. I would like to say that was principle, and the honest version is smaller: I chose a campaign level export for other reasons, and the score lives at keyword level, so it fell out as a by-product.
The argument above says a monthly page is the wrong place for it, so the outcome is right. It was not a decision, and the difference is worth writing down rather than letting the absence read as design.
Related reading
- Performance Max reports arrived without their labels is the same shape: a number computed one way and read another
- Google Ads counts your leads twice by default
- Auction insights, and what the share numbers are share of
- Breakeven ROAS: the one field no calculator asks for
Questions people actually ask
It is arithmetic, not a Google publication. Every row is one minus five over your score: 1 - 5/10 is 50 per cent, 1 - 5/9 is 44.4, 1 - 5/8 is 37.5. It falls out of Ad Rank if you hold the competitor's Ad Rank fixed and normalise to the average score of 5.
It is right about the direction and wrong about the situation it describes. It assumes the Ad Rank you have to beat does not move, and a better score often moves you up a position where the advertiser above you is harder to beat. One of the nine guides says so: you “could actually pay more than you were before”.
Better ad quality does. Google's own help centre says higher ad quality generally leads to better positions and lower cost, and the same page says the scores are not inputs in the ad auction. The improvement is real; the score is the readout of it rather than the cause.
The question treats it as a target, and Google's page says it is not a key performance indicator and should not be optimised or aggregated. It is most useful where the score is low and the volume is high, because that pair names an ad group worth an afternoon.
An account average is a mean of diagnostics weighted by nothing, so it belongs in the working view rather than the monthly page. Watch cost per conversion instead, which survives the position effect that makes cost per click move the wrong way after an improvement.
Expected click-through rate, ad relevance and landing page experience, each with its own above average, average or below average status. They are the visible part of the auction's quality assessment, which also uses signals the score never shows you: device, location, time of day and assets.
WRITTEN BY
Olha · clinic data analyst
I build the reporting our managers open every morning at a multi-branch medical clinic — and package it so other practices don't have to start from scratch.
Published on 7 September 2026. Three limits worth stating outside the body text. The sample is two search phrasings on one day, and one of the eleven refuses this location, so rerun the method rather than trust the counts. I have not proved the discount table is wrong, only that nobody printing it says where it came from and that Google’s own documentation contradicts the mechanism it implies; those are different claims and the article keeps them apart. And this piece is deliberately built out of blocks the rest of this blog had never used, because four articles in a row had come out of the same mould.