How much of your Google Ads lead volume is junk?
About half. In the three accounts where every lead was graded by the person who had to follow it up, 49% to 54% of what Google Ads counted as a lead was not one the business wanted. Here are the numbers and what I think causes it.
What I counted
Google Ads reports conversions. A conversion is whatever the account was told to count: a form submitted, a phone button tapped, a thank-you page loaded, a file downloaded. Nobody at Google reads the form. So the only way to know how many of those conversions were real leads is for a person to grade each one afterwards. In three accounts, someone did.
- A law firm's intake team grades every website form fill as a quality lead or not, in batches, as part of the work I do for them. It is the same form the ads send people to. Over six batches between November 2025 and September 2026 they graded 353 form fills.
- On my own consultation form, I graded every lead from September 20 to October 11, 2026 as quality, maybe or junk. That is 16 real leads, with one test entry taken out.
- An industrial equipment maker's account counted phone calls from ads as conversions. The call log has a length for each of the 109 calls from January to September 2026, and after I took the account over the owner reviewed the long calls himself.
What the numbers were
| Account | What Google counted | Not a lead | Share |
|---|---|---|---|
| Education law firm | Website form fills, graded by the firm | 174 of 353 | 49% |
| My own consultancy | Consultation form fills, graded by me | 8 junk of 16 (6 maybe, 2 quality) | 50% |
| Industrial equipment maker | Phone calls from ads, by length | 59 of 109 under one minute | 54% |
| Same account, after I took it over | Calls of a minute or more, reviewed by the owner | 5 of 5 were not buyers | 100% |
The law firm's share moved with the calendar. In the batch that covered June to mid-July, 70% of the graded form fills were marked not a lead. In the batch that ran from mid-July into the start of the school year, it was 46%. Same form and the same ads, with different people searching.
On my own form, 2 of the 16 were leads I wanted to call back. Both had a Google click ID on them. So did 10 of the junk and maybe leads. The click ID tells you the ad was clicked. It tells you nothing about who clicked it.
The equipment maker's account is the sharpest case. For nine months it counted a tap on the ad's phone button as a conversion worth $10,000. More than half of the calls that followed lasted under a minute, and once the owner listened to the five longer ones that came in after I started, none of them was someone buying equipment. The report said the account was converting. It was collecting phone calls.
What this does to the number on your report
If your report says $120 per lead and half the leads are junk, you are paying $240 for a real one. Every comparison that uses the reported number, against last year or against another channel, is off by the same factor. The number itself is not wrong. It just counts something other than what you wanted.
It also shapes what Google does next. The bidding learns from the conversions it is given. When half of them are junk, it goes looking for more of the same mix, because that is what it was told success looks like.
Why half of the leads are junk
I see the same causes in nearly every account I open, and the three above have all of them.
The searches
Broad matching puts your ad in front of searches that share words with your keyword and nothing else. In the equipment maker's account, the most expensive calls came from people looking for other companies' support lines. In my own account, 689 of the 1,737 searches Google showed me over a month were from people who were never going to be a client: students looking for a course, people looking for a job, people looking for Google's own support line, people who wanted it free. Those searches get clicks, and some of those clicks fill in forms.
What you told Google to count
A conversion that is easy to satisfy will be satisfied by junk. The equipment maker's old setup counted every phone tap as a $10,000 conversion, so Google's bidding did exactly what it was told and paid up to $146 for a single tap. A thank-you page that loads for anyone who submits anything is the same problem in a quieter form.
Nothing goes back
Google never finds out which form fills became cases or customers unless someone tells it. In most accounts nobody does, so the bidding optimizes toward form fills and the mix stays where it is. Of the three accounts here, only the law firm's grades go back into Google as uploaded quality leads. I don't claim that alone explains its better autumn batch, since the school calendar did part of that, but it is the only one of the three where the bidding has ever been shown a real lead.
The form asks nothing
A form with a name and an email invites everyone. A question about budget or case type does not stop junk, but it lets you see it on the day it arrives instead of after someone has spent an hour chasing it. My own form asks for a monthly budget, and that one answer sorts the inbox before I open it.
What to ask your team
You don't need to log into Google Ads to find out where your account stands. Four questions do it.
- What share of last month's Google Ads leads did sales or intake accept? If nobody can answer within a day, nobody is grading, and the bidding is learning from the junk along with the rest.
- Which conversion actions is the bidding optimizing toward, and what does each one mean? Ask to see the list. A phone tap, a page view and a signed quote should not count the same.
- When were the search terms last read, and how many negatives were added? Every week is the right answer. "At launch" is the common one.
- Do graded leads go back into Google? Offline conversion uploads are how. If the answer is no, Google has never seen one of your real customers.
Method and sample
Three accounts: two clients and my own. The clients are not named, by my choice. Both have seen their own numbers.
- Law firm. Twelve grading batches exist between November 2025 and September 2026. I counted the six where the firm had graded at least 80% of the rows, and inside those only rows marked yes or no: 389 form fills, 353 graded, 36 left blank or marked uncertain and excluded. Counting every graded row in all twelve batches instead gives the same answer, 49%. The form does not record where a visitor came from. In the June to September window Google Ads attributed 182 thank-you-page conversions to the ads against 265 form fills on the sheet, so roughly two thirds of these form fills came through Google Ads by Google's own count.
- My own account. Every lead in my lead sheet from September 20 to October 11, 2026, graded by me on October 12. Quality means someone I would take a call with and could help. Maybe means unclear. Junk means no fit at all. 17 rows, one test submission removed, 16 counted. 12 of the 16 carried a Google click ID. The search term figure is from the same account, September 14 to October 15, 2026, and counts only the searches Google shows; it hides the majority of spend behind "other search terms".
- Equipment maker. Google Ads call reporting for calls from ads, January 1 to September 30, 2026: 109 calls, 59 shorter than 60 seconds. September 28 to October 13, 2026: five calls of 60 seconds or more, reviewed by the owner, none a buyer.
These are small samples from accounts I happen to have graded data for, not a survey of the industry. I will add to this page as more graded leads accumulate, and the date at the top changes only when the numbers do.
Find out what your Google Ads budget is really buying.
A free call, and you'll talk to me.