What are invalid clicks in Google Ads, and do you pay for them?

Google’s ad traffic quality team uses automated filters, machine learning and manual reviews, and says it catches most invalid traffic shortly after the ad interaction. What happens to a click depends on when it’s caught:

• Before the billing cycle ends. Google takes the click out of your campaign’s figures and your bill, and restores its cost to the campaign’s budget to be spent on valid clicks. You pay nothing, so Google says “no refunds are necessary.”
• After you’ve been billed. Google issues a credit in the next billing cycle, “when appropriate and possible.” It appears on your invoice and in your account’s transactions as an Invalid activity adjustment.

To see a month’s credits:

1. 01Go to Billing › Summary.
2. 02In the month’s card, open the Adjustments drop-down. Credits for invalid traffic are labeled Invalid activity.

A credit doesn’t rewrite the month it belongs to. Google freezes a campaign’s figures at the end of the month so they match the invoice, and a credit issued later leaves them as they were. To see campaigns with their credits taken out, open the Report editor from the Campaigns menu and, under Template gallery, select Invalid Activity Credit Report: Search & PMax. It adds each campaign’s credited clicks, credited interactions and credited amount, and covers Search and Performance Max campaigns only.

Two smaller rules sit beside these. Google charges for no more than 2 clicks per ad impression. And in rare cases a report shows more conversions than clicks: a click led to a conversion and was filtered afterwards, so the click went and the conversion stayed.

Google counts the clicks it filters in columns you add to the campaigns table yourself:

1. 01Go to Campaigns › Campaigns.
2. 02Select the columns icon above the table, then the search icon next to Modify columns for campaigns.
3. 03Search for Invalid clicks, and select Apply.

• Invalid clicks and Invalid click rate. The number and the percentage of clicks “classified as invalid and automatically filtered from your account.” Google’s API reference gives the sum behind the rate: the filtered clicks, divided by filtered and unfiltered clicks together.
• General invalid clicks and General invalid click rate. The part caught by routine filtering: known data-center traffic, bots, spiders and other crawlers, and irregular patterns. These two are available for Search and Display campaigns only.
• Invalid interactions. Google’s page on invalid traffic names this column beside Invalid clicks, as traffic filtered before you were billed.

Google says Invalid clicks and Invalid click rate are available on the Campaigns page only, so there is no count for an ad group, a keyword or a placement. Its API reports them for a campaign or the whole account, by day, device or network, and no finer.

They count clicks already taken out: none of them is in your Clicks or your Cost. A campaign showing 960 clicks and 40 invalid clicks had 1,000 clicks made, of which 960 were counted and charged: an invalid click rate of 4%.

So read a high number the way Google does: “An increase in invalid clicks means Google’s defenses actively protect your budget.” It is money you didn’t pay. It also isn’t your credits. The columns count what was filtered before you were billed, a credit covers what was found after, and Google calls the two numbers “not directly related.”

Google says a sudden rise in traffic, or a daily budget spent early, “doesn’t always indicate invalid traffic.” The ordinary causes come first:

• A change in the account. A raised budget or bid, added keywords, most of all general ones, or new audience and location settings can bring more impressions and more clicks. Campaigns › Change history lists what changed, and when.
• Demand. A season, a news event, or a competitor who stopped advertising, which lets your ads show more often.
• A new placement. On automatic placements, Google’s Ad Traffic Quality site says an ad’s traffic “can significantly increase if it starts showing on a new publisher site.”
• Clicks from outside your area. Your ad can show to someone outside the area you target when their search names it. Google’s answer is the Presence setting, which the location targeting guide covers.

Your own analytics isn’t proof either. Google names two patterns there that look like bots and often aren’t:

• Zero-second visits. Many analytics tools need a second click to time a visit, so a visit of zero seconds can be a person who read the landing page and clicked nothing more.
• Many clicks from one IP address. Universities, offices, coffee shops and mobile carriers put hundreds or thousands of devices behind one public address.

Your logs can also count more than Google bills. Google filters a click after it happens, so “Google Ads removes filtered invalid clicks from your reports, but your website analytics might still record the visit.” Its Ad Traffic Quality site adds that “charges are based solely on Google’s metrics,” and that several log entries carrying the same GCLID are charged as one click at most. The GCLID guide covers that click ID.

The same goes for a click-fraud tool. Google says such tools “often flag invalid traffic that Google defenses already detected and filtered,” that it can’t evaluate their reports without knowing their detection criteria, and that it can’t compare with them click by click. So a tool’s count of flagged clicks isn’t a count of clicks you paid for. Set it beside the Invalid clicks column before you act on it.

The Invalid clicks column stops at the campaign, and accidental clicks start somewhere smaller: one site or one app where the ad sits beside something people tap. For a campaign that shows on the Display Network, the place to look is the report of where its ads showed.

1. 01Go to Campaigns › Insights and reports › When and where ads showed and select the Where ads showed tab.
2. 02Set the date range to the last 30 days, have Impr., Clicks, CTR, Cost and conversions in the table, and sort by CTR.
3. 03Take a site or app whose CTR stands far above the rest, and work out the rate of the others: the clicks of every other site and app in the campaign, divided by their impressions.
4. 04Multiply that rate by the placement’s impressions. The answer is the clicks it would have drawn if people clicked there as they do everywhere else.
5. 05Set that beside the clicks it drew, then read its conversions. Clicks that convert as the others do are customers, however many there are.

Brickfall Arcade drew 648 clicks from 15,400 impressions, 4.2%. The campaign’s other 23 sites and apps drew 1,763 from 414,600, 0.4%. At their rate the app’s impressions were worth about 65 clicks, and it drew ten times that.

Clicks are counts, and luck moves a count by about the square root of what was due. Where 65 clicks are expected, anything from about 49 to 82 is an ordinary month. And you aren’t reading one placement. This report lists 24, so a one-in-twenty fluke turns up about once among them. To keep the chance of any false alarm at 5%, each has to be rarer than 0.2%. Twice the expected clicks is that rare only once about 12 were expected. A site with 3 clicks where 1 was due proves nothing: luck gives that 8% of the time.

Then the clicks themselves. On the other sites and apps a click became a sale 1.9% of the time, so 648 clicks would have brought about 12 sales. The app brought none. Its clicks cost $0.10 each where the others’ cost $0.62, which is why $64.80 looks too small to matter.

No other row comes near. And daily-trivia-quiz.example is a different problem: its 381 clicks sold nothing either, at a click-through rate no higher than the rest. That is a placement that doesn’t sell, which the Display campaign guide reads from the same report by what each placement’s spend bought.

One caution before you add up the damage. Google says not to use this report as “an accurate representation of billing”: it has invalid clicks removed only where they were detected immediately, so it can still count clicks Google’s filters took out later and never charged. Treat it as where to look. What you paid is on the campaign’s own row and your invoice.

Performance Max has no such reading. Google’s placement report for it says where the ads served and how many impressions each placement received, and Google says it “shouldn’t be used to evaluate performance.”

A placement like that is yours to remove. Google names placements “where your ads are showing, but not performing well” among the ones to exclude.

1. 01Go to Campaigns › Audiences, keywords, and content › Content.
2. 02Select Exclusions, then Edit exclusions on the right.
3. 03Choose where the exclusion applies: Campaign, Ad group or Account.
4. 04Enter the placement, a site by its URL or an app by its app ID, and select Save.

The Display campaign guide covers the rest: exclusion lists and the account-level exclusion, and excluding mobile apps and app categories.

Where you suspect invalid clicks aren’t being filtered, Google says “you can manually exclude specific IP addresses from triggering your ads.” An exclusion can sit at two levels:

• For the account. Go to Admin › Account settings, open the IP exclusions section, enter the addresses and select Save. Google says this covers all campaigns in the account, Performance Max, Demand Gen, Search, Shopping, Display and YouTube among them.
• For one campaign. Go to Campaigns › Campaigns › Settings, select the campaign, then Additional settings and IP exclusions. A campaign takes up to 500 addresses. Video, hotel, App, Performance Max and Smart Display campaigns have no IP exclusion of their own.

The two lists add up: a campaign is kept from every address on either. An asterisk in place of the last three digits, as in 198.51.100.*, excludes a block of addresses. An address also comes in two forms, IPv4 and IPv6, and Google says to exclude every version of the one you mean.

It is a blunt tool. Google’s own example is a company excluding its office network, so staff checking the search results don’t click its ads. An address isn’t a person: internet providers sometimes assign one to a large number of computers, and mobile carriers rotate them among their users, so an address you exclude today may be a customer’s tomorrow.

If you still believe you paid for invalid clicks, Google will look. Its click quality form asks its specialists to review your traffic from the last 60 days. Google’s page on invalid traffic lists what to gather first:

• Your customer ID, and the exact dates of the activity.
• The campaigns and ad groups where you saw it, and any keywords that drive it.
• From your web server logs: the IP addresses that click repeatedly or never convert, their browsers and devices, and the GCLIDs of the clicks you suspect.
• For invalid leads, a spreadsheet with each lead’s GCLID and what happened when you emailed or called.
• Why the traffic looks wrong, briefly. Google’s example is “a 300% spike in clicks without more conversions.” Where you suspect one site, its Ad Traffic Quality site asks for the publisher’s domain name too.

Google says the review “typically takes several business days,” and that you’ll get an email with its findings. If it finds invalid traffic its filters missed, it tells you of any credit. It won’t say which clicks it judged valid or invalid, since sharing that could show the people behind them how they’re caught.

Fake form submissions are a separate matter. Google’s filters run after an ad serves, so a bot can reach your form whether or not you pay for its click. Google’s advice is on your side of the form: reCAPTCHA, server-side validation, and a double opt-in email.

Every day Goldbeater reads where each of your enabled Display, Video and Demand Gen campaigns’ ads showed over the last 30 days, and sets every site and app’s clicks against what its impressions would draw at the click-through rate of the campaign’s other sites and apps. It raises one that drew over twice those clicks on at least $20, where so many is unlikely after allowing for every site and app tested in the account, and where its clicks converted under half as often as the others’ did, or not at all where the others converted nothing. A site or app whose clicks convert is never raised, however high its rate.

That is the app read above: the finding’s 648 clicks, the 65.5 it expected and its two rates are the table’s first row, and the 24 placements it allows for are the report’s 24. The finding is advice: it says where the exclusion goes, and Goldbeater drafts no change. For a placement you picked yourself it says to take it out of the ad group’s placements, and for a Demand Gen campaign it names the account-level exclusion.

Five more checks look at the same problem from other sides:

• Children’s content, games and job sites. Placements a published list or their own name marks as one of those, where together they cost $20 or more: one finding for a campaign and kind, saying whether they converted far below the rest of the campaign or no measurably worse. On the campaign above it raises Brickfall Arcade as its one placement in games.
• The same three kinds in Performance Max. Raised where together they showed more often than one of the campaign’s placements typically does, by more than chance explains, with their impressions, and saying that Google reports no cost, click or conversion for them.
• A campaign spending far past its usual day. A day, or a run of days to yesterday, each spending at least twice the campaign’s highest day of the four weeks before, on more clicks than chance explains, and at least $20 past its usual day. The finding sends you to Change history first, as Google does.
• A campaign that spent and converted nothing. No conversion counted in 30 days and no sale or lead recorded, on at least three times its target CPA, or what your typical converting campaign pays for a conversion, and at least $20. A campaign that bids for impressions or views is left out. The finding lists what its clicks recorded instead, and counts an invalid-traffic tool’s flags as a figure of their own, apart from engaged visits.
• No exclusion anywhere. A Display or Video campaign, enabled or waiting to launch, that excludes no site, app or video itself or in an ad group and carries no list, on an account that excludes none. And a Display campaign that excludes no app category anywhere, so it can show in any app.

The alerts guide shows a spike and a stop as Goldbeater reports them, and the not-converting guide a campaign whose clicks bring nothing. The audit checklist has every one of these as a line to work through by hand.

What are invalid clicks in Google Ads?

In Google’s words, clicks on your ads that don’t represent a genuine interest in your business: accidental clicks caused by where an ad is placed, clicks from bots and automated tools, and manual clicks meant to raise your costs or a publisher’s income. Google filters the ones it detects out of your reports and your bill.

Does Google Ads charge for invalid clicks, or refund them?

It doesn’t charge for the ones it catches. A click found invalid before the billing cycle ends is removed from your figures and your bill, so Google says no refund is necessary. One found after you were billed comes back as a credit in the next billing cycle, where appropriate and possible, shown as an Invalid activity adjustment on your invoice and under Billing, Summary.

What is a normal invalid click rate?

Google says there is no industry standard for a normal invalid traffic rate. The Invalid click rate column is the share of all clicks, filtered and unfiltered, that Google filtered, and you weren’t charged for any of them. Google reads a rise in invalid clicks as its defenses protecting your budget.

Can a competitor run up my costs by clicking my ads?

Google names competitors who click to increase your spend among the invalid traffic it filters, and it charges for no more than 2 clicks per ad impression. Many clicks from one address aren’t proof on their own, since offices, universities and mobile carriers put many people behind one IP address. If your logs show a pattern, send them with the click quality form.

How do I block an IP address in Google Ads?

For the whole account, go to Admin, Account settings and the IP exclusions section, which Google says covers all campaigns, Performance Max among them. For one campaign, open its settings, then Additional settings and IP exclusions, where it takes up to 500 addresses. Video, hotel, App, Performance Max and Smart Display campaigns have no campaign-level IP exclusion.

How do I report click fraud to Google?

Use the click quality form, which asks Google to investigate activity from the last 60 days. Have your customer ID, the dates, the campaigns and keywords, and from your server logs the IP addresses, devices and GCLIDs of the clicks you suspect. Google says a review typically takes several business days and ends with an email. It won’t tell you which clicks it judged invalid.

Why does my click-fraud tool report more invalid clicks than Google Ads?

Google says third-party tools often flag traffic its own defenses already detected and filtered, so part of what a tool counts was never billed. It can’t evaluate a tool’s report without knowing its detection criteria, and it won’t compare click by click. Your own analytics differs for the same reason: it records the visit, and Google removes the click afterwards.