Google Ads invalid clicks: what Google refunds, what it doesn't

“Invalid clicks” is one of the most discussed and least understood concepts in Google Ads. Some advertisers assume Google already cleans out every fake click; others assume it cleans out nothing. Both are incomplete. This guide explains systematically how Google defines invalid clicks, when it automatically does not charge or issues a credit, and when the loss stays with you.

How Google defines invalid clicks

Google classifies clicks on ads that it considers illegitimate as invalid clicks. The definition in the Help Center covers three main groups:

• Manual clicks intended to artificially increase an advertiser's costs or the revenue of the site hosting the ad.
• Clicks generated by automated tools, robots or other deceptive software.
• Accidental clicks that provide no value to the advertiser, such as the second click of a double-click.

The definition matters because it draws the line by intent and source. A competitor's deliberate click, a click farm or a botnet fall inside it. A user who is genuinely interested in your ad but does not buy falls outside it; that click may be “bad”, but it is not invalid.

The main types of invalid activity

Activity that falls under the definition is not uniform. It behaves differently depending on its source and is caught by different methods.

Bots and automation

Simple bots click from data-centre networks without behaving like a real browser and are relatively easy to catch. Sophisticated bots use real browser engines, imitate mouse movement and scrolling, and come from residential IP addresses. This second group is where both Google and third-party tools work hardest, which is why you need to look at a combination of signals rather than any single one.

Click farms

Organisations where real people on real devices click ads for pay. Because the devices and behaviour are human, they can look legitimate at the moment of the click; the giveaway is usually repeated groups of devices, fast exits and location patterns that do not match the target.

Competitors and deliberate manual clicks

A rival or someone with bad intentions clicking manually to drain the budget. Because volumes are low, the chance of slipping past filters is high, but the impact in local, competitive sectors can be large. We cover this separately in our article on competitor clicks.

Publisher-driven invalid traffic

Clicks on Display Network and app inventory made or commissioned by a publisher to increase their own revenue. Google does not pay the publisher or charge the advertiser in such cases, and may close the publisher account.

Accidental clicks

Double clicks, accidental taps while scrolling and similar repeats. Google automatically collapses some of these into a single click. They are more common in mobile app placements.

How Google catches invalid clicks

1. Real-time filters

Every click passes through Google's automated systems before it is billed. Signals such as known bot signatures, abnormal click frequency and known malicious sources are evaluated at this stage. Clicks found to be invalid are removed from your reports and your bill; you never pay for them.

2. Retrospective analysis

Some patterns only become visible over a longer period. For clicks Google identifies as invalid after billing, it applies a credit to your account. These credits appear in the billing section as an “invalid activity” adjustment.

3. Manual review

If you notice suspicious activity as an advertiser, you can ask Google for a review. Google compares the request against its own data and credits the part it finds invalid. The process is described step by step in our refund request guide.

Where to see invalid clicks in your account

In the campaigns table you can modify columns and add “Invalid clicks” and “Invalid click rate”. These columns show the clicks Google filtered and did not charge for. Credits applied later appear not in the campaign report but as separate lines in the transaction history under Billing.

What Google refunds and what it doesn't

The table below summarises the most common situations. The “Covered by policy” column shows whether the policy applies; whether Google detects every individual event is a separate question.

Why do non-refunded losses matter?

The bottom half of the table consists of items that fall outside the invalid click definition but drain the budget all the same. Expecting a refund from Google for these is a mistake; the fix lies in account configuration. Making this distinction saves you from pointless refund requests and lets you focus on genuinely invalid clicks. We cover the budget protection side in our Google Ads budget protection strategies article.

The area Google's filter cannot see

Google has a very strong view of the moment of the click and what came before it: account history, search behaviour, network information. What it cannot see on its own is what happens on your site after the click; a conversion tag only reports the moment of conversion. These patterns can therefore slip past the filters:

• Visits that land and leave without any interaction, yet looked normal at click time.
• The same device returning repeatedly during the day with different IP addresses.
• Bots that fill in forms with fake details; these can count as conversions and mislead your bid strategy.
• Low-volume deliberate clicks hidden behind shared IPs on mobile networks.

This is why a layer that measures post-click behaviour is a complement to Google's filter, not a competitor. It measures the behaviour Google does not see and turns it into evidence.

Why Google's filter cannot catch everything

This does not mean Google's filter is weak; it is cautious by design. Google must also avoid wrongly classing a legitimate user's click as invalid, or publishers and advertisers would lose real traffic. Filters are therefore tuned to keep false positives low and tend to treat borderline cases as valid.

Filters also only work with the signals Google can see. Google cannot know on its own what happened on your site after the click, whether the visitor filled a form with fake details, or how many times the same device came back that day. Data collected on your site is complementary: not against Google, but to shed light on the side Google cannot see.

In short, Google's automatic credit is the first line of defence; your own measurement and, where needed, a review request form the second. Working together, they make the loss both more visible and more manageable.

Invalid clicks versus “bad traffic”

These are the two concepts most often confused. An invalid click is illegitimate by intent or source. Bad traffic is legitimate but useless to you. An example: a student searching “free moving price list” for information who clicks your moving company's ad is not an invalid click. A bot clicking the same ad is. The first calls for negative keywords, the second for detection and refunds.

Risk differs by campaign type

Invalid click risk does not look the same across campaign types. Search campaigns have high CPCs and relatively low volume; this is where competitor and deliberate clicks hurt most. Display Network and app placements have cheap clicks but high volume; accidental clicks and publisher-driven traffic are more visible. Performance Max combines all these inventories in one campaign with more limited reporting and needs separate attention; we cover it in detail in our Performance Max fake clicks article.

This difference also shapes your monitoring strategy. Focusing on the quality of individual clicks in Search, on placement quality in Display, and on the cleanliness of conversion signals in PMax is usually the most efficient approach.

Practical steps to close the gap

1. Add the invalid click columns to your campaigns table and check them weekly.
2. Track invalid activity credits in your billing transaction history.
3. Review the search terms report regularly; cut bad traffic with negative keywords.
4. Set the location option to “people in or regularly in your targeted locations”.
5. Use a protection layer on your site that measures post-click behaviour.
6. File a review request with evidence for clicks you believe Google missed.

wall.click takes care of the last two steps: it scores every ad visit using behaviour, device and network signals, feeds the result back to Google as exclusions and conversion signals, and prepares a periodic evidence report for refund requests. See the full list on the features page.

Frequently asked questions

Do I need to apply separately for invalid clicks?

Not for those Google detects itself; these are either never charged or credited back automatically. You only need to apply for activity you believe Google missed.

My invalid click rate shows zero. Is that normal?

It can happen in low-volume accounts. But a zero rate does not prove there are no fake clicks; it shows Google filtered nothing in that period. Compare it with your post-click behaviour data.

Does Google show which individual clicks were invalid?

No. Google reports the total number and rate of invalid clicks but does not share which individual click was filtered or why. This is a deliberate choice to make it harder for bad actors to reverse-engineer the filters.

My invalid click rate went up. Should I worry?

A rising rate shows Google is catching more invalid activity and not charging for it, which is good news in itself. If suspicious behaviour on your own site is rising in the same period, though, consider that part of the attack may be slipping through and look at click-level data.

Which account do credits go to?

Credits are applied to the billing of the Google Ads account where the invalid activity occurred. Depending on your payment setup, they are deducted from the next invoice or added to your account balance.

Are invalid clicks on Display and YouTube covered too?

Yes, the invalid click policy covers all Google Ads networks. Low-quality placements on the Display Network may not count as invalid under the policy, however; for those you need placement exclusions.