Getting your Google Ads money back.

ValidVisit doesn't file this claim for you — Google only accepts invalid-activity reports from the advertiser. What it gives you is the case: visit-level scores, dated counts per campaign, and the source breakdown reviewers ask for. Everything below is the sequence for putting that case to work.

Google runs two separate protections, and conflating them wrecks more claims than weak evidence does. Automatic filtering happens before billing: clicks the system classifies as invalid in real time never reach your invoice and never appear as a line item. Invalid click credits come after the fact: a click passes the live filter, gets billed, and is later reclassified during extended analysis, at which point the cost is credited back automatically. You file nothing to receive routine credits — they appear on their own, and they offset future spend rather than returning cash.

The credit is a floor, not a ceiling. Traffic engineered to look human — automated browsers dressed up as ordinary devices, clicks routed through residential proxy networks, human click farms, slow automation that paces itself like a reader — regularly passes the live filter, lands on your page, and stays billed. Those are the clicks a claim exists for. The precise term for all of it is invalid traffic, which covers deliberate abuse, automation and junk placements alike; framing your claim in those terms, not as an accusation of fraud, is what a reviewer can work with.

Hold one more framing steady before you write anything to Google: platforms grade their own homework. That is not an accusation — it is a structural fact about any system that both sells the clicks and polices them, and Google does not publish the thresholds it applies. An independent count is your side of the argument, which is why the work starts with what invalid traffic actually is and the broader Google Ads click fraud picture, not with the complaint form.

File nothing until the evidence file exists. Reviewers respond to documented, dated, specific records — not to a bounce-rate screenshot and a suspicion.

Keep the units straight, because the claim depends on it. Google bills clicks; what proves a click invalid is what happened after it — the visit. ValidVisit scores every visit that lands from your campaigns 0–100, weighing more than 100 independent data points spanning the network the visit arrived through, the device behind it, and how the visitor behaves on the page. Each scored visit ties back to the click that produced it through gclid and your campaign tokens via the tracking token library, so the file speaks Google's own attribution language. The scoring method itself is documented at how it works.

Assemble all of the following before you open the form:

• The affected campaign, ad group and keyword, with an exact date range. Vague windows get vague reviews.
• Timestamped records of the suspect clicks and the visits they produced, so patterns can be cross-referenced against Google's own records.
• Geographic distribution against your targeting — clicks from regions you never bid on, or a spike from one country with zero conversion history.
• Repetition — the same sources hitting the same keyword inside short windows, across days.
• Engagement evidence — sessions with no scroll and no movement, calls to action fired faster than a person can read a headline, durations measured in seconds.
• Network origin — a concentration of visits tracing to server farms and known server ranges instead of ordinary home and mobile connections.
• The per-visit quality scores for the campaign and window, filtered so the low-scoring share and its slice of spend are unmistakable.

Export it as a dated CSV per campaign. An undifferentiated IP list gives a reviewer nothing to act on; a scored, timestamped, source-attributed export gives them exactly the structure they can check against their own data.

There is no report-click-fraud button. The route is a billing inquiry through your own account, and it is the only route Google accepts.

1. Sign in to the Google Ads account that was charged.
2. Open Help and choose Contact us.
3. Select Billing and payments, then the invalid clicks or suspicious traffic topic.
4. State the campaign, the exact date range, the approximate number of suspect clicks, and the pattern — a volume spike from one placement, a cluster of clicks with zero on-site engagement, sources reappearing across days.
5. Attach the evidence export — timestamped visits, quality scores, source breakdown.

Write it like an incident report. Describe the anomaly in numbers and dates, not adjectives. A reviewer can act on "campaign X, June 3–17, clicks concentrated on two sources, the resulting visits scored in the bottom quality band with no on-page engagement." A reviewer cannot act on a general sense that the traffic feels fake.

Set expectations before you submit:

• Google may credit documented invalid activity it agrees with, applied against future spend — not returned as cash.
• Google will not share the internal thresholds it used, and will not give you a click-by-click breakdown of the decision.
• No credit is guaranteed for any claim. The decision is Google's, every time.
• Clicks already showing in your invalid-clicks columns were filtered before billing — focus the claim on charged clicks only.

Timelines are not published; responses commonly arrive within a few business days to two weeks. File promptly — the further you are from the click dates, the thinner your supporting data gets. Review adjustments monthly and aim to file within 60 days of the period in question.

Run the claim dual-track. While the review is pending, exclude the flagged placements and ranges yourself — the sections below cover how — so the same sources are not billing you while Google deliberates.

Verify before and after you file. Three places in the account tell the story.

1. Add the invalid-traffic columns. In any campaign table, open the columns picker, search for invalid, and add Invalid clicks and Invalid click rate. These figures come from the pre-billing filter — clicks removed before they were charged. They are context, not credits.
2. Check the billing summary. Under Billing, the summary shows an adjustment line labeled Invalid activity — the post-hoc credits, listed as a negative amount against the period's charges.
3. Check transactions. Filter the transactions view to adjustments for dated, itemized credit entries you can map to specific campaign periods.

Two caveats keep the ledger honest. First, an invalid click credit offsets future ad spend; it is not a cash refund. Second, a credit repairs the invoice, not the data — invalid clicks that fired conversion events already fed Smart Bidding before the credit landed, and no adjustment un-teaches the model. A click fraud refund recovers past spend; cutting the source protects every campaign after it, which is why the exclusion work below matters more than the credit itself.

Then compare Google's number against your own. Your independently scored visit count for the same campaign and window is the audit line: when your low-quality share runs well past what the credit implies, that gap is your next claim, already documented. Trend the gap per campaign against your own history — your own baseline, not an industry average — and within a cycle or two you will know exactly when filing is worth the effort.

IP exclusion is the cut-the-source lever built into Google Ads. Use it — with a clear view of what it can and cannot do.

1. Open the campaign and go to Settings.
2. Expand Additional settings.
3. Open the IP exclusions panel.
4. Add single IPv4 addresses or CIDR ranges.
5. Save, then repeat for every campaign that needs protection.

Know the constraints before you build the list:

• The cap is roughly 500 exclusions per campaign. Against a coordinated source spread across many subnets, that fills fast — spend the slots deliberately.
• IPv4 only. Suspicious traffic arriving over IPv6 cannot be excluded here at all.
• Exclusions are campaign-scoped. There is no shared account-level list; Search, Display and Performance Max each need their own, maintained separately.
• Propagation takes hours, and clicks from an excluded address can still land in that window.
• Exclusion never refunds anything. It is forward-looking only — it stops your ads being shown to that address again. Money already spent comes back only through the credit process above.

Then know where the lever actually bites. It works against persistent offenders — known server ranges that send consistently bad traffic from a stable block. It fails against rotation: operations cycling through thousands of addresses have moved on before your entry propagates. And it backfires against residential proxy traffic, where the address belongs to a real household — banning the range excludes genuine customers who happen to share the block. When the problem lives at a specific Display site or app rather than an address, use placement exclusions instead; that acts on the source rather than a symptom.

Broad ranges deserve extra care. A wide CIDR block covers thousands of addresses, and the interface does not show you the blast radius. Exclude the narrowest range the evidence supports.

Exclusion lists built on gut feel fill 500 slots with noise. Run exclusion as the output of a detection process, not the process itself.

1. Instrument the landing pages. The pixel scores every arriving visit from day one and attributes it to your campaign tokens.
2. Let several days of data accumulate, then sort the source and placement report by low-score share. The worst offenders separate from the pack quickly.
3. Exclude the worst placements first. For Display and Search Partners traffic, placement exclusions usually out-earn IP entries because they act at the source.
4. Spend IP slots only on persistent, stable, low-scoring ranges — the server-farm blocks that keep reappearing. Skip anything that looks residential.
5. File the credit claim with the same export that drove the exclusions. One evidence file feeds both tracks.
6. Review weekly. Sources rotate, placements deteriorate, and last month's clean zone is this month's problem. Prune entries that stop mattering so the slots stay free.

The division of labor is deliberate and worth stating plainly. ValidVisit detects, scores and attributes — it never blocks a click, never edits a campaign, and never pushes an exclusion on your behalf. Every flagged visit is held with its score and its source so you act on evidence, in your own account, at your own bar. Scoring happens after the visit arrives, on your page — nothing sits in the click path, and nothing about your funnel changes. The Google-specific setup is covered in click fraud on Google Ads, and the underlying method in bot traffic detection.

Most invalid traffic is impersonal — automation and low-grade placements monetizing volume. A different profile earns a different response: the same handful of sources returning after exclusion, clicks concentrated on your highest-value keywords, activity clustered in business hours in your own market, budget drained early in the day so your ads go dark for the traffic that matters.

That pattern reads less like background noise and more like a deliberate campaign against your spend, and it changes the play. Keep every dated export — recurrence across billing periods is itself evidence, and each claim you file strengthens the next. Tighten the loop from weekly review to daily during an active wave. And treat the investigation as its own discipline: the competitor click fraud chapter covers how to distinguish a rival's clicking from organic invalid traffic, and what the pattern file for it looks like.

For network-side repeat offenders — a Display placement or Search Partners domain that keeps shipping low-scoring visits no matter what you exclude — escalate past your own account settings. Placement-level evidence with per-visit scores and dates is exactly what a network's traffic-quality team can act on, and it beats a campaign-level complaint every time.

The through-line of this playbook is the file. Credits are Google's decision, exclusions decay, sources rotate — but a running, dated, source-attributed record of scored visits per campaign survives all of it. Build it once, and every refund claim, every exclusion list and every escalation draws from the same well.