Retail advertisers are heading into the most automated holiday season on record, and new data suggests the traffic filling those automated campaigns is getting measurably worse.

Lunio's Invalid Traffic Impact Report for Retail analyzed more than 414 million ad clicks across 88 retail ad accounts between Q4 2025 and Q2 2026, the largest dataset in this report series to date. The average invalid traffic (IVT) rate across the sector was 5.00%, and it climbed in every quarter measured. The single most consequential finding sits inside Google Search: campaigns running Google's AI Max feature suite went from the cleanest traffic in the channel to the dirtiest in the space of nine months.

This article covers the key findings, what they mean for retail paid media teams, and what to check in your own accounts before Q4 budgets surge. The full report, including methodology, placement tables, and channel-level data, is free to download.

What Counts as Invalid Traffic?

Invalid traffic (IVT) refers to any ad clicks or website visits that don't come from a real person with genuine interest. That includes bots (both benign and malicious), accidental clicks, competitor clicks, and other zero-intent visitors with no chance of ever converting into customers.

Ad platforms filter some of this activity, but the data in this report comes from accounts running in unprotected, monitor-only mode. In other words, it measures the invalid traffic that gets past the platforms' native defenses and reaches advertisers anyway.

And that matters even more than it used to. Here's why:

LLMs have collapsed the cost of generating sophisticated fake traffic. An agent that runs a synthetic browsing session, produces a plausible click pattern, and passes basic post-click engagement checks can now be deployed for cents per session. The cost of defense has stayed roughly flat while the cost of offense has fallen through the floor, and retail's product feeds, promotions, and affiliate programs make it a natural target.

Retail IVT Is Rising Every Quarter

The headline 5.00% average understates the trend. Retail IVT stood at 4.18% in Q4 2025, climbed to 5.28% in Q1 2026, and reached 5.54% in Q2, the highest figure in the dataset.

Counterintuitively, the holiday quarter was the cleanest quarter in relative terms.

The most plausible explanation is dilution. Q4's surge of genuine shoppers swells the denominator, so the same underlying invalid activity accounts for a smaller share of total clicks. The absolute numbers back this up: invalid clicks held at roughly 6.6 million in Q4 and kept growing to 7.2 million in Q2, even as total traffic fell from its holiday peak.

And there is a lesson buried in that. If your traffic quality metrics look better in November, treat the improvement with suspicion. Invalid volume rarely takes the holidays off - it just gets harder to see.

IVT by Ad Platform

Across the four major platforms in the retail sample:

• Meta recorded the highest average IVT of any platform at 5.99%, peaking at 6.61% in Q1 2026. With more than 83 million clicks in the sample, this is the largest concentration of invalid clicks outside Google by absolute volume. It also tracks with Meta's own transparency reporting, which estimates fake accounts at roughly 3% of monthly active users, more than 100 million accounts at Meta's scale.
• TikTok averaged 5.56% and posted the sharpest climb of any social channel: 3.94% in Q4 2025 to 6.61% in Q2 2026, a 68% increase. As TikTok Shop matures into a core retail acquisition channel, the trajectory matters more than the average.
• Google averaged 4.75%, lower than Meta and TikTok but on a far larger sample (318 million clicks, 77% of the dataset) and with a clear upward trend, from 3.95% to 5.43%. Because Google carries most retail budgets, that drift represents more wasted spend than any other platform's headline rate.
• Bing (Microsoft Ads) was the only platform to improve, easing from 4.70% to 4.18%. Retail spend on Bing skews toward Shopping and branded search rather than broad audience-network prospecting, which likely limits exposure.

Inside Google: Every Major Channel Got Worse

Every major Google channel in the retail sample ended the period with a higher IVT rate than it started with. The differences are in how fast.

Google Shopping carried the highest average IVT of any Google channel at 6.33%, climbing 80% from 4.16% to 7.51%. By the end of the period, roughly one in thirteen Shopping clicks was invalid. Feed-driven ads serve against high-purchase-intent product queries, which is exactly the traffic that competitor scrapers, price-monitoring bots, and affiliate arbitrage operations target. One example from the report: when KURU Footwear ran its initial traffic audit, its primary branded Shopping campaign carried a 19.85% IVT rate, more than three times the account average at the time.

Google Display recorded the steepest climb, more than tripling from 2.45% to 8.13% in six months. Two forces drive it: holiday dilution unwinding, and a 68% contraction in Display volume that left remaining spend concentrated in always-on network inventory.

Google Search averaged 4.73% and rose every quarter, a 53% increase across the period on 114 million clicks. Search has traditionally been the channel retailers trust most. The data suggests that trust needs re-examining, and the AI Max analysis below identifies the fastest-growing reason why.

Google Demand Gen averaged 4.59% but more than doubled from 2.81% to 5.95%, the same pattern seen in other sectors as advertisers broaden Demand Gen audience targeting.

Performance Max looked calm at 4.43%, the lowest and flattest of the major channels. Two caveats apply. PMax is the largest channel in the sample at 150 million clicks, so even a low rate means more invalid clicks in absolute terms than Shopping and Display combined. And because PMax blends Search, Shopping, Display, YouTube, Discover, Gmail, and Maps into a single figure, clean and dirty inventory average each other out in ways advertisers cannot fully separate. Google's recent transparency improvements only allow blocking of Search and Shopping inventory, not the Display, YouTube, and app placements where a disproportionate amount of IVT hides.

The AI Max Effect: Automation's Rising Invalid Traffic Cost

Google introduced AI Max for Search campaigns in May 2025: a one-click feature suite that expands existing Search campaigns beyond their keywords, using broad match and keywordless AI to match ads against queries the advertiser never specified. It left beta in April 2026, and Google describes it as its fastest-growing AI-powered Search ads product.

Retail has adopted it faster than any sector Lunio tracks. AI Max went from a third of all retail Search clicks in Q4 2025 to a majority (55%) by Q2 2026, even as total Search volume fell.

The report contains the first robust, like-for-like measurement of what that shift is doing to traffic quality. The comparison is between retail Search campaigns with AI Max enabled and retail Search campaigns without it, across the same customer base and the same nine months. Both cohorts are Search campaigns, so campaign type, sector, and season are held constant. The only structural difference is the feature itself.

How much invalid traffic does AI Max carry?

When retailers first adopted AI Max, their campaigns were markedly cleaner than standard Search: 2.46% versus 3.72% in Q4 2025. Nine months later the picture had fully inverted. AI Max campaigns more than doubled to 5.28% (a 114% increase), while standard Search campaigns in the same accounts held broadly flat and ended the period 17% lower than they began.

By Q2 2026, AI Max campaigns carried an invalid traffic rate 72% higher than standard Search. The concentration of waste flipped too: AI Max went from a third of Search clicks and a quarter of invalid clicks in Q4 2025 to 55% of Search clicks and 68% of invalid Search clicks by Q2 2026.

Mechanically, this is what AI Max's design predicts. Its core features widen a campaign's reach into queries the advertiser never chose and pages the advertiser never nominated. The gap between the two cohorts is, quite precisely, the cost of everything AI Max reaches that standard Search does not. That additional territory is measurably dirtier.

Two caveats: First, this is an observational comparison, since advertisers choose whether to enable the feature. But the crossover pattern cuts hard against a simple selection story: the campaigns that adopted AI Max started cleaner, then deteriorated quarter after quarter within the same accounts. Second, some AI Max traffic is genuinely incremental, reaching demand standard Search would never capture. Part of the elevated IVT is the price of that new inventory. The point is to weigh the incremental reach against the incremental waste rather than assuming the reach is clean.

Worth noting: Google's headline performance claim for AI Max, 14% more conversions or conversion value at similar CPA/ROAS, is explicitly sourced from non-retail advertisers. Retailers evaluating the feature sit outside the population that benchmark describes.

The September 2026 auto-upgrade

From September 2026, Google begins automatically upgrading eligible legacy Search campaigns to AI Max, including those using Dynamic Search Ads, automatically created assets, and campaign-level broad match.

Many retail advertisers who never opted in will be moved onto the feature in the weeks immediately before Black Friday, at the exact moment their budgets and their exposure peak. Auditing traffic quality before that convergence, rather than after the Q4 invoices land, is the single most time-sensitive recommendation in the report.

The Problem Placements: Junk Games and an MFA Network

The report ranks the 25 worst-performing retail ad placements by IVT rate. The single worst offender, a mobile VPN app, recorded an average IVT rate of 90.8%. Nine in ten clicks from that placement were invalid.

The list falls into two recurring groups. The first is low-quality mobile gaming and utility inventory: disposable casual-game apps and free-VPN utilities running multiple ad units per screen, some positioned to catch accidental taps mid-game, with IVT rates from the mid-60s upward.

The second is more distinctive: a cluster of templated, business-software-themed content domains with near-identical naming patterns, all sitting in a narrow 52 to 57% IVT band. This is the signature of a made-for-advertising (MFA) network, batches of interchangeable domains built to arbitrage ad spend while masquerading as legitimate editorial inventory. The Association of National Advertisers estimates MFA sites drain around 15% of total programmatic ad spend.

Every placement in the named table sits above 50% IVT, and the list skews heavily toward mobile app inventory. That reflects where retail prospecting budgets actually flow: into app-network and automated Display reach at enormous scale, through exactly the channels where manual placement reviews cannot realistically keep up.

What Retail Marketers Should Do Before Q4

The Commercial Stakes

For a retail business spending $5 million per year on paid media at a representative $3.70 CPC, a 5.00% IVT rate translates to roughly $250,000 in directly wasted spend annually.

Factor in a conservative 3:1 ROAS and the lost revenue opportunity approaches $750,000 a year.

And that's before the second-order damage: remarketing pools seeded with fake visitors, smart bidding trained on polluted conversion data, and forecasting built on add-to-carts that never came from a human.

The full report covers all of the above in detail, including the complete quarterly datasets, the full top-25 placement table, campaign-level findings, and methodology.

Frequently Asked Questions:

What is the average invalid traffic rate in retail advertising?

Across 414 million retail ad clicks analyzed between Q4 2025 and Q2 2026, the average invalid traffic rate was 5.00%. The rate rose every quarter, from 4.18% in Q4 2025 to 5.54% in Q2 2026, the highest level recorded for the sector.

Does Google AI Max increase invalid traffic?

In Lunio's retail dataset, Search campaigns with AI Max enabled carried a 72% higher invalid traffic rate than standard Search campaigns in the same accounts by Q2 2026. AI Max campaign IVT rose 114% over nine months while standard Search fell 17%. The comparison is observational, but because both cohorts are Search campaigns in the same accounts over the same period, the AI Max feature is the only structural difference between them.

Which ad platform has the highest invalid traffic rate for retail?

Meta recorded the highest average IVT rate of any major platform in the retail sample at 5.99%. TikTok averaged 5.56% and rose fastest, Google averaged 4.75% on the largest sample, and Bing (Microsoft Ads) averaged 4.46% and was the only platform to improve across the period.

Which Google campaign type has the most invalid traffic?

Google Shopping carried the highest average IVT rate of any Google channel at 6.33%, rising 80% across the period. Display recorded the second-highest average at 6.05% and the steepest climb, more than tripling from 2.45% to 8.13% in six months.

When does Google auto-upgrade Search campaigns to AI Max?

From September 2026, Google begins automatically upgrading eligible legacy Search campaigns to AI Max, including campaigns using Dynamic Search Ads, automatically created assets, and campaign-level broad match. For retail advertisers, this lands in the weeks immediately before Black Friday 2026.

Why does retail IVT look lower during the holiday quarter?

Because of dilution. The surge of genuine shoppers in Q4 swells total click volume, so the same underlying invalid activity accounts for a smaller share of clicks. In the dataset, invalid click volume held at roughly 6.6 million in Q4 2025 and grew to 7.2 million by Q2 2026 even as total traffic fell from its holiday peak.

How much does invalid traffic cost retail advertisers?

For a retailer spending $5 million annually on paid media at an average $3.70 CPC, a 5.00% IVT rate equates to roughly $250,000 in directly wasted ad spend per year, and approximately $750,000 in lost revenue opportunity at a conservative 3:1 ROAS.