TLDR
TrafficGuard audited 40,021 ChatGPT ad clicks across 47 enterprise advertisers and found a 4% blended invalid rate, well below Google's 10.8% on the same campaigns. Clicks from server, proxy and malicious infrastructure ran 3.6 times higher on ChatGPT than on Google, and impossible device configurations appeared nearly ten times as often.
KEY TAKEAWAYS
The numbers and what the gap means for spend
TrafficGuard, the ad-verification platform of Australian Securities Exchange-listed Adveritas Ltd (AV1), analysed 40,021 paid clicks on ChatGPT ads across 47 enterprise advertisers over three weeks from 24 August to 14 September 2026.[1] The blended invalid-click rate came back at 4%, or 1,592 clicks. On the same advertisers' Google Ads campaigns, the invalid rate was 10.8%.
ChatGPT's blended rate is 63% lower than Google's, yet the type of fraud arriving on ChatGPT is harder to spot and almost entirely non-human in origin. By advertiser, the range ran from 0.1% to 34%.[1]
TrafficGuard chief executive Mathew Ratty said the worst-performing account had nearly 90% of its flagged clicks coming from server infrastructure with no human user behind it. OpenAI began testing ads in ChatGPT on 9 February 2026, initially for users in the United States, then expanded to Australia from April 2026.[2] TrafficGuard's three-week window ran roughly six months after the channel opened.
Machine traffic composition
Clicks from datacenter and proxy infrastructure arrived 3.6 times more often on ChatGPT than on the same advertisers' Google campaigns, and clicks from impossible device configurations appeared nearly ten times as often.[1] In raw terms: the hosting, proxy and malicious IP rate was 3.34% on ChatGPT against 0.92% on Google. Impossible device configurations ran at 0.86% against 0.09%.
Machine-origin fraud is harder to filter without dedicated verification tooling. A server cycling through impossible device signatures at scale generates clicks that a platform's own dashboard may count as real.
Ratty said the pattern was coordinated and persistent. "Even though ChatGPT advertising is in its early days, automated traffic has already found it," Ratty said. "It is coordinated, it is persistent and it is clicking every day. Most advertisers have no idea how much of their spend is going to it."[1]
TrafficGuard chief product officer Miguel Lopes said OpenAI is not uniquely exposed to the problem. "OpenAI is not uniquely vulnerable, it is simply the newest place to sell clicks, and low-quality traffic follows the budget wherever it goes," Lopes said.[1]
Four hosting and proxy networks accounted for roughly half of all invalid clicks recorded, reaching 25 of the 47 advertisers studied.[1] Those four networks produced 742 flagged clicks in total.
| Metric | ChatGPT Ads | Google Ads |
|---|---|---|
| Blended invalid-click rate | 4.0% | 10.8% |
| Hosting / proxy / malicious IP rate | 3.34% | 0.92% |
| Impossible device configurations | 0.86% | 0.09% |
| Total clicks analysed | 40,021 | 5,594,614 |
| Invalid-click range by advertiser | 0.1% to 34% | Not reported |
Australian advertisers on a six-month-old channel
Australia received ChatGPT ads from April 2026, so local buyers have operated on the channel for less time than the US cohort. Click-through rates (the share of ad impressions that result in a click) here are modest: Australian ChatGPT ad click-through sits at 0.67%, roughly one-third the rate recorded in Brazil.[3] Cautious local spending limits absolute dollar exposure to fraud, though it does not change the composition of the traffic that does arrive.
TrafficGuard is the flagship software-as-a-service product of Adveritas Ltd (ASX: AV1), as disclosed in the company's February 2026 ASX filing.[4] The report was dated 30 September 2026. OpenAI has not commented publicly on whether clicks flagged by TrafficGuard's analysis were charged to advertisers, and no external audit has verified TrafficGuard's methodology.
The sample of 40,021 ChatGPT clicks compares with 5,594,614 Google Ads clicks from the same advertiser set. For media buyers running pay-per-click campaigns across both platforms: the bigger bill for wasted spend is still on Google, but the fraud arriving on ChatGPT is the kind that standard platform reporting is least equipped to catch, and TrafficGuard's numbers argue for third-party verification before ChatGPT budgets are scaled.
SOURCES & CITATIONS
FREQUENTLY ASKED QUESTIONS
Is ChatGPT ad fraud worse than Google Ads fraud?
When did ChatGPT ads launch in Australia?
Who ran this analysis and are the figures independently verified?
Xaviery Malinao writes for Prompt the Market on how brands and agencies are adapting to answer engines, drawing on Bushnote's work with clients across search, AI search and content.
