Advertisers are only just beginning to explore the opportunity presented by ChatGPT Ads, and automated traffic has wasted no time finding it.
TrafficGuard’s recent analysis of early ChatGPT Ads account data across 47 advertisers found that up to one in three clicks on some campaigns was invalid, with rates ranging from 0.1% to 34% by advertiser. Automated invalid traffic (IVT) is already reaching campaigns, consuming budget that advertisers expect to be reaching genuine potential customers. Much of it came from a small number of coordinated, automated sources that appeared within weeks of the channel’s launch.
It is a familiar problem arriving in a completely new environment. Wherever advertising budgets move, invalid traffic follows, and every major ad platform has contended with it in its early days. The rapid growth of AI chatbots has also created another valuable target. IVT can generate clicks relentlessly, often without advertisers realising how much of their spend is being consumed by non-human activity.
On a channel this new, the consequences go beyond wasted budget. With few performance benchmarks currently established by marketing teams, distinguishing genuine user behaviour from invalid activity becomes harder, and bad traffic risks becoming the data that advertisers use to decide whether ChatGPT Ads are actually working.
Distorted Metrics and Wasted Spend
ChatGPT ads have given advertisers the chance to establish themselves in a brand new and rapidly growing market. With new markets come new budgets, and invalid traffic follows the money. ChatGPT Ads is simply the latest channel to attract it, as search, social and programmatic did before it.
The problem is that with how new ChatGPT Ads are, there isn’t enough historical data to compare performance against. The benchmark for metrics like click-through rate (CTR) or conversion lag hasn’t been established yet.
This makes it even more difficult to notice invalid traffic is impacting results. If a campaign receives an influx of clicks from a datacentre, it can be written off as an early test as there’s no data to check it against.
Detecting IVT is already becoming more challenging. AI has made it easier for bad actors to disguise automated traffic as human users to carry out click fraud. More sophisticated bots can probe a brand’s protection and, if a behaviour is flagged as “suspicious”, adjust so their future clicks and impressions appear more human.
This behaviour can include hovering over buttons, making and then fixing typos, scrolling or pausing to mimic reading and making curved mouse movements instead of straight lines. Basic filters often miss sophisticated IVT like this, allowing it to affect a brand’s campaigns without raising alarm bells.
Residential proxies and datacentre infrastructure allow automated traffic to operate at scale, and the cost adds up. Juniper Research estimated that $172 billion in global ad spend will be lost to ad fraud by 2028.
A pay-per-click (PPC) campaign has a limited budget and clicks from IVT can drain it, so the ad stops serving before genuine human users see it. At the top of the range we observed, an invalid click rate of 34% on a $20,000 campaign would mean roughly $6,800 worth of clicks that will never convert.
The damage doesn’t end there however, as the even bigger risk of IVT is the impact it has on campaign decisions. ChatGPT Ads offers conversion optimisation to target the clicks that are most likely to convert. However, if invalid clicks make their way into a campaign’s data, any bidding system can learn from these false signals and optimise in the wrong direction, prioritising traffic that won’t provide real value.
IVT distorts key metrics and reporting. Advertisers may see a low conversion rate on a campaign and cut the channel thinking that it isn’t working. This can result in advertisers scrapping a promising channel even though their results are actually being influenced by invalid traffic.
Identifying Invalid Traffic
Invalid traffic’s greatest strength is its ability to go undetected. If advertisers take an active role in monitoring their traffic and metrics, they can limit wasted spend and protect campaign integrity. Advertisers should run regular checks like the following:
- Comparing ChatGPT Ads to other channels: You should separate ChatGPT Ads from your other Search campaigns running the same offer and compare their performance. If there is a considerable difference in performance, it may be a sign of invalid traffic. In our analysis, ChatGPT Ads clicks came from datacentre and proxy infrastructure at 3.6 times the rate seen on Google Ads for the same advertisers.
- Check where your clicks are coming from: Traffic from datacentres, proxy networks and hosting providers should be thoroughly investigated as a large share could potentially be IVT. In TrafficGuard’s analysis, just four hosting and proxy networks reached 25 of the 47 advertisers and accounted for nearly half of all invalid clicks.
- Check for repetition: You should look out for IP ranges that pop up across different campaigns, or across multiple days as these can indicate automated traffic.
- Highlight impossible devices: Spoofed devices are a strong indicator of invalid traffic. These are devices that are reported with a combination of processors, operating systems and screen sizes that simply can’t exist.
- Match clicks and visits: You should compare clicks against real landing page sessions and engagement. If a click doesn’t turn into a meaningful visit, it should be investigated.
Advertisers can also consider independent IVT detection that analyses every click against hundreds of signals and applies the same standard across ChatGPT, Google and Meta. If IVT is caught early instead of in a report after the fact, advertisers can stop paying for clicks that will never convert.
Checking Quality to Ensure Growth
Advertising inside AI chatbots like ChatGPT has handed advertisers a genuinely new way to reach audiences. As with every new channel, it also attracts invalid traffic.
Invalid traffic has already found ChatGPT Ads, and on some campaigns it is taking a meaningful share of clicks that will never lead to conversions. Advertisers should be inspecting the quality of their traffic before they increase their spend. A problem is far cheaper to find while the test is small than after the budget has scaled.
______
About Mathew Ratty
Mathew Ratty, a seasoned professional with 7 years in digital ad tech, currently leads as CEO of Adveritas. Formerly part of a mobile ad network, he’s also an avid tech investor with a decade of diverse investments. Under his leadership, Adveritas launched its flagship product, TrafficGuard, using innovative strategies and assembling a top-tier C-level team. Holding a First Class Honours Finance degree from Curtin University, Australia, Ratty steers TrafficGuard’s mission. This pioneering ad fraud prevention solution employs AI and advanced machine learning, revolutionizing business operations. Trusted by major brands like Disney, Tab Corp, and HelloFresh, TrafficGuard, accessible on Google Cloud Marketplace, upholds transparency and security in digital advertising, setting industry benchmarks.
Join our LinkedIn group Information Security Community!
