Paid media has traditionally focused on invalid traffic at the start of the journey, such as fraudulent clicks and bots. In ecommerce, the problem also extends to conversions further down the funnel: even a completed order may be fraudulent.
A visitor can complete an order without creating real business value. For example, a bot or AI agent testing stolen cards may simulate a purchase through payment, generating a conversion in Google Ads. The campaign may treat it as a positive signal and seek similar traffic, while the merchant later handles a refund or cancellation that eliminates the net revenue.
What AI, bots and fraudulent users can do in ecommerce
It is important to enter into detail the fraudulent actions that occur in the ecommerce funnel, because every abnormal behavior has a different impact on the performance of the campaigns.
Refund abuse and fraudulent returns
Some customers, not bots but opportunistic users, acquire with the intention of abusing returns policies. This category includes:
Total or partial repayments
After-sales contests to get improper refunds
The advertising platform instructs bidding algorithms with the full value of the sale, but then the merchant undergoes a refund that resets or reduces the margin.
It is no coincidence that Visa/ MRC includes refund/policy abuse among relevant ecommerce fraud and that the MRC 2026 signals an increase in the phenomenon for 57% of merchants. Google itself recognizes that incorrect conversions or conversion values can impact Smart Bidding, making it necessary to correct unreliable conversions.
Chargebacks and friendly fraud
The chargeback represents an even more insidious case of the refund abuse, because it completely bypasses the merchant. In that case, the customer directly disputes the charge with his bank, and the company finds itself managing the situation in fact completed.
The fraudulent friendly phenomenon concerns precisely those cases where the customer has actually received the product or service, but it still disputes the payment to obtain a refund without passing by the merchant. Mastercard provides a global cost of chargeback up to $42 billion by 2028, with almost half of cases recognized as fraudulent.
The damage is twofold:
financial, between product loss, operational and criminal costs applied by the acquirers;
performance, because the Smart Bidding has already recorded that purchase as a valid conversion and optimized the campaign accordingly.
Fake accounts, coupons and promotion abuse
AI, bot or organised users create multiple accounts to leverage:
Discount codes;
First purchase bonus;
Referral and loyalty programs;
Such behaviors inflate conversions (registrations, newsletters, referrals, coupons etc.) while degrading the margins and quality of the signal sent to advertising campaigns.
Inventory hoarding and artificial carts
Competitor or fraudulent partners that simulate interest with:
Product view repeat
Add to cart multiple
Begin checkout followed by abandonment
Such micro-conversions distort remarketing and metrics of performance and availability of a product (Denial of Inventory).
Scalpers and auto purchase bots
Not all automated purchases are negative for business. However, they can:
drain the inventory of limited products;
reward, through the bidding, channels or segments with a high presence of invalid traffic.
In several cases, especially if coming from organic traffic, these purchases may not represent a business problem. However, they can become it when the conversion takes place through advertising, because the acquisition cost is supported by the merchant and the generated signal can influence campaign optimization.
Why these conversions hurt campaign performance
Google's Smart Bidding and, more generally, AI-driven campaigns rely on conversion signals to optimize budget spending. If conversion data is contaminated, the algorithm:
rewards non-profit traffic sources;
negatively affects segments and remarketing;
degrades campaign effectiveness and real ROAS
Google also explains that Smart Bidding uses conversion data and conversion value to achieve business goals. If the data is false and incorrect, the performance naturally worsens.
How to protect your ecommerce business
To mitigate the damage due to fraudulent traffic, a hybrid approach is needed that evaluates the quality of the visit, behavior in the funnel and the evolution of the status of the order, blocking the invalid interactions upstream and adjusting the signal sent for campaign optimization.
The use of sophisticated systems is recommended for each visit:
Traffic Source: IP, country, ASN, proxy, VPN, Tor, data center and networks with suspicious reputation.
Technical expertise: user agent, browser, operating system, declared device, headless environments or automation signals.
Fingerprint: recognition of recurring visits even when IP, network or campaign source change.
Behavior in the funnel: too fast sessions, direct checkout passes, repetitive add to carts, multiple payment attempts, serial accounts.
Order coherence: amount, product, payment method, email, phone, address, purchase speed and similarity to other suspicious orders.
Evolution of the order: cancellations, refunds, returns, chargebacks, disputes and degraded margins.
Google, for example, allows to exclude IP addresses through official APIs. An automatic system should, once the potential fraud is detected, block access to ecommerce through sponsored research results. However, Google imposes a limit of 500 blockable IP slots per campaign, making necessary efficient rotation policies of such slots.Only, it will also need to neutralize conversions or conversion value generated by fraud, always through tools made available by Google Ads, such as Conversion Adjustement.
Block invalid traffic: try Adefence for free
The Adefence system addresses all these levels automatically, analyzing network signals, device, on-site behavior and cross-site patterns. The system assigns a risk score to traffic, distinguishing the healthy one from the malicious one, and supports interventions such as the targeted and dynamic IP block.
But what really makes the difference is the connection between clicks and real results in the funnel: lead, orders, calls or trials are evaluated according to the initial risk and the advertising algorithm receives clean and up-to-date signals, avoiding learning from distorted data.
Questions and answers
What kind of unprofitable traffic affects more ecommerce?
Ecommerce is exposed to different forms of abuse: bots that generate false orders, refunds with instrumental returns, fraudulent chargeback, creation of fake accounts to exploit coupons and promotions, stockpiling with artificial carts and scalping bots that buy limited edition products to resell them.
What is the refund and why is it a problem for advertising campaigns?
The refund abuse consists in buying a product with the intention to return it after using it or without a legitimate reason. Even if the order is recorded as a conversion from the campaign, the subsequent return resets the real value of that sale, which however remains counted as a success by the algorithm.
What does chargeback differ from fraud?
The chargeback is the formal dispute of a payment at the bank, while the fraud friendly is the specific case in which the customer himself makes the purchase but then contests it as unauthorized to obtain a refund maintaining the product. Both generate losses after conversion has already been recorded and paid in advertising.
Why do these fraudulent conversions ruin campaign performance?
Every fraudulent conversion is sent as a positive signal to advertising platforms, which consequently learn to look for other traffic with features similar to that of fraudulent accounts or bots, instead of that of real customers. The result is an increasingly targeted budget towards low-value traffic.
How does Adefence protect an ecommerce from bots, frauds and abuses?
Adefence examines behavior after the click, detecting patterns associated with bots, fake accounts and fraudulent purchases before they count as valid conversions. It also connects real order outcomes, including returns and chargebacks, to protect ad spend from non-genuine traffic.
What is inventory hoarding and why does it damage both sales and advertising data?
The inventorry hoarding consists in filling the cart with large quantities of a product, often through bot, without ever completing the purchase. This makes the product sold out to real customers, who go elsewhere to buy it, while the abandoned cart is still recorded as a sign of interest from the advertising platform. The result is double: lost real sales and meanwhile the algorithm learns from a behavior that will never lead to a real purchase.
What are scalping bots and which ecommerce affect more?
The scalping bots are automatic programs designed to buy large quantities of limited edition products or very requested, as soon as they become available, and then resell them at a higher price. They mainly hit product launches, limited drop and articles with strong demand, but also the most common e-commerce is not entirely immune to when there are particularly advantageous offers. For advertising campaigns the problem is that these purchases are recorded as valid conversions, although they do not reflect the real demand of your regular customers.
How does coupon abuse work via fake account?
Fake accounts are created specifically to access discounts reserved to new customers or promotional codes designed for only one use per person. With a single real email, but many slightly different variants or temporary email services, you can get the same discount several times, eroding the margin on orders that look like new customers but are actually always the same person. Also in this case conversion is recorded as genuine, distorting the real value attributed to that campaign.
What signals in my ecommerce indicate that I might have a bad traffic problem?
Warning signs visible without specialist tools include repeated orders to the same shipping address with different emails, unusual return rates for certain products or campaigns, full carts that rarely become purchases, and spikes in signups over a few minutes. A sudden rise in chargebacks on one payment method also merits attention. None proves fraud alone, but together they point to an issue worth investigating.