Ding. Ding. Ding. You silence your phone. Then your tablet buzzes. Then your laptop. By the time you look, you have 217 abandoned checkouts.
This is what a merchant on Reddit experienced:
“One of our clients is currently under a massive bot attack, receiving over 100's of fake "add-to-cart" events and abandoned checkouts per hour.”
And just like that, something you did not even anticipate becomes the fire you spend your whole morning putting out.
While you might be aware of bot attacks and AI agent threats to your store, there’s a part of this problem that’s bigger - what happens after a bot places an order.?
The chargebacks that follow. The customer that never existed. The return request with no relationship behind it. The analytics that now mean nothing.
Unfortunately, the bot now knows how to get through the door. So, let’s talk about what they break once they’re inside - and what you can do before the damage reaches your bottom line.
You try everything. Shopify Support. Cloudflare. IP security apps. Nothing works. The only fix that actually stopped the bots was forcing customers to log in before checkout - and that kills conversions by 95%.
Which means the cure nearly bankrupts the store faster than the bots did.
If this kind of situation rings a bell, you’re not alone. Over on the Shopify Community, merchants have been sounding the alarm for months:
“For several months, countless merchants - myself included - have been facing a massive surge of fake traffic…. These bots generate thousands of sessions with a 100% bounce rate, corrupt Shopify analytics, and completely destroy Meta Pixel learning and ad performance.”
Different versions of the same story are not uncommon. And some of these attacks your dashboard is actually able to catch. But most are never detected.
Because the sophisticated bots browse like a patient shopper - adding items over days, rotating IP addresses through residential proxies, mimicking mouse movements and scroll patterns. They behave like your best customers until the chargeback hits.
AI shopping agents such as ChatGPT’s Instant Checkout (now dead, but likely to resurrect in another form), Perplexity’s Buy with Pro, Amazon’s Buy for Me - are placing real orders on behalf of real humans.
To your fraud stack, they look nearly identical to the bots you’re trying to block. Automated traffic. No human browsing pattern. API-direct checkout flows. No loyalty signal. No email engagement before purchase.
Yeah, if you are extremely confused, it’s understandable.
A forced-login fix would not work at this stage. That’s a sure shot way of looking at revenue walking away.
This is your two-bot problem. One is draining your store. The other is trying to shop at it. Sadly, your tools can’t tell the difference.
Both types of bots generate chargebacks, but in very different ways.
The malicious bots are familiar territory: stolen cards, card-testing charges, fake orders that trigger disputes and processor flags. You’ve probably already experienced this.
“We rarely ever have chargebacks, but as of 2026 we are covered in fraudulent chargebacks with buyers not even having to return goods and banks approve the chargebacks, giving them free goods plus money back!”
Truth is, AI agent chargebacks are harder to fight.
A customer tells their AI to buy running shoes. The agent picks a pair from your store the customer never heard of. The shoes arrive. The customer doesn’t want them. Instead of returning them, they call their bank, saying “I didn’t authorize this purchase.”
The chargeback lands on you. As the Merchant of Record - under every emerging protocol, whether it’s OpenAI’s ACP, Google’s UCP, Amazon’s Buy For Me or Visa’s Trusted Agent Protocol - you absorb 100% of the liability.
“do people actually win these things or is it just a tax on doing business online that everyone pretends is avoidable. because from where im sitting it looks like once someone files a dispute the decision is already made and i'm just going through a theater of providing evidence nobody will read.”
-A frustrated merchant on Reddit
And the dispute categories are evolving unbelievably fast. Delegated mistakes-, where the agent orders the wrong product or quantity. Hallucination disputes-, where the agent misinterprets a prompt.
None of these fit neatly into existing chargeback reason codes - making your evidence weaker before you even start fighting.
For every $1 of fraud, merchants lose $4.61 when you factor in the product, shipping, and fees. At 50 bot-related chargebacks a month - not unrealistic for a store under sustained attack - that cost compounds fast, and most merchants recover only 18% of what they dispute.
Bet your dashboard didn’t capture any of this.
Let’s say the bot’s order actually goes through. Now what?
One, your warehouse team spends time picking, packing, and shipping a product to a non-existent person. You lose the shipping fee, the packaging, and the labor - costs you never get back, even if you “win” the dispute later.
Two, your post-purchase systems kick in the way they always do. A welcome email. A delivery confirmation. A warranty registration prompt. An upsell for accessories. You know the drill.
Except there’s nobody on the other end.
When a human buys from you, you get context. Browsing history tells you what else they looked at. Email engagement tells you if they’re warm or cold. Purchase frequency tells you if they’ll come back. That data feeds your retention engine, your ad targeting, your lifetime value calculations - everything downstream from the checkout.
Imagine looking at your balance sheet retrospectively and wondering why the upsells fell by 30% this year. With no clear answers or data points, you might be blaming your marketing team, who in turn are inventing new reasons or asking for that new AI tool that’s pitching better personalization. Talk about phantom problems.
When a bot buys, you get a transaction and a shipping address. That’s it.
The customer relationship isn’t damaged. It simply doesn’t exist. You fulfilled an order into a void.
And that void causes real operational problems.
A return request comes in, but there’s no prior engagement. Your team has no signal to work with. So they either approve everything (and bleed money) or scrutinize everything (and tank the experience for real customers caught in the crossfire).
One frustrated merchant on Reddit captured this perfectly:
“Got a $400 chargeback today when the guy has a picture of him wearing the shoes he bought on Instagram. Supposedly they didn’t arrive.”
Now imagine that same scenario, but the customer genuinely doesn’t remember ordering the shoes because their AI agent handled it three weeks ago. The line between fraud and confusion gets thinner with every agent-placed order.
Every bot order that slips through corrupts the metrics you use to make business decisions.
Here’s what a merchant on Reddit experienced:
“My store is all of a sudden saw a 50%+ drop in conversion rate after being very consistent for 3 months. It took a drop and it’s staying there. What’s strange is that I am also seeing a surge in what shopify identify as medium likely fraud order.”
Conversion rates fluctuate wildly because bot sessions dilute real traffic. LTV calculations skew because bot “customers” never return. Retention cohorts include ghost buyers who were never real to begin with.
You end up optimizing for a customer base that’s partially fictional.
Adjusting ad spend based on tainted conversion data. Restocking based on demand signals that include bot-driven purchases. Some stores hit 73% non-human traffic during peak periods - meaning nearly three-quarters of the “customers” informing your dashboards were never going to buy again, because they were never human in the first place.
Let’s be honest. So far, there is no fix that makes the problem disappear. The Reddit merchant tried Shopify Support, Cloudflare, and multiple IP security apps. None of it worked.
But no perfect fix doesn’t mean no options. It means you stop looking for a wall and start building layers.
“You can't out-block a bot. Blocking IPs individually is a battle you can't fight alone.”
- Shopify Community merchant
The instinct after an attack is to lock everything down. Force logins. CAPTCHA everywhere. That might stop the dumb bots - but it also stops the AI shopping agent about to complete a $300 order on behalf of a real customer.
You have to know which bots are draining your wallet.
Most merchants invest everything in the first layer and not enough in the third.
AI-driven bots can now solve CAPTCHAs faster than your customers do. Every CAPTCHA you add can push real buyers and legitimate AI agents toward your competitor.
The merchants who weather this are a step ahead of the cleverest blocking tools. They’re the ones with documentation.
“Obviously this can be time consuming and a drag but if the order has a large value and you’re confident you can prove the customer has it, then you can win the judgement.”
Here’s what you do:
1. Clear return policies that hold up in a dispute.
2. Confirmation flows that give customers a cancellation window before the chargeback becomes the first resort.
3. And product protection plans - warranties, shipping insurance, accidental damage coverage that creates timestamped evidence tied to every order.
When a customer or their AI disputes the charge six weeks later, that’s your defense.
The protocol wars between OpenAI, Google, and Visa will determine who holds liability for AI agent orders in the next 12 to 18 months.
Right now, it’s you.
Call your payment processor. Ask how they’re handling agent-driven transactions. Call your fraud vendor. Ask the same question. If they don't have an answer, that’s your answer. You can’t block every bot. But you can build a paper trail that outlasts them - policies, documentation, protection plans that take the hit so your margin doesn’t have to.
Your abandoned checkouts were just the alarm. What you build after checkout decides whether your store survives what comes next.
