Amazon Demands Perplexity Stop AI Agent from Making Purchases
bloomberg.com
bloomberg.com
I used to make Amazon purchases through ChatGPT Agent before Amazon blocked them. I could take a picture of a wrapper and say "buy a new one on Amazon" and it would handle the whole process. Awesome. I actually started shopping at other retailers when Amazon blocked their agent.
One theory I have is that an AI agent can more efficiently price-compare the dozens of different listings of the same item and order the cheapest, cutting into Amazon's margins.
Doesn't Alexa let you buy stuff? It's probably more to do with the walled garden scenario: use our AI to shop on our platform for our products.
If people find a benefit from using AI to shop on Amazon at scale then there's really no reason to keep Amazon in the loop at all. Especially for products that are high margin.
It's the logistics that matter.
If an agent can go and buy exactly and only the thing I need that is going to crush those other sales.
Because it bypasses their efforts to promote preferred products in the short term and, more importantly, in the long term someone else becoming the front door to purchasing also enables that actor to commodify Amazon the same way Amazon commodifies other sellers, and before long to cut out Amazon as a middleman entirely and become the marketplace where buyers purchase from a diverse array of functionally anonymous, interchangeable sellers.
More on topic, my personal suspicion would be that AIs get a lot of stuff wrong, unsurprisingly, and so these orders end up with a higher percentages of returns or people claiming they didn’t buy them or other customer service issues.
For the same reasons Amazon intentionally removed their search box's ability to require or exclude terms and target phrases. More efficient, faster searches for YOU reduces their opportunities to shovel more "not what you're looking for right now" in front of your eyeballs to increase impulse sales of adjacent goods.
1. When you visit amazon.com looking for a product, even if you are looking for a very specific product, they'll put other products in front of you. You might end up buying one of these other products, and that might make someone more money than if you bought the product you initially planned to buy.
2. You might buy more than you initially intended if you visit the site yourself, via the "Customers Also Bought" links, and other advertising.
3. Agents aren't perfect, and it's possible that they were making ordering mistakes at (much?) higher rates than humans were doing, which was costing Amazon more processing returns or other customer service inquiries.
4. Amazon wants to control the experience of buying a product through them, from start to finish. If they can't do that, they just become a commodity product aggregation site. An agent making purchases for someone can find the same product from another retailer, perhaps with better pricing, faster shipping, a better return policy, etc.
Of course, if other sites allow agents to make purchases, and that eventually becomes the way a significant number of customers want to shop, Amazon will have to get on board.
It means ChatGPT just gives me more of just what I'm asking for rather than giving Amazon the opportunity to push something else on me.
If AI enables regular people to sleeplessly and ruthlessly exploit the market, like large companies do, it would be a really really good thing.
Then, when the agent buys stuff, my guess is that returns are a higher %age (because naturally agents have no preference model so end users might not like what they purchased eventually and returned). I don't know if enough people returned for it to matter, but as a user, if an agent is making a purchase and i know its super easy to return, i won't check everything agent did, just buy and then return if I dont like it.
I don't think it's always a price comparison or margin thing at this point. Given their margins and volume, the agent purchases woudl need to be an order of magnitude more to matter on that front.
Can I ask my partner to buy a product on Amazon?
Can I ask my personal assistant to buy a product on Amazon?
Can I hire a contractor to buy products on Amazon?
Can I communicate with a contractor via API to direct them what products to buy?
What if there is no human on the other end and its an LLM?
Same issue with LinkedIn. I know execs who have assistants running their socials. Is this legal?
Like, where do we draw the line? In the future, would the only way to shop on Amazon be with approved VR goggles that scan your retina to verify you are a human?
And they even provide a definition of what an Agent is:
"Agent” means any software or service that takes autonomous or semi-autonomous action on behalf of, or at the instruction of, any person or entity.
Though to me it raises even more questions. What is a software that takes "autonomous" action on my behalf. Is curl "autonomous"?
Only when you supply -L
What if I set up a cron job to buy a certain product every month - is that not autonomous? What if it is first querying my live toilet paper sticks to make the decision?
What's the criteria that makes one function "autonomous" and the other one "manual"? I feel it really boils down to this.
That is why you have personal credentials to log in to Amazon. If you want to have delegating capabilities you can open an Amazon business account.
Based on this article, I'd think not?
Even the SEC would be against it, as it would inflate the user base of Amazon.
My wife and I used to share 1 account, but then I wanted to buy her a gift that had to be a surprise - so had to create a new account and add it as part of the family to the original one…. Then kids grew up and wanted to make small orders themselves, and I didn’t want them to see our order history…
1Password shared vaults are there for a reason - people share credentials all the time, business or personal.
This way Amazon can keep track of your separate buyer profiles.
Perplexify has shown itself to be a bad actor [1][2][3], and possibly incompetent, too [4].
We need to draw a line, eventually. But it’s far from urgent. And I don’t think Perplexity should be the one deciding.
[1] https://blog.cloudflare.com/perplexity-is-using-stealth-unde...
[2] https://www.reuters.com/legal/litigation/perplexity-ai-loses...
[3] https://arstechnica.com/tech-policy/2025/10/reddit-sues-to-b...
I'm not sure why you would feel entitled to make a purchase on their site outside of their (whatever reasonable or unreasonable) rules may be.
I am not against people using agents to do shopping although I won’t do that until I see a reliable agent. From the title I thought Amazon might have a point to protect users. But oh no. I should have known better that Amazon must have been cooking their own shopping agents. It’s already hilarious at this point.
Anyway, this excerpt from Perplexity has so many issues. It just looks like an ugly fight.
The official Amazon search is so bad and deliberately pushes "promoted" or overly expensive products to the top that I would say it's only a matter of time until someone reverses the enshittification by building a 3rd party Amazon client. And it looks like that's precisely what Perplexity did here: They offer an agent that works for you, not for Amazon. And, predictably, Amazon hates it. It reduces their power to strong-arm brands into purchasing advertisements for their own product names.
Well, that is if you use a real, functional search engine
Thanks
It’s actually the idea behind the decentralized marketplace I’m building. It uses MCP-UI to bring the whole storefront and checkout into the chat.
I’m keeping a close eye on e-commerce and AI and the recent deal Paypal made with OpenAI and Amazon getting aggressive, it’s clear they want to make AI powered commerce a walled garden.
Perplexity sitting on AWS and doing this is like the tenant (Perplexity) turned their apartment into an Airbnb and making tons of money with the landlord (Amazon) getting very upset.
Perplexity may have to think about moving off of AWS.
I'm not a lawyer, but the "security measure" wording sounded just slightly odd to me, and I wonder whether it's meaningful, and came from Amazon.
Sounds familiar?
Prudent or not, self serving or not, amazon is right to argue its access controls be strict and that it as owner of website should control it with industry wide accepted rules of engagement, this argument from perplexity is a bad one, human or not while browsing web is a low bar already, captcha etc. were easily circumvented even before ai agents, now if we argue agents are human adjacent, it is a direct case for removing humans of any agency (largely philosophically speaking) and not to mention, imagine the horror show of scams & ransomware it unleashes on millions of users (even engineers can’t recognize or stop them today, prompt injection etc)
This argument that websites controlling their access to bots & agents is a good idea. It should be the way it is, for businesses (amazon or not) and for internet blogs and open web associated sites, if they choose to exclude themselves from upcoming lovely silicon valley stories of ai utopias, they should be able to do so. No one should force them to ‘get with the program’.