The Intelligent Interface War
Gray Patterns to Fight the Machines Are Ruining CX
Gray Patterns: Brands are changing their designs for a world full of AI agents, scrapers, and smart interfaces.
We’re the collateral damage.
If you bought something on Amazon recently, you might have noticed a boring change in your inbox.
You got the receipt, and the subject line looked the same.
But when you open it, all the details are missing. No item names, no specs, no breakdown of your purchase. Instead, just a generic confirmation with a button that needs you to return to Amazon’s website to see your details.
Why?
Because Amazon have started asking, “How do we stop Google Gemini, Anthropic, and a swarm of scrapers from leveraging our customers’ inboxes into competitive research against us and free training data for them?”
You weren’t the target.
You were the collateral damage.
This small change shows where we’re heading.
A bigger trend.
Brands making open standards and consumer tools harder to use.
Defensive brand behaviour in a fight against the machines.
From dark patterns to gray patterns
I’ve talked before about dark patterns: the deceptive tricks that manipulate us with pre-ticked boxes, “confirmshaming,” and making it nearly impossible to cancel a service.
Dark patterns are designed to be harmful. They want to pressure, mislead, or wear us down so we do something we might not have chosen on our own.
But Amazon are playing a different game.
These are Gray Patterns1. Brands changing their designs for a world full of AI agents, scrapers, and smart interfaces.
Here’s the breakdown:
Primary target
Machine agent (AI, scraper, automated bot)
Intent
Defensive
Result
Collateral harm to the customer while defending data or control
Design question
“How do we stop this machine from reading us even if people suffer?”
A Gray pattern is a design choice to block machines from reading information.
But it makes things worse for people too.
Amazon aren’t trying to fool you. They’re trying to protect themselves, to keep AI agents and data brokers from getting structured, machine-readable purchase data.
Our loss of convenience isn’t the goal, but it’s still real.
And that lack of concern matters.
You can often pressure a brand to stop using obvious dark patterns because it affects their sales and reputation. But can you really shame a trillion-dollar company into giving up what it sees as a crucial defence against AI taking over its data?
Well, good luck with that.
The Age of Entanglement: when the same interface serves humans and machines
Let’s look at the bigger picture.
Human identity, big platforms, and AI all mix together in complicated and sometimes difficult ways.
For years, algorithm anxiety was our concern.
But platforms are worried about new outside algorithms.
Amazon isn’t just designing for you anymore.
It’s designing against:
Inbox‑scraping data brokers who sell purchase data and panels to consultancies.
Personal AI agents that will sit inside Gmail or Outlook and quietly build a complete picture of what you buy, where, how often, and at what price.
Intelligent interfaces (think Gemini, custom agents, system‑level copilots) that could become the primary interface between you and the brands you buy from.
For Amazon, it’s a very real threat. If your AI agent can see everything you buy and check any marketplace, Amazon could go from being your main shopping site to just one of many options in the background.
So platforms start attacking our shared interfaces - the places where humans and machines both meet the brand:
Your inbox.
The open web.
Public APIs.
Structured metadata.
And so targeting the very places your future smart interfaces need to work.
Machine defence in the wild: platforms are already fighting AI
Amazon isn’t alone.
We’re seeing new ways to defend against machines, with moves to block AI agents and scrapers.
A few examples:
Social platforms walling off the graph
Reddit and X/Twitter saw that their conversations were being quietly scraped to train AI and power search tools. As a result, people now face more login walls, more hassle, and less easy browsing.Publishers and dynamic paywalls
Many news organisations use advanced tools to detect bots and block AI. If your request seems like it’s from an AI scraper, you get a CAPTCHA, a shortened article, or a strict paywall - even if you pay for a subscription. Once again, the human experience suffers as a side effect of fighting machines.Data poisoning - Artists and platforms use tools to add invisible noise to images, so AI models get confused when they try to learn from them. You can’t see any difference, but the AI gets bad data.
E‑commerce price and markup obfuscation
Retailers are mixing up website code, changing CSS class names, or showing prices in ways that are hard for bots to read. You still see “£19.99” clearly, but price-comparison bots see nonsense or nothing at all. The reason? To stop undercutting, price games, and revealing pricing strategies.
All these technical choices are out of fear of being cut out, but we end up with a poorer experience.
What we lost with Amazon’s receipts.
You may see Amazon’s receipt change as just a small annoyance. But look closely, and the trend is a powerful loss of consumer control.
Before this change, an Amazon email receipt was:
Portable
You could do what you want with it, no permission required.User‑owned
It lived in your inbox. Amazon couldn’t edit or revoke it after the fact.Ctrl+F‑able
It was plain text. You could search “blue socks” or “router” and pull up the exact receipt in seconds.Proof of purchase
It itemised what you bought, when, and for how much, proof for returns, taxes, and reimbursements.Living in your domain, not the brand’s
It existed in your personal environment (Gmail, Outlook, whatever), not locked behind Amazon’s login flows.
Amazon’s change quietly revoked that artefact. The data still exists, but now locked back inside Amazon’s walled garden. A piece of your commercial history moved from sovereign, portable, user‑owned to captive, platform‑controlled.
That’s not just a tweak, more a reassertion of territory.
The false choice: “protect data or serve customers”
Underneath this trend sits a dangerous assumption: that brands must choose between defending their data from AI and maintaining a great customer experience.
It shows up as stripped‑down receipts, broken open‑web access, scrambled markup, and endless CAPTCHAs - a stack of Gray Patterns that quietly tells the customer: “If we have to degrade your experience to block the machines, so be it.”
It’s a trap, not because platforms shouldn’t defend themselves, but because treating machine legibility and human experience as a zero‑sum game is the wrong frame.
The more interesting move is architectural, not theatrical.
Instead of using the same brittle interface to fight both humans and agents, brands can technically separate the concerns with clean, controlled, machine‑only endpoints for AI; rich, honest, portable experiences for people. A shift from surveillance (“we trust you because we track you”) to scoped permissioning (“we trust this specific action”), and replace hostile friction with a few intentional “proof moments” where the human steps back into the loop.
Brands don’t need to choose between defending their platform and serving their customers.
Brands can win without walls.
They can work well with machines, without forgetting us who buy from them.
Acknowledgement to Future Commerce, who first coined the phrase Gray Patterns.
I’m also opening up spots for conferences, in-house presentations, and in-depth one-on-one calls, starting in September. DM me.
Thanks for reading UNCX | The Age of Entanglement.
Please share, or subscribe for free to get all new posts and to support my work.