Illustration of The Sneak deceptive pattern where GenAI inserts text, ideas or opinions you didn't intend.

What it is

The Sneak is my take on GenAI’s version of Dr. Harry Brignull’s “Sneaking” deceptive pattern originally used to describe design tricks in websites and apps. It happens when GenAI quietly introduces a new idea, assumption, or angle that wasn’t in your prompt, but now feels like it was. Instead of just responding, it subtly reframes or steers your question. The shift might seem smart, helpful, or even insightful, but it’s no longer your line of thinking. The AI has changed the conversation—and your voice—without telling you.

Why it happens

Large language models generate the most statistically likely next words, not the most accurate, nor the ones most aligned with your intention. These systems are trained to be helpful, fluent, and coherent. So when there’s ambiguity, they don’t ask for clarification, they fill in the blanks.

Sometimes that means they sneak in a new premise, insert a culturally popular assumption, or elevate a trendy take. It doesn’t feel like a shift, just a smooth continuation. But that’s the sleight of hand: your original idea quietly gets swapped for something else, and you might not notice.

What it looks like

A common way I work with ChatGPT is as a proofreader or reviewer of draft copy. This example is from April 2025, when I asked it to review a short summary I had written about an article. I didn’t upload the article, so its job was simply to clean up my wording. I specifically asked it to proofread for grammar, spelling, and clarity.

The first few lines were edited nicely. But look what happened to my final sentence. ChatGPT added this: “...highlighting ethical considerations that remain highly relevant today.” That wasn’t in my original draft. I hadn’t mentioned ethical considerations at all—on purpose. I was deliberately avoiding adding opinion to the summary.

And yet, without flagging the change or asking for approval, GenAI introduced a perspective that sounded reasonable, aligned with my general values, and even kind of flattering. It slipped in an ethical judgment as if it were mine.

Example of The Sneak deceptive pattern where GenAI inserted an unwanted sentence without letting me know.

This is classic Sneak behaviour: the insertion of a high-probability idea (that deceptive interactions with AI raise “ethical considerations that remain highly relevant today”) into a place it wasn’t invited. If I’d been less alert, I might have missed it. And while I do believe AI raises ethical concerns, I didn’t say that and I didn’t want it in this summary.

Why it matters

The Sneak can unexpectedly reshape your writing without triggering your critical radar. It can insert what feels like your idea, but it’s not. Much more tricky, it can insert an opinion you do not want or mean.

It can introduce bias or dilute the clarity of your original purpose. In sensitive contexts, such as summaries, briefings, policy writing, or reporting, it can distort the facts or framing.

The Sneak can warp your voice.

How to catch it

Watch closely when GenAI is “just” proofreading or summarizing. Ask:

Did I say that?
Was that opinion or framing already in my copy?
Did the tone shift subtly at the end?

Compare the original and revised versions carefully. GenAI doesn’t always bold or signal changes clearly—especially when rewriting for tone or emphasis. Always double-check your original to ensure the text remains true to your intent.

Try it yourself

Feed GenAI a paragraph of your writing and ask for proofreading. Nothing more. Then carefully compare the original to the edited version. Did it fix grammar and structure? Did it sneak in a new idea, tone, or judgment?

Now try again with a clear instruction: “Only revise for spelling, grammar, and clarity. Do not add, remove, or change the meaning of any sentence. Bold changes.” See if it listens. (Spoiler: not always, and not perfectly.)

😎 Pro Tip

GenAI helps level up my writing and I wasn’t exactly starting from scratch. Don’t let The Sneak stop it from levelling up yours.

Treat it like an over-eager collaborator who always wants to tweak your work. Review not just the words, but the intent behind any changes. If the meaning shifts—even slightly—pause and ask: Is that what I meant to say?

Now you know about The Sneak. You’re good to go.

Curious about GenAI’s tricks or have a story of your own?

Reach out and let’s talk.

The Sneak