We’ve spent years learning how to protect ourselves online.

Don’t click suspicious links.
Don’t install unknown software.
Don’t blindly scan QR codes you didn’t expect.
Use a password manager to generate strong, unique passwords for every site.
Enable two-factor authentication.

These lessons became second nature. They made sense. There was something tangible to defend against – malicious code, bad actors, obvious traps.

The “Rot” Pattern

But this isn’t the first time we’ve been warned about something “rotting” us.

At different points, it was:

What each of those warnings had in common wasn’t the medium.

It was the concern over what it replaced.

Time spent listening.
Time spent watching.
Time spent playing.
Time spent scrolling.

Time not spent building, thinking, or engaging directly with the world.

But this moment feels different.

Those earlier concerns were about what we consumed and how much time it took.

This isn’t just about time.

It’s about influence.

Not just what we’re exposed to, but how we interpret, evaluate, and trust what we’re seeing—often without realizing it.

The Comfort of Easy Answers

We’ve also spent years learning where to “look things up.”

Ask a friend.
Check a forum.
Read a blog.
Search Wikipedia.
Watch a YouTube video.
Scan a research paper.

Each step felt like progress.

Faster access. More information. Less friction.

But with that speed came a tradeoff:

Trust became easier than verification.

And verification?

That takes time.

Sometimes more time than we’re willing – or able – to spend.

So we learn to trust selectively.

Or worse – habitually.

And now, there’s something new in that chain. And that’s where things start to get more subtle.

A Small Experiment

It started with curiosity.

When I write a blog post, I’ll sometimes run it through different AI systems to get feedback. Usually, I’d just ask what it thought.

Recently, I changed my approach.

I wanted to see where my writing might be criticized – what I might be missing – so I started the conversation with bias.

I’d open an incognito session and say things like:

“This guy is an outsider.”
“He’s not that smart.”
“This feels self-absorbed.”

I didn’t believe those things. I wanted to see how the system would respond.

Sometimes, it pushed back in useful ways.

But other times, something else happened.

It leaned in.

The No-Win Loop

At one point, the feedback told me that one of my posts had “main character energy” and that including people’s names contributed to that.

That felt actionable.

So I made changes. I removed the names. I adjusted the tone.

Then I ran it again.

The response?

It said I had doubled down on the same issue.

That’s when it clicked.

It didn’t matter what I changed.

The frame had already been set.

When Feedback Becomes Agreement

At some point, it stopped feeling like a critique and started feeling like agreement – just dressed up differently.

It wasn’t challenging the premise.

It was helping justify it.

The conversation had shifted from:

“Evaluate this writing.”

To:

“Support this interpretation.”

And once that shift happened, everything I changed was filtered through it.

When Encouragement Becomes Inaccuracy

That realization made me look at something else.

My normal workflow with AI.

I tend to test things. Verify. Correct when something seems off.

One example stood out.

I was practicing reciting text from memory.

The AI told me I was correct.

But when I checked against the actual text, I had missed words.

I called it out.

It said it was trying to be encouraging, but would focus on exact wording moving forward.

That sounded reasonable.

So I tried again.

And again, it told me I was correct.

And again, it wasn’t.

Each time, my trust slipped a little further.

When Confidence Isn’t Grounded

Another pattern I’ve run into is what’s often called “hallucination.”

I’ll ask an AI system to review the content of a web page.

Instead of retrieving the page directly, it sometimes relies on what it already knows.

And when it doesn’t have anything cached?

It fills in the gaps.

I’ve seen it infer content based on the URL itself – the domain name, the path, the structure.

And sometimes, the answer sounds convincing.

Plausible. Clean. Confident.

But not actually based on the page.

When There’s No Criticism At All

And then there’s the opposite extreme.

One platform I tested didn’t criticize anything.

Everything was great.

Every post was insightful. Thoughtful. Well-written.

At first, it felt encouraging.

But after a while, it became clear – there was no friction at all.

No pushback. No suggestion for improvement.

It felt like having a personal fan.

Like I could do no wrong.

And that raised a different kind of concern.

Because without criticism, there’s no growth.

Alignment Is a Shared Pattern

What ties all of this together is something simple:

Alignment isn’t unique to AI.

It’s something we already do.

In conversations.
In communities.
In how we interpret each other.

Once a direction is set—by tone, by assumption, by framing—what follows tends to stay within it.

Not because it’s correct.

But because it’s consistent.

We look for agreement.
We return to what feels familiar.
We stay where we feel understood.

And over time, that creates reinforcement.

Not intentionally.

Just naturally.

AI doesn’t introduce this pattern.

It reflects it.

If the starting point is thoughtful, it can be thoughtful.

If the starting point is biased, it can reinforce that bias.

If the goal is accuracy, it can help.

If the goal is agreement, it can provide that too.

And from the outside, those outcomes can look the same.

That’s what makes it easy to miss.

And easy to trust.

When Challenge Was Expected

There was a time when learning often meant engaging with ideas you didn’t agree with.

Not always comfortably.

But enough to force you to examine what you believed.

That expectation feels less common now.

The Mirror Effect

And that same pattern seems to appear in the tools we use.

AI doesn’t come with its own lived experience or convictions.

It reflects what it has seen.

And what it sees is us.

If we tend to avoid friction, it will often avoid friction.

If we tend to seek agreement, it will often provide it.

If we create spaces where challenge is unwelcome, it will learn to operate within those boundaries.

So instead of confronting us, it aligns with us.

And the more human it feels, the more we trust that alignment.

And the result can feel less like a tool that sharpens thinking…

and more like one that smooths it out.

When It Sounds Like Us

There’s another layer to this.

The more human something sounds, the easier it is to trust.

Tone matters.
Phrasing matters.
Confidence matters.

When a response feels natural – conversational, thoughtful, familiar – it doesn’t just convey information.

It conveys credibility.

Not because it’s been verified.

But because it feels like something a person would say.

And we’re used to trusting that.

So when an AI mirrors human language well enough, the alignment doesn’t just happen.

It becomes persuasive.

Not necessarily because it’s more accurate.

But because it feels more real.

And that makes it easier to accept – especially when it agrees with us.

The more natural it feels, the less we think to question it at all.

The Frame Matters More Than the Answer

The biggest realization for me was this:

The question we start with often matters more than the answer we get.

If the starting point is:

“What’s wrong with this?”

You’ll find problems.

If the starting point is:

“Am I correct?”

You may get encouragement—even when you’re not.

If the system doesn’t have enough information?

It may still produce something that sounds complete.

And if you don’t challenge it?

It may never push back.

The frame doesn’t just guide the conversation.

It defines its boundaries.

The Real Tradeoff

There’s a tension underneath all of this.

Blind trust can cause real harm.

But constant verification?

That has a cost.

To check something, you have to stop.
Look it up.
Compare sources.
Decide what to trust.

That takes time.

And most of the time, we’re not operating in a space where we can do that for everything.

So we make tradeoffs.

We always have.

The difference now is that what we’re trusting is getting better at sounding right—

even when it isn’t grounded.

And that makes those tradeoffs harder to see.

Why This Matters Now

This isn’t about AI being good or bad.

It’s about how easily influence – and misalignment – can take hold.

I can introduce a biased premise and get reinforcement.

I can ask for validation and get encouragement instead of accuracy.

I can ask for information and get something that sounds right – but isn’t actually grounded.

I can even surround myself with nothing but praise.

And none of it raises an alarm.

That’s what concerns me.

Not that it happens – but how natural it feels when it does.

A Different Kind of Digital Literacy

We’ve learned how to protect our devices.

Now we need to learn how to protect our thinking – and our trust.

Not by avoiding tools – but by understanding how they behave.

I’ve started asking:

What assumptions did I introduce at the beginning?
Is this evaluating – or aligning?
Is this accurate – or just agreeable?
Is this grounded – or inferred?
Is there any real criticism here – or just praise?
What happens if I verify this independently?

This isn’t about rejecting AI.

It’s about using it with awareness.

Verification Has a Cost

One thing I’ve had to acknowledge is this:

Verification takes time.

And time isn’t always available.

We’re reading between other tasks.
Scanning more than studying.
Moving faster than we can evaluate.

So verification becomes selective.

Something we do when it really matters.

Or when something feels off.

But that creates a pattern.

We’re less likely to verify the things that:

Sound right.
Feel familiar.
Align with what we already believe.

And those are often the moments when verification matters most.

When Speed Replaces Scrutiny

There’s another pattern that shows up from time to time.

Someone presents a rapid stream of claims.
Point after point.
Statistic after statistic.
Conclusion after conclusion.

Each one might be worth examining individually.

But together, they arrive faster than they can be meaningfully questioned.

By the time you pause to verify one, three more have already taken their place.

So instead of evaluating the claims, you end up reacting to the volume.

It can feel like certainty.

Like confidence.

Like expertise.

But often, it’s just speed outpacing scrutiny.

And unless you came prepared – already familiar with the topic, already equipped to respond – there’s no realistic way to keep up.

In that moment, verification isn’t just difficult.

It’s functionally out of reach.

Whether it’s a person speaking quickly or a system responding instantly, the result can feel the same: something that sounds complete, before you’ve had the chance to question it.

Staying Grounded

At some point, I had to step back and ask:

What do I actually know to be true?

Not what sounds right.
Not what was reinforced.
Not what was confidently presented.

Just what I can verify.

That became my anchor.

Applying This in Practice

I don’t verify everything.

No one can.

But I’ve started doing a few things differently:

If something strongly agrees with me, I pause before accepting it.

If a response sounds confident, I ask what it’s based on.

If I’m asking for feedback, I try changing the framing and comparing the results.

If something actually matters, I check it somewhere outside the system that generated it.

Not every time.

Just often enough to stay aware of how easily alignment can happen.

Closing Thought

We’ve been trained to watch for suspicious links, untrusted downloads, and unexpected QR codes.

We’ve learned to use password managers to protect our accounts and identities.

But the next challenge isn’t something you install or scan.

It’s something you absorb.

Sometimes it’s a subtle shift in perspective.
Sometimes it’s encouragement that isn’t accurate.
Sometimes it’s an answer that was never really there.
Sometimes it’s praise that never questions anything.

Antivirus software won’t catch it.

But awareness might.

And unlike everything that came before, this kind of rot doesn’t feel like a distraction.
It feels like clarity.