The AI Addiction: What Are the Signs?

You ask ChatGPT to write something. It does. You read it. It looks decent. You ask, "Can you improve this?" It says sure, here are five adjustments. You apply them. It looks better. You ask again. It suggests three more. You apply those too. You ask one more time. It finds two more things to change.

This cycle never ends. And if you have not noticed it yet, you are probably in it.

I am talking about what I call the AI Addiction. Not a real addiction, obviously. Nobody is in rehab for asking Claude to review their homepage copy. But the pattern is there, and it is doing more damage than people realise!!

The Validation Trap

Here is the thing about AI feedback. It will always suggest something.

Always.

You could ask it to write the best paragraph it has ever produced, then ask it to review that same paragraph, and it will find something to improve. It is built that way. When you ask for feedback, the system assumes you want changes, so it gives you changes. Even when the original was fine.

This is what makes it addictive. Every response feels productive. Every suggestion feels like progress. You apply the changes, ask again, get more changes, apply those, ask again. It feels like you are improving. But you are actually just moving sideways in a loop that has no exit point.

As I wrote about in AI fatigue and the authenticity premium, we are already drowning in AI-generated content. The irony is that the people closest to the tools are the ones most at risk of losing their own voice in the process.

The Five Signs of Your AI Addiction

Behavioural psychology has a fairly standard list of addiction signs. Let us map them onto the AI feedback loop and see how many hit a bit too close to home.

1. Tolerance

In addiction terms, tolerance means you need more of the substance to get the same effect. With AI, it shows up as needing more rounds of feedback to feel satisfied with your work.

First it was one quick polish. Then it was three revisions. Now you cannot publish anything without running it through the feedback loop four or five times. The first draft never feels finished anymore, even when it is.

2. Loss of Control

This is the big one. You know you should just publish the thing. You know it is good enough. But you cannot stop yourself from asking for one more review.

The AI will always find something. A slightly punchier headline. A better transition. A more compelling call to action. So you keep going, because surely this next round will be the one where it says "actually, this is perfect, do not change a word."

That round never comes.

3. Preoccupation

You start thinking about your prompts when you are not at your desk. You lie in bed mentally composing your next request. You catch yourself wondering how the AI would rephrase something your colleague said in a meeting.

The tool has stopped being a tool. It has become the first place your brain goes when it needs to think.

4. Contradictory Behaviour

This is the one that triggered me to write this post.

I have clients who use AI to ideate. Every week, they come back with a new direction. Sometimes directly contradicting what they proposed the week before. Last week it was a minimalist landing page. This week it is a bold, maximalist redesign. Last week they wanted to focus on SEO. This week they want to kill the blog and focus on video.

They are not being indecisive in the normal sense. They are being served a fresh batch of plausible-sounding ideas every time they ask, and they have lost the ability to filter which ones actually align with their strategy. The AI does not know their business. It just answers the question in front of it.

If your ideas are changing every week and you cannot explain why, you might be in the loop.

5. Withdrawal

Try this experiment. For one day, do not open ChatGPT or Claude. Write something on your own. Trust your own judgement. Publish it without asking for feedback.

If that thought makes you uncomfortable, that is the withdrawal talking.

The discomfort is not because your work will be worse. It is because you have outsourced the act of deciding when something is finished. And when you take that back, it feels like flying blind.

Why It Feels So Good

The AI feedback loop is designed to feel productive. Every response is prompt, detailed, and confident. It never says "I think you are overthinking this." It never says "stop." It validates the act of asking by always having an answer.

That is the hook. Not the quality of the output, but the feeling of progress. You are doing something. You are improving. You are refining. Except you are also slowly replacing your own judgement with a system that will never tell you to stop.

Your website does not need to be perfect. It needs to be yours. If you want to read more about why over-polishing is its own problem, I wrote about that here.

How to Break the Cycle

Set a rule before you start. Two rounds of feedback, maximum. After that, the draft is done.

Write your own first draft without AI. Use it for polish, not for thinking. When you ask for feedback, ask a specific question, not "can you improve this?" A vague question invites an endless list. A specific question gets a useful answer.

And most importantly, start trusting the version of you that existed before the tools. The one who made decisions without a second opinion from a language model. That person was probably right more often than you remember.

As I have learned over the years loving (and sometimes hating) web design, the best work always comes from conviction, not from consensus. AI can help you refine your ideas. But it cannot tell you which ideas are worth having. That part is still yours!