Victoria Okor
Victoria Okor
How many times have you turned to AI for answers today? This week? This month?
When was the last time you read a book and challenged what the author said? Have you had a conversation with someone on politics or education without first confirming what an AI had to say about it? How about your groceries? That project you’re procrastinating?
How many times do you jump from one LLM to the other—Gemini, ChatGPT, Claude, Copilot—an endless list to seek validation from a system that was trained on humans, with humans? Even with the persistent disclaimer—“AI can make mistakes”—somehow it gives that illusion of authority.
Look at the numbers. According to the Gallup Q1 2026 Workforce Report, at least 28% of professionals in the US alone use AI multiple times a week, with a core 13% using it daily for almost every cognitive task. According to reporting from TechCrunch, using data first reported by Axios, ChatGPT processes 2.5 billion prompts every day—roughly 1.7 million requests every minute. That’s ChatGPT alone.
What’s fascinating is that despite the disclaimers, and the hallucination rates (as tracked by the Vectara Hallucination Leaderboard) still hovering between 15% and 25% for complex reasoning, nearly 62% of users admit they trust AI outputs as a dependable indicator of authenticity during their initial interactions. This is absolute psychological dependency or codependency at its peak.
Snapping the Illusion
Generative AI feels authoritative because it sounds familiar. It feels novel, even original, because it is rearranged fragments of human language and a recombination of thoughts dolled up as insight, but an imitation nonetheless. LLMs follow pattern mimicry. They predict the next word, the next phrase, the next sentence not because they understand, but because statistical probability tells them what usually comes next. “We” trained them.
This mimicry is seductive precisely because it is fluent. rhythms we recognise and structures we understand, but beneath, there is no cognitive reasoning, only mimicry.
Creating an original thought stems from recognising them for what they are: generative mimicry. Tools to enhance or remind, not replacements for intellect and/or critical thinking.
Cognitive Detox
Contrary to popular opinion, creating original thoughts is a skill. Like a physical muscle, it needs to be exercised and challenged. You can train your mind to create premises, challenge premises, and resist premises through curiosity, research, and imagination.
Relying totally on AI is a cognitive shortcut—and cognitive shortcuts are where original thoughts go to die. Think of the daily habits:
- Checking AI before you write a single sentence—instead of drafting your own idea first.
- Waiting for others to post their opinions on social media before you decide what you think.
- Agreeing with a polished answer without evaluating it, because it sounded ‘intelligent’, not because it’s proven.
These shortcuts drown you in information but leave you illiterate.
Here’s the pivot: humans have an innate capacity to generate infinite novel thoughts. And when we bypass the AI’s pragmatic bias—thinking whatever sounds coherent and easy to process is relevant—we raise our threshold for originality.
Creating an Original Thought
If our human value lies in our unique perspective, and that perspective is laundered through LLMs trained on the “global average,” then isn’t that value diluted in and of itself? The ultimate metric of AI Insobriety isn’t that the machine makes mistakes; it is that we no longer have the cognitive autonomy to care.
Literature expands our worldview. We must reintroduce friction into our intellectual lives. Pick up a book, research an interest, take a walk, listen in on politics, education, or tech. Have a conversation with someone, or yourself, or a pet. Observe nature, learn a new language, and document your findings.
You could incorporate these three practices:
- Pre‑commitment drafting: Write your position on something before querying any source, AI or otherwise. Even three sentences can force you to actually think before introducing external input.
- Adversarial reading: When you read anything, identify the one claim you most want to disagree with and build the case against it.
- Delayed synthesis: Gather information, then sit with it for a defined period before forming a conclusion.
Sit with the information you have gathered. Wrestle with it. Try to make sense of it and form your own opinions without seeking a digital latch. When it starts to get uncomfortable, you know you’re doing the right thing.
