8-minute read
Why AI should strengthen our questions rather than replace our thinking
A marketing student recently told me that he had to meet me in person. He and others in his class had begun to wonder whether I - and perhaps Azurlis™ itself - were artificial intelligence.
I was not sure whether to feel complimented or mildly insulted. Had my work appeared unusually consistent and well informed? Or had I somehow become too polished to be plausible as a human being? Either way, I apparently required physical verification.
For the record, I am real.
I founded Azurlis™ in 2008. I formulate its products, make decisions, change my mind when evidence warrants it, muddle the occasional sentence and laugh far too loudly.
I am also 68, which means my existence predates generative AI by a comfortable margin.
I do use AI. I use it to test ideas, organise complicated material, improve clarity and challenge my own assumptions. What I do not do is hand over responsibility for what I publish. That distinction matters, because the important question is no longer simply whether somebody used AI.
We need to ask how it was used, what knowledge guided it and whether a human being remained willing to think.
Why genuine expertise can now look artificial
The internet has always contained a mixture of knowledge, advertising, confidence and nonsense. Generative AI has intensified the problem because it can produce fluent language in seconds. A weak idea can now be dressed in excellent grammar. A shallow explanation can arrive with headings, balance and apparent authority. A fabricated expert can look remarkably organised.
This changes how we interpret professional communication. Writing that is clear, detailed and consistent may now trigger suspicion. People sometimes assume that a real human being should be slightly chaotic, while polished language must have come from a machine.
That is an understandable reaction, but it creates an odd penalty for genuine expertise. A scientist who has spent decades learning how to assess evidence may sound methodical. A founder who has worked with the same principles since 2008 may communicate consistently. A formulator should be able to explain why an ingredient is present, what it can reasonably do and where its limitations lie. None of those qualities proves the use of AI. They may simply reflect years of thought.
At the same time, a smiling face on camera does not prove authenticity. Human beings can repeat generated scripts, rent authority through borrowed credentials or manufacture an apparently candid moment. A shaky video and an unfiltered face can be just as calculated as a studio advertisement.
Authenticity is therefore harder to establish from surface clues. It becomes visible over time through coherence between what a person says, what they make, what they know and what they are prepared to correct.
The greater danger is intellectual passivity
Public discussion about AI often rushes towards the spectacular: superintelligence, autonomous weapons, disappearing professions and machines that might one day control decisions of enormous consequence. These questions deserve serious attention. Yet a quieter risk is already present in homes, classrooms and workplaces.
We may stop asking questions because an answer arrives quickly and sounds convincing.
Fluency is not evidence. A generated answer can be useful, insightful and beautifully expressed. It can also be incomplete, outdated or wrong. The same is true of human answers, of course. The difference is that AI can produce plausible explanations at extraordinary speed and scale, often without showing the uncertainty that should accompany them.
The danger is not that a machine can form a sentence. It is that we may confuse a well-formed sentence with a well-founded conclusion.
This concern is not confined to people who are suspicious of technology. UNESCO has argued that learning in the age of generative AI requires independent judgement, critical thinking and emotional intelligence. The OECD similarly describes AI literacy as the ability to engage with AI, create with it, question it and manage it critically. These are not instructions to avoid AI. They are instructions to remain mentally present while using it.
Thinking is slower than retrieving. It requires us to compare sources, notice contradictions and tolerate the discomfort of not knowing immediately. It also asks us to recognise the boundaries of our own expertise. Those habits can feel inefficient when a confident response is available in seconds, but efficiency is not the only value that matters. A fast wrong answer can waste more time than a careful question.
Access to information is not the same as understanding
We have heard a version of this argument before: if everything can be found online, why do we need educators or experts? AI has made the claim more seductive because it does more than retrieve information. It can summarise, explain, compare and imitate reasoning.
But access to information does not automatically create understanding. Knowing a collection of facts is different from knowing which facts matter, whether they apply to the present case and what might invalidate the conclusion.
In skincare, an ingredient list can tell us what is present, but it cannot reveal every detail of concentration, processing, stability, compatibility or product performance. A study may report an impressive percentage, but the number means little until we ask about sample size, controls, duration, measurement and relevance to the claim being made. Expertise lives partly in those questions.
The same principle applies outside science. A legal answer depends on jurisdiction and current law. Health information depends on the person, the evidence and the limits of general advice. A business recommendation depends on assumptions that may never appear in the final paragraph.
AI can help us explore these matters. It can identify questions we overlooked, translate technical language and expose gaps in an argument. It cannot remove our obligation to decide when a claim needs verification or when a qualified professional must take responsibility.
Using AI without outsourcing judgement
I find the idea that we must choose between embracing AI and rejecting it unhelpful. Tools do not become wise merely because they are powerful, and refusing a useful tool does not make a person more thoughtful.
A better practice is to decide which parts of a task can be assisted and which parts must remain accountable to human judgement.
I may use AI to help organise a long article, suggest a clearer sequence or argue against my first conclusion. I still need enough knowledge to recognise an error, reject an inflated claim and preserve the nuance that matters. If I cannot evaluate the result, I should not present it as my authority.
This is especially important when someone claims expertise. AI can imitate the language of competence, but it has no professional reputation to protect and no personal consequences when advice fails. The person or organisation publishing the result remains responsible.
A useful AI relationship should increase intellectual friction at the right moments. Ask it for the strongest objection. Request the assumptions behind an answer. Look for evidence that could disprove the preferred conclusion. Check original sources. Separate what is known from what is inferred. Then apply the judgement that comes from education, experience and context.
Used this way, AI can expand thought rather than contract it. It can be a tireless research assistant and an occasionally exasperating debate partner. It should not become an oracle.
What trustworthy human expertise looks like
If polished language no longer proves expertise, what should we look for? I would begin with specificity. A trustworthy person can usually explain how they reached a conclusion, what evidence informed it and what limitations remain. They distinguish observation from interpretation and do not turn uncertainty into a marketing inconvenience.
Consistency matters, but so does correction. Someone who never revises a position may be protecting an image rather than following evidence. Genuine expertise includes the confidence to say, 'This is what the evidence supports at present,' and the humility to update that view when better evidence arrives.
Accountability matters most. Who formulated the product? Who checked the claim? Who will answer questions when something is unclear? Technology can support each step, but responsibility cannot disappear into a software system.
For Azurlis™, this means that I remain visible in the decisions behind the brand. I explain why I choose certain ingredients, why I reject others and why formulation matters more than a fashionable ingredient name. I do not claim that natural automatically means safe, that synthetic automatically means harmful or that a cosmetic can reverse time. AI did not create those standards. They come from my scientific training, formulation experience and values.
Perhaps I should show more of the person around that knowledge: the working process, the second takes, the laughter and the occasional magnificent verbal disaster. Not because imperfection is proof of humanity, but because trust grows through a fuller relationship than polished copy alone can provide.
Keep the questions alive
My unexpected identity investigation by a marketing class was funny, but it exposed a serious change in how we judge information. We can no longer assume that fluent writing came from a knowledgeable person. We also should not assume that fluent writing could not have come from one.
The answer is not to demand that humans become less articulate so they look authentic. It is to become more curious about the substance beneath the presentation.
Ask who is speaking. Ask what they know. Ask what evidence supports the claim. Ask what has been omitted, what remains uncertain and who accepts responsibility for the result. Apply the same questions to a person, a brand, an influencer, a news report or an AI response.
I am not frightened by the existence of a tool that can help us think. I am concerned by any culture that encourages us to stop thinking because the tool is convenient.
So yes, I use AI. I also question it, correct it, laugh with it and occasionally argue with it. The final judgement remains mine.
And should anyone still require confirmation: I am Dr Fernanda. I am a scientist, formulator, founder and very real human being. My unsuccessful AI voice clone, which apparently makes me sound like a moron, may be submitted as supporting evidence.
Sources and further reading
UNESCO. Artificial Intelligence in Education. https://www.unesco.org/en/digital
education/artificial-intelligence
UNESCO. Generation AI: Navigating the Opportunities and Risks of Artificial Intelligence in
Education. 22 July 2024. https://www.unesco.org/en/articles/generation-ai-navigating-
opportunities-and-risks-artificial-intelligence-education
OECD. Designing Safe AI Systems for Education. 23 January 2026.
https://www.oecd.org/en/blogs/2026/01/designing-safe-ai-systems-for-education.html
NIST. AI Risk Management Framework. https://www.nist.gov/itl/ai-risk-management
framework
