How Not to Get Lost in the Age of the Internet and AI
The problem stopped being access. The problem is judgment. Learning to tell a fact from a well-dressed guess may be the single most useful skill of this decade: not finding information, but grading it.
Essay · Everyone · EN
Olga Goodman, MD · Rheumatology in Perspective
Published · Updated October 1, 2026
Fifty years ago, finding a medical fact took a library card, a patient librarian, and an afternoon.
Today it takes four seconds and returns forty answers — written with equal confidence by a professor, a supplement company, your cousin on social media, and a chatbot that has never, to its own knowledge, been wrong.
The problem stopped being access. The problem is judgment. Learning to tell a fact from a well-dressed guess may be the single most useful skill of this decade: not finding information, but grading it.
Good information and nonsense now wear the same outfit
A confident answer is not the same as a correct one. A long answer is not the same as a complete one. Ten websites repeating the same sentence are not ten sources — they may be one press release wearing nine disguises. And an AI that explains something beautifully can be explaining the wrong thing beautifully. Fluency was never evidence. It just used to be rarer without substance behind it.
Before “is this true,” ask “what am I even reading”
| What you’re reading | What it actually tells you |
|---|---|
| A patient’s story | What happened to one person — not how often, or whether it would happen to you |
| A randomized trial | Whether the treatment beat the comparison, in that group, for that outcome |
| A guideline | What a panel considered reasonable, as of its last review date |
| An ad dressed as an article | What a company wants you to decide |
| An AI summary | What its sources said, if they existed, and if it read them correctly |
Each of these can be entirely true and still be answering a different question than the one you asked.
Check the date. No, the other date.
A well-designed page from 2016 can outrank a more accurate page from 2026. Search ranking is not the same thing as medical reliability. Ask when something was written. Then ask the sharper question: when was the information in it last checked? A freshly redesigned page can still be serving ten-year-old medicine under a new coat of paint.
Walk one step backward
“Studies show…” Which studies? “Cuts the risk by 50%.” Fifty percent of what, exactly? “Current guidelines recommend…” Whose guidelines, from what year, for which patients? You don’t need to chase every claim like a dissertation topic. But the bigger the decision, the more that one step backward is worth taking.
Be suspicious of total certainty
Honest answers in medicine often sound less confident than dishonest ones. “We know this fairly well.” “The evidence is mixed.” “We genuinely don’t know yet” is not a failure — it is often the most accurate sentence on offer. If a source is never unsure about anything, be a little more unsure about that source.
AI deserves respect, and a specific kind of suspicion
AI can untangle a mess of information, explain a lab result in plain language, compare options, and turn a vague worry into a sharp question. That is a real gift, not a party trick.
But AI has one habit worth knowing: it tends to answer the question you actually asked, in the spirit you asked it. Ask “is this medication dangerous?” and you’ll get a thoughtful list of risks. Ask “is this medication safe?” about the same drug, and you’ll get a thoughtful list of reassurances. Mention that you already don’t trust your doctor, and an AI response may start from that framing rather than independently challenge it. That does not mean the AI is lying; it may simply be following the frame you supplied. That is not quite the same as being a neutral referee for the truth you did not ask about.
So ask the same question from opposite directions before trusting either answer. Then ask: “What evidence would argue against the answer you just gave me?” And ask for what can be checked — a DOI, a study name, a real guideline — not just a well-turned paragraph. Polish is not proof.
Disagreement is not incompetence
Two good physicians can study the same incomplete chart and land on two reasonable, different conclusions. That is not a crisis of medicine. It is medicine. The useful question is rarely “who’s right.” It is “why did they land somewhere different?” That question teaches more than a winner ever could.
And sometimes, just stop
Searching is the one activity with no natural finish line. Every answer opens three more tabs, and at some point more information buys less clarity, not more. The useful shift is from collecting everything to sorting what matters: what’s established, what’s genuinely unsettled, and what would actually change your decision. That sorting, not a bigger pile of links, is most of what expertise is for.
The internet and AI are not the enemy
Used well, both make people sharper, not lazier. But they changed the job. Finding information used to be the hard part. Now filtering it is. The goal was never to trust nothing. It is to know what deserves trust, what deserves a second look, and what deserves to be ignored, politely and without guilt — including, now and then, an AI’s first answer, offered a little too eagerly to please.
Educational content from the author’s personal website; not individual medical advice or a clinic portal.