The internet did not invent bad medical information. It just gave it infinite shelf space, excellent lighting, and eventually a machine capable of producing more of it than any human could ever check.
THE PATIENT ARRIVES WITH A FOLDER.
Not an actual folder.
A digital one.
Screenshots.
TikToks.
A Reddit thread.
Three articles.
An Instagram carousel.
A wellness account explaining cortisol.
A website that looks medical enough to make you nervous.
One podcast clip.
A screenshot of somebody else's lab results.
And one question:
Could this be what I have?
Doctors have always had to compete with rumor.
Your grandmother knew a woman whose cousin cured something with vinegar.
A neighbor swore that dairy caused every disease known to man.
Somebody at church knew a guy who knew a guy.
The difference is that your grandmother's cousin did not have an algorithm.
Now misinformation arrives searchable, beautifully formatted, professionally narrated, emotionally optimized and available at 2:17 AM when the person reading it is terrified enough to believe almost anything.
Then AI shows up and says:
What if we made more?
The Rabbit Hole Has a Waiting Room
UCLA physician Aparna Sridhar describes a pattern that became familiar in her practice: patients arriving more anxious after going down online health-information “rabbit holes,” sometimes requiring appointments to spend additional time undoing inaccurate information before actual treatment could move forward. She connects this to cyberchondria, the cycle in which repeated medical searching escalates anxiety rather than resolving it.
This feels like one of those terms invented for something everybody already recognizes.
You feel a strange pain.
Google it.
The internet suggests:
Muscle strain.
Gas.
A rare neurological disorder.
Terminal cancer.
Wonderful.
Search again.
Maybe the first website was wrong.
Now you have eight diseases.
The internet is very good at answering the question you typed. It is much worse at knowing whether that was the question you needed to ask.
Health anxiety changes the way people consume information because uncertainty is painful.
The person searching does not necessarily want the most nuanced answer.
They want relief.
Certainty.
A name.
A plan.
Unfortunately, certainty is also one of the easiest things for medical slop to manufacture.
Science Says “Maybe.” Slop Says “Tuesday.”
Real medicine contains an irritating amount of ambiguity.
This drug helps some people.
That symptom could mean several things.
Evidence is mixed.
The trial was small.
The effect is modest.
Your personal history matters.
Please speak with a clinician.
This is excellent medicine and catastrophic content.
Online health misinformation arrives with much better copywriting.
ONE FOOD DESTROYS INFLAMMATION.
THIS DEFICIENCY EXPLAINS EVERYTHING.
DOCTORS MISSED THIS.
STOP EATING THIS IMMEDIATELY.
No confidence interval.
No differential diagnosis.
No seventeen-minute explanation of what “association” means.
Just certainty.
APA researchers describe health misinformation as especially dangerous because false information can rush into the blank spaces while science is still working out the answer. Their reporting notes that low-quality information often travels well precisely because people encounter it in emotional environments where simple explanations outperform complicated reality.
That sentence should terrify anyone who has ever waited three weeks for a biopsy result.
Science is slow because reality is difficult.
Slop treats slowness as a market opportunity.
The Website Has a Stethoscope, So It Must Be Fine
Before AI doctors, before TikTok wellness gurus, before ChatGPT could confidently explain your bloodwork at midnight, the web already had a credibility problem.
The National Center for Health Research gives a wonderfully boring example.
Type cancer.com into a browser and you might reasonably assume you have arrived at the American Cancer Society.
You haven't.
The American Cancer Society uses cancer.org.
Cancer.com, the organization points out, is a commercial site associated with Janssen, a pharmaceutical company owned by Johnson & Johnson. That does not automatically make the information false, but it does mean the person reading deserves to know who built the room they just walked into.
This is the older version of medical slop.
Nothing needs to be fake.
It only needs to look more neutral than it is.
Health information has always had sponsorship, incentives, selective framing and commercial interests.
AI does not create these incentives.
It industrializes their ability to produce content.
The medical website does not have to lie to mislead you. Sometimes it only has to forget to explain who paid for the furniture.
This is why “just look for sources” has never been enough.
Who owns the source?
Who funds it?
What does it sell?
What information was included?
What disappeared?
A citation can be real.
The framing can still be slop.
Then the Machine Reads the Internet
This is where the garbage-in problem becomes literal.
Generative systems are very good at producing answers from patterns in enormous bodies of human information.
Unfortunately, human information contains us.
Our research.
Our reporting.
Our medical guidelines.
And also:
our miracle cures,
our conspiracy theories,
our bad summaries,
our affiliate marketing,
our old information,
our oversimplified headlines,
our wellness grifts,
and our guy on Facebook who says his knees stopped hurting after he stopped eating tomatoes.
The machine does not encounter “the internet” as a clean medical library.
It encounters a landfill with several excellent libraries somewhere inside it.
Then we ask it for a summary.
The Lie Does Not Have to Look Stupid
One of the more comforting myths about misinformation is that we will recognize it.
Bad grammar.
Crazy font.
Suspicious website.
Guy screaming into a webcam.
Easy.
The APA's work on misinformation psychology is less reassuring.
People often focus first on understanding new information rather than evaluating its truthfulness, and plausible falsehoods can be learned as fact. Anxiety can also increase susceptibility. Repetition matters. Familiarity matters. Source cues matter.
Which means modern medical slop is becoming dangerous at exactly the moment it becomes aesthetically competent.
The voice is calm.
The graphics are tasteful.
The claim includes a study.
The account has a blue check.
The website says Institute.
Nothing looks like a scam.
That is the scam's career goal.
False Information Has Better Timing
There is another unfair advantage.
Medical science often arrives after uncertainty.
The patient wants to know today.
The evidence may take years.
Health misinformation does not suffer from this scheduling problem.
During the Covid era, one small, nonrandomized study involving only 36 patients helped drive enormous interest in hydroxychloroquine before stronger evidence caught up. The APA uses cases like this to show how incomplete or weak evidence can spread rapidly when it supplies an answer people desperately want.
This pattern is bigger than Covid.
When reality says:
“We don't know yet,”
the slop economy hears:
“Available advertising space.”
The Patient Has to Unlearn Before They Can Learn
This may be the most expensive part of the problem.
Once misinformation enters someone's mental model, correcting it is not the same as simply adding the right answer.
The clinician may have to explain:
why the claim is wrong,
why it sounded plausible,
why the study does not mean what the post said,
why this influencer is not qualified,
why your symptoms do not necessarily mean what the forum says they mean,
and why the doctor telling you this is not merely part of the conspiracy.
APA's consensus recommendations emphasize that misinformation correction works better when it does more than say “false.” Effective responses include explaining why a claim is wrong, providing an alternative explanation, using trusted messengers and sometimes prebunking—teaching people how manipulation works before they encounter it.
In other words, medicine now has to provide treatment and media literacy.
Very efficient system we've built here.
We Keep Blaming the Patient
There is an irritating moral tone that often appears around health misinformation.
Why did they believe that?
Why didn't they check?
Why would anyone take medical advice from TikTok?
This misses the architecture.
People search for health information because healthcare is scary, expensive, confusing, inaccessible, rushed or all five at once.
UCLA notes that good online information can genuinely reduce anxiety and help patients understand their care. The problem is not that people look online. The problem is that useful, evidence-based information arrives in the exact same environment as commercial promotion, anecdote, misinformation and algorithmically rewarded panic.
A patient should not need a graduate seminar in epistemology to figure out whether a rash requires a dermatologist.
And yet.
The Algorithm Has Terrible Bedside Manner
Social media has another problem.
Its objective is not health.
It is attention.
The platform does not ask:
Will this improve the patient's understanding?
It asks:
Will this person keep watching?
Those goals overlap occasionally.
They are not synonyms.
APA researchers note that low-quality and false content often receives substantial engagement and that misinformation spreads partly because emotionally compelling claims are easier to share than dry corrections.
Fear works.
Hope works.
Anger works.
Secret cures work spectacularly.
“Talk to your physician” does not exactly slap.
This means medical slop is not merely an information problem.
It is an incentive problem.
The truthful answer is competing inside a machine that may structurally prefer the more exciting wrong one.
AI Doesn't Need to Invent the Garbage
This is the part I think matters most.
We talk about AI medical slop as though the machine is going to wake up one morning and invent a thousand dangerous lies.
It does not have to.
We already made them.
AI's more transformative role may simply be taking the existing ecosystem of weak evidence, health anxiety, commercial incentives, miracle claims and context-free advice and giving it production capacity.
Turn one bad claim into:
a video,
a podcast script,
a carousel,
an article,
ten social captions,
a fake doctor,
a patient FAQ,
a newsletter,
and twenty slightly different versions optimized for different audiences.
The machine does not need creativity.
The garbage already exists.
It only needs logistics.
The medical misinformation crisis was already burning. AI arrived with an industrial fan.
The Internet Starts Quoting Itself
Then the problem becomes recursive.
Synthetic summaries are published.
AI-generated articles proliferate.
Other systems retrieve those articles.
Creators use AI to summarize the synthetic summaries.
Search results fill with derivative explanations.
The information environment becomes increasingly difficult to trace back to first principles.
Who conducted the study?
Who interpreted it?
Who summarized the interpretation?
Who summarized that summary?
At what point did “may” become “does”?
At what point did “associated with” become “causes”?
At what point did one mouse become all human beings?
This is medical slop's most boring horror.
Not a robot doctor murdering somebody with a spectacular hallucination.
A thousand tiny distortions becoming normal because nobody remembers where the sentence came from.
Trust the Person Who Can Tell You Who They Are
There is no perfect defense.
But the old boring signals still help.
Who wrote this?
What are their credentials?
Who owns the website?
Is the claim supported by more than one reputable source?
Does the article explain uncertainty?
Is somebody selling something?
Can you find the original study?
Would a serious medical organization phrase the claim this confidently?
APA recommends relying on trusted sources, correcting misinformation with evidence, and building resilience before exposure. UCLA emphasizes evidence-based, credible information rather than clickbait and anecdote. The National Center for Health Research says transparency about ownership and funding matters because apparently neutral health sites may have commercial origins.
None of this is glamorous.
That may be a feature.
The boring answer is often boring because nobody optimized it for virality.
Medicine Has Always Had Garbage
Snake oil existed before broadband.
Fraud existed before ChatGPT.
Bad science existed before social media.
Medical misinformation is not new.
What changed is the machinery.
The old rumor had to travel person to person.
The new one can be generated, packaged, personalized, translated, narrated and distributed before breakfast.
And because health sits so close to fear, the garbage does not merely fill the feed.
It enters appointments.
It changes what patients ask.
It changes what they fear.
It changes what they refuse.
It changes how much time clinicians spend repairing information before they can repair anything else.
That is the garbage-in problem.
Not that machines occasionally say something wrong.
That healthcare now operates downstream from an information system in which the cost of producing convincing medical-looking material is collapsing faster than our ability to verify it.
The answer is not to stop people from looking things up.
The answer is to remember that information has provenance.
Someone made it.
Someone paid for it.
Someone interpreted it.
Someone omitted something.
Someone may be wrong.
And now, increasingly, someone may not be involved at all.
