What you’ll find here

Every post is a lesson. We take the AI slop flooding your feed, explain exactly why it spreads, and give you the tools to recognize it before you hit share. Digital literacy, wrapped in humor. No tech background required.

The Sloppies are our monthly awards - celebrating the internet’s best worst AI-generated content. The community nominates. The community votes. The Slopfather presides. Nominations are always open at thesloppies.com.

Where to start

• The Field Guide - five tells that catch 80% of what your feed is serving you.

• Want to understand why smart people fall for it? Start with Why Smart People Share Slop

• Ready to see the hall of fame? Browse The February Winners

The Slopfather is watching. You should be too.

Subscribe free. Stay sharp.

Subscribe now

Someone made it in about four seconds. Someone else shared it. Then someone else shared it. By the time you saw it, it had already passed through enough hands to feel like news.

This guide is for you. Not the tech person. Not the cybersecurity expert. You. The person who uses the internet every day and does not want to be the one who accidentally spreads something fake.

You do not need special tools. You do not need to understand how AI works. You need five things to look at. That is it.

Tell #1: Look at the Hands

AI is extraordinarily good at faces. It has seen billions of photographs of human faces and it has gotten very good at making faces look real.

Hands are a different story.

Hands are complicated. Every hand is slightly different. The way fingers bend, the way knuckles sit, the way a fist closes — these things vary endlessly, and AI has not figured them out the way it has figured out eyes and noses and smiles.

So before you share an image of a person, look at the hands.

Count the fingers. Look at how they sit. Ask whether a real hand could actually be in that position.

If the hand has six fingers and the person is not a cartoon character, someone generated that image. If the fingers look melted into each other, someone generated that image. If the hand is gripping something in a way no hand has ever gripped anything, someone generated that image.

The face will lie to you. The hands almost always tell the truth.

Tell #2: Read the Signs

AI knows what signs look like. It does not know what signs say.

Think about that for a moment. If you asked AI to generate a photo of a busy coffee shop in Paris, it would give you golden light and wooden tables and a chalkboard menu on the wall. The chalkboard would look exactly like a chalkboard menu.

But if you read what the chalkboard says, it might say something like: “ERRESS & COFFEE PUBIR.”

That is not a language. That is AI doing its best impression of what letters arranged on a sign look like, without understanding that letters are supposed to mean something.

This works on everything. Storefront windows. Newspaper headlines. Book covers. Name tags. Product labels. Menu boards.

The rule is simple: zoom in and read the text. If the words are nonsense — not a foreign language you do not recognize, but actual nonsense — the image is almost certainly AI-generated.

Real photographs have real words in them. AI-generated images frequently do not.

Tell #3: Ask Why This Exists

This is the most important one. The others are about catching fakes. This one is about catching the ones that want something from you.

Every piece of content on the internet was made by someone with a reason.

Before you share something, ask: what is this asking me to do?

Sometimes the answer is obvious. Like the page. Follow the account. Share this to raise awareness. Donate to this cause. Sometimes the answer is invisible, which is actually more concerning. An image that just wants you to feel something — outrage, sadness, inspiration, fear — and share it is generating what the people who study this call “engagement.” Engagement means money. The more you react and share, the more money the account makes from ads and partnerships.

Ask yourself: does this image want me to do something?

If the answer is yes, slow down. Ask who made it and why. Look for a source. If there is no source, that is information.

The fake flood with the crying child wanted you to comment “praying.” The inspirational quote attributed to no one wanted you to follow the account. The veterans’ page with AI-generated photos of wounded soldiers wanted you to share, so that more people would see the ads running alongside it.

They all had a reason. Knowing that they have a reason is the first step to not doing what they want.

Tell #4: Check the Background

AI spends most of its effort on whatever the prompt asked for. If someone asked for a photo of a woman at a dinner party, AI will put a lot of work into the woman and the table and the food.

The background gets less attention.

Look at the edges of AI images. Look at what is happening behind the main subject. Walls that warp. Furniture that fades into nothing. Windows that look out onto a blurry smear that used to be a city. Shelves where the books have no titles. Staircases that go nowhere. Doorways that do not quite connect to the floor.

The center of the image is often convincing. The edges are where AI runs out of instructions and starts making things up.

This is especially useful for indoor scenes: kitchens, living rooms, offices, restaurants. The table will look perfect. Walk your eye around the edges of the room. Something will be wrong.

Tell #5: Does the Story Check Out?

AI can generate images. It cannot generate reality.

This tell is the simplest one: if an image is claiming to show something real, ask whether that thing actually happened.

This does not require any technology. It requires one question.

If someone shares a photograph of a celebrity at an event, that event was either a real event that was covered somewhere, or it was not. If you search the celebrity’s name and that event, and nothing comes up, the photograph probably did not come from that event.

If someone shares a news photograph of a natural disaster, that disaster either happened and is being covered by news outlets, or it did not. If you search for that disaster and the only results pointing to it are the viral post, something is wrong.

Most people skip this step because the image looks real and the caption says it is real and fifty people already shared it. Fifty people sharing something is not evidence that it is true. Fifty people may have all made the same mistake. They were trusting each other. Nobody checked.

You can check. It takes about thirty seconds and a search.

The Five Tells, All in One Place

1. Hands. Count the fingers. Look at how they sit. Hands are where AI falls apart.

2. Text. Zoom into any signs, labels, or headlines. If the words are nonsense, the image is fake.

3. Why does this exist? Every piece of content wants something from you. Know what it wants before you give it.

4. Background. The main subject is convincing. The edges are where AI runs out of instructions. Look at the edges.

5. Does the story check out? If the image is claiming to show something real, thirty seconds of searching will usually tell you whether it happened.

That is the whole guide.

You do not need to know anything about AI. You do not need any apps. You do not need to be young or technical or fluent in anything digital.

You need thirty seconds and a habit of asking one question before you hit share: does this make sense?

Most of the time, if you look, it does not.

The Sloppies exist because looking matters. Nominations for March are open at thesloppies.com. You have already seen slop today. Send it in.

Which of the five tells has already caught you out? Share the slop you almost shared in the comments.

The Slopfather keeps the records so you don’t have to swim in the slop alone. If this was useful, share it with someone who needs it.

Subscribe now

Everyone’s instinct is to ask how they missed it.

That is the wrong question.

The actual question

Your brain is not built to verify everything. It cannot be. You absorb hundreds of pieces of information every single day. If you stopped to fact-check each one, you would not make it to lunch.

So your brain shortcuts. It pattern-matches. It asks: does this look like a thing I already trust? If yes, it files it and moves on. That is not a flaw in your intelligence. That is your intelligence working correctly under load.

The problem is that AI learned to make things that look exactly like things you already trust.

The slop did not beat anyone’s intelligence. It beat the part of the brain that decides whether to use intelligence at all.

The Reddit post had the right format. The right subreddit. An account with history. Early upvotes from other accounts with history. By the time most people saw it, the post had already passed through fifty people who did not flag it. That is fifty reasons not to look closely. The fifty-first person saw fifty votes of implied approval and started reading from inside a false sense of safety.

None of those fifty people were stupid. They were just earlier in the same trap.

Worth noting: Hard Fork, Feb. 13, 2026

Kevin Roose and Casey Newton had a New York Times reporter on to talk about writers using AI to pump out romance novels. Readers are buying those books thinking a human wrote them. Not because the books fooled any serious test. Because nobody ran one. When you buy a book, you assume it was written by the person whose name is on the cover. The platform does not correct you. The assumption just becomes your experience.

The fatigue nobody is talking about

Hard Fork’s March 13 episode introduced something called “AI brain fry.” A researcher named Julie Bedard studied what happens to workers who deal with AI output all day. They get worn down. They stop checking things. Not because they stopped caring. Because checking everything is not something a human body can sustain at the volume AI produces.

Think about the last time you were genuinely exhausted at the end of a workday. Now imagine that state is the condition under which you are expected to spot AI-generated misinformation in your social feed.

That is the actual situation.

I spent fifteen years in cybersecurity watching this exact pattern destroy otherwise solid security programs. You build a system that generates alerts. The alerts multiply. The team starts missing things. Not because the team got worse. Because the volume broke the team’s ability to care about each individual alert. The industry has a name for it: alert fatigue. The solution was never “be more careful.” The solution was always fixing the system that generates the alerts.

Also worth noting: Hard Fork, March 13, 2026

Casey Newton found out Grammarly had used his name and identity in an AI feature without asking him. A tech journalist who has been reading the AI industry closely for over a decade. He did not catch it until after the fact. That is not carelessness. That is what it looks like when the attack surface for this stuff gets bigger than any one person’s ability to watch it.

The part nobody wants to admit

Some of the shares were not even about the content.

People share things to perform an identity. The person who shares a good deal is being the person who finds good deals. The person who shares the hot restaurant recommendation is being the person who knows about hot restaurants. The content is a costume. The share is the statement.

Checking whether the restaurant actually exists would have ruined the whole point.

This is not new human behavior that AI created. AI just made it cheaper to produce the costume.

The Verdict

Smart people share slop because the platforms built systems where social proof looks identical to actual proof. AI learned to manufacture social proof at scale. And humans can only run manual verification on so much before the tank runs empty. This was never about intelligence. It was about infrastructure that was designed, tested, and optimized to make verification feel unnecessary.

Telling smart people to be more careful is not a fix. It is an assignment of blame that lets the platform off the hook.

The slop is not the problem. The slop is the symptom. The problem is a system that profits every time you share something, and has no financial interest in whether that something is real.

Subscribe now

The Slopfather covers the AI content flood so you do not have to swim in it alone. If you shared something that turned out to be slop, the designers say thank you! You were the infrastructure working as designed.

Cover art: by myself using Microsoft Paint. I figured you did not need another piece of AI art.

Here it is.

Nine categories. Nine winners. One through-line that connects all of them: AI is extraordinarily good at looking right and catastrophically bad at being right. The aesthetic passes inspection. The logic does not survive thirty seconds of attention.

That gap is the tell. It is always the tell.

Finger Crimes: Stubbie Adult Schoolboy Action

39 votes

An AI Indiana Jones type. Sitting in a classroom. Hand on a desk. The caption says “Prove It.”

The hand on that desk is committing a quiet anatomical felony.

Finger Crimes is the category everyone understands immediately, because hands are something every human being has spent their entire life looking at. We know what five fingers looks like. We know it without thinking. AI does not know it the same way, and the result is visible the moment you look at the hand instead of the face.

The tell here is not just the finger count. It is the confidence. This image dares you to look closer while the evidence is sitting right there in plain view. That is the AI equivalent of a con artist asking you to check their references.

The lesson: AI generation prioritizes the face. The hands are an afterthought. When you suspect an image, stop looking at the eyes. Look at the hands. They will tell you everything the face is trying to hide.

Alphabet Soup: Coffee Sl0p

45 votes

A Parisian coffee shop. Perfect aesthetic. Golden light. The kind of place you want to sit in for three hours with a book.

The window says “ERRESS & COFFEE PUBIR.” The door says “CITY ACCIT SEVIE BANIS.”

“COFFEE SHOP” lands correctly, which is suspicious in retrospect. That one phrase lulled the eye into assuming the rest would follow. It did not follow. The rest went somewhere unknowable.

AI understands what a sign looks like. It does not understand what a sign says. The visual grammar of language, the shapes of letters arranged in rows, is something AI can replicate with confidence. The semantic content of those letters is a different problem entirely.

The tell: zoom into any text in an AI image. Storefront signs. Newspaper headlines. Product labels. Menu boards. If the words are real, they are almost certainly in the title position where the prompt specified them. Everything else is decoration that looks like language without being language.

The practical application: the next time you see a restaurant review with a photo of the storefront, look at the menu board in the window. If you cannot read it, the photo is probably not real.

Uncanny Valley: Uncanny Babie

40 votes

A baby. A bib. A bowl of what appears to be wet cement.

AI has mastered the visual vocabulary of mealtime. The mess. The splatter. The small fist gripping a spoon. Every element is present and correctly deployed. The food is structural concrete.

The Uncanny Valley category exists for exactly this: content that is almost right in every detail and catastrophically wrong in one specific way that the human brain registers immediately without being able to articulate. Something is wrong. Something is deeply wrong. You cannot say what it is until you look at the bowl.

The tell: AI optimizes for what things look like in aggregate, not what individual components are made of. The scene is correct. The substance is not. When an image feels off and you cannot name why, look at the specific materials. Ask what each element is actually made of. AI frequently substitutes visual texture for physical reality.

Fakefluencer Slop: AI Vet

35 votes

A young woman. A wheelchair. A military uniform with garbled insignia. Generated in seconds to harvest real human empathy from real human beings who genuinely care about veterans.

This is the category that should make you angry, and I say that as someone who tries to keep the tone here precise rather than emotional.

The other categories are lazy. Bad prompts. Careless outputs. Accidentally funny. The Fakefluencer Slop winner is none of those things. It is engineered. The wheelchair is deliberate. The youth is deliberate. The uniform is deliberate. Every element was selected because it triggers a specific emotional response that overrides the inspection instinct.

The tell for this category is different from the others. It is not visual. It is behavioral. Ask why this image exists. What is it asking you to do? Like, share, comment “respect,” follow the page. The engagement ask is built into the image before you ever see it. AI slop of this kind is not accidental content. It is a system for converting human compassion into platform metrics.

The security framing here is exact: this is a social engineering attack. The payload is your attention and your credibility. When you share it, you are the delivery mechanism.

LinkedIn Slop: Biznass Prof

5 votes

A confident AI businessman. An unlabeled brown bottle. The word “inspiration” underneath.

Five votes won this category. That tells you something about the state of LinkedIn Slop submissions in Month One. The competition was not fierce. The category will get better.

The winner is perfect in its specific way. A man asked AI for wisdom. AI told him: inspiration. He is now holding a mystery liquid to celebrate this revelation. No context. No product. No explanation. The bottle exists because the image needed something for his hands to do and AI defaulted to a vessel of undefined significance.

The tell for LinkedIn Slop is the thought leader posture: the knowing look, the aspirational caption, the complete absence of anything being said. A human with something to communicate communicates it. An AI asked to generate a professional post generates the visual grammar of communication without the content.

The content exists to look like content. That is its entire job.

AI Slop Art: fAIry Slop Princess

26 votes

A warrior princess. Armor that has never met actual armor. A sword grip suggesting she learned to hold weapons from a YouTube thumbnail. Background fog because environments are hard.

She looks incredible. She would lose a fight to a strongly-worded letter.

AI Slop Art as a category is specifically about the failure of functional logic. The aesthetics are extraordinary. The engineering is not. The armor is engraved with impossible detail and would provide no protection to the wearer. The sword hilt is held in a way that would dislocate a wrist on the first swing.

The tell: AI knows what things look like in art. It does not know how things work in practice. Armor, weapons, tools, machinery: AI generates the visual impression of function without the underlying mechanical logic. Ask whether the thing depicted could actually do what it is depicted as doing.

The princess looks like a warrior. The armor disagrees.

Physics Doesn’t Work That Way: Bldg Amiss

23 votes

A stunning architectural photograph. A reflection pool.

The building says “ERREESS M COFFEE PUBIR.” The reflection says “INSPIRATION.”

These are not the same sign. They are not the same word. They are not the same building. The reflection pool is showing you something more motivational than what’s physically present. The AI generated a mirror that decided to editorialize.

This is the category for content where AI understood the premise and failed the execution at the physics level. A reflection pool reflects. This one interprets. The tell is to look at any mirror, reflection, or shadow in an AI image and ask whether it is showing you the same reality as the object casting it. It almost never is.

The bonus tell for this winner: the text. “INSPIRATION” is a word. “ERREESS M COFFEE PUBIR” is not. The reflection corrected the sign. The AI knew the sign should say something legible, put that understanding into the reflection, and forgot to apply it to the building itself.

Historical Crimes: Faaake Floodie

10 votes

A crying child. A soaking wet puppy. A flooded neighborhood.

Nobody was rescued. The puppy does not exist. The child is not real. The flood was generated in four seconds to collect “praying” comments from people who were genuinely moved.

Ten votes won this category. That is the lowest vote count across all nine winners. I do not think it is because this was the weakest entry. I think it is because this category is harder to sit with than the others.

Historical Crimes covers AI content that fakes real events: disasters, news moments, historical photographs. The tell is the emotional architecture. Real disaster photography is chaotic and imperfect. The lighting is wrong. The framing is accidental. Someone had a camera in a terrible situation and pointed it at something.

AI-generated disaster content is optimized. The child is centered. The puppy is adjacent and perfectly lit. The water is at the ideal level for maximum pathos. Everything is where it needs to be for maximum emotional impact because the image was designed for maximum emotional impact.

Real grief does not compose itself. This kind does.

Shrimp Jesus: Shrimp Jesus...Need I Say More

126 votes

Jesus. A shrimp. The beginning of something.

126 votes. The next closest winner had 45. This was not a close race. This was the community recognizing the origin point of a cultural moment and voting accordingly.

The Shrimp Jesus Memorial Award exists because this image spawned a genre. 120 Facebook pages built content empires on this format. The theology is unclear. The engagement was not. Real churches collected fewer “Amen” comments than this image of a savior riding a crustacean toward the surface of a body of water.

The tell for Shrimp Jesus content is the theological coherence test. Does the image make sense within any established religious tradition? If the answer requires a long explanation, someone generated it for engagement rather than devotion.

The image is technically flawless. The theology is a little unclear. Those two things existing simultaneously is the entire Shrimp Jesus Memorial Award.

What Nine Winners Tell You

Every winner in February failed in the same fundamental way. The surface was correct. The underlying reality was not. The armor looked like armor. The reflection pool looked like a reflection pool. The food looked like food.

AI is solving the visual problem. It is not solving the logical problem. It does not know how armor works. It does not know what reflections do. It does not know what babies eat.

That gap, between how things look and how things work, is where every AI tell lives. You do not need a detection tool. You need thirty seconds and a question: does this make sense?

Most of the time, it does not.

Which February winner surprised you most - and which category do you think March will make worse? Let’s hear your predictions.

The Slopfather keeps the records so you don’t have to swim in the slop alone. If this was useful, share it with someone who needs it.

Subscribe now

March nominations are open now at thesloppies.com. The community found nine extraordinary failures in Month One. Month Two will be worse. Submit your finds before March 21.

Next issue: the Slopfather’s field guide to spotting AI slop in the wild. The five tells that catch 80 percent of what your feed is serving you.

There is a video on YouTube of a cartoon elephant swimming in a pool.

The elephant enters the water at a physically impossible angle. Its trunk briefly detaches. A cheerful synthetic voice announces the letters of the alphabet in no particular order. The elephant's face cycles through four expressions in two seconds. The video is forty-one seconds long. It has nine million views.

This is not a failure of the content moderation system. This is the content moderation system working correctly and leaving the video alone because it does not technically violate any policy that currently exists.

A New York Times investigation published last week documented what happens after a toddler watches a single CoComelon video on YouTube. Within a fifteen-minute session, more than forty percent of the recommended Shorts contained AI-generated visuals. The clips featured warped faces, extra limbs, garbled text, and no narrative structure whatsoever. None ran longer than thirty seconds. The channels behind them were largely anonymous. They were earning ad revenue until the Times asked YouTube for comment, at which point five channels were suspended from the Partner Program. Reactive, not proactive. The journalism did the moderation.

This is worth sitting with. YouTube's algorithm had the information needed to identify these channels. It had view duration, engagement patterns, channel age, upload frequency, and content fingerprinting. It chose not to act on any of it because the videos were performing.

Here is what a good children's video actually does.

Ellen Doherty, the chief creative officer at Fred Rogers Productions, explained the structure of a Daniel Tiger episode to the Times. Two short stories per episode. Songs that reinforce themes and can be memorized by both child and parent. Pacing calibrated to children who do not yet have cinematic language. Long pauses built in deliberately, because the pause is where learning happens.

"That spark of human connection is everything," she said.

A forty-one-second AI elephant has no pauses. It cannot afford pauses. A pause is a moment in which the child might look away, and the algorithm needs the child not to look away.

This is the distinction that matters. Good children's media is built around what children need. AI slop is built around what the algorithm rewards. These are not compatible goals, and the algorithm does not pretend otherwise.

Spend fifteen years in security operations and you develop a specific relationship with signals. Not all signals carry equal weight. Some signals are loud and fast and mean nothing. Some signals are quiet and slow and mean everything. The ability to distinguish between them is the job.

What YouTube is feeding children is all noise and no signal.

Developmental psychologists call the phenomenon displacement. It is not only that the AI content is bad. It is that the AI content occupies the time and attention that would otherwise go to something that builds something. Reading. Play. Interaction with other humans. Even passive media with real narrative structure. The sheer volume of low-quality content is crowding out the conditions under which development actually happens.

The volume is not accidental. Creators, many operating anonymously, have built reliable income streams on this content. The barrier is low. The upload cadence is mechanical. The feed keeps filling.

Here is what the AI elephant knows about your child: nothing.

It was not designed with your child in mind. It was designed by someone who watched a tutorial about passive income and downloaded a video generation tool. The tutorial did not include a section on object permanence or language acquisition or the developmental importance of narrative repetition. It included a section on SEO and thumbnail optimization.

The elephant's face cycles through four expressions in two seconds because four expressions in two seconds tests better in the first three seconds of watch time than one expression held for the duration of the video. The alphabet is announced in no particular order because order requires structure and structure requires intention, and the creator has no intention beyond the next monetization threshold.

This is not cynicism. This is the business model, stated plainly.

The American Academy of Pediatrics recommends that media use for children under two be very limited. About sixty percent of parents with a child under two report the child watches YouTube. About one-third report the child watches it every day.

YouTube told researchers there is nothing to worry about.

YouTube's algorithm continues to recommend AI elephant content.

McCall Booth, a developmental psychologist at Georgetown University, raised the concern that will outlast this news cycle. Children exposed to improbable but aesthetically realistic content may develop cognitive frameworks that include those impossibilities as baseline. The elephant detaches its trunk. The alphabet is non-sequential. Physics is decorative. These are not lessons anyone intended to teach.

But the algorithm is not in the business of intentions.

There is a reason we talk about screaming at children as a failure of parenting. It is not because screaming fails to get attention. It gets attention immediately and completely. The reason screaming fails is that attention without meaning teaches only that the world is loud.

The AI elephant gets attention. It holds attention. By every metric the algorithm tracks, it succeeds.

The metrics do not track what happens to the child after the video ends.

YouTube suspended five channels after the Times ran its investigation. It removed three hyperrealistic videos from YouTube Kids. It declined to extend the AI disclosure requirement to animated children's content.

Five channels. The investigation identified five channels.

The feed is not five channels deep.

The parents who figured this out first did what parents always do when institutions fail them. They built playlists of vetted content by hand. They removed the app entirely. They made individual decisions about individual videos while the platform that profits from their children's attention issued a statement.

None of this is their fault. They did not build the algorithm. They did not set the monetization policy. They did not decide that disclosure requirements would not apply to cartoons. They opened an app that told them it was safe for children and found out, eventually, that "safe for children" is a promise the platform made to itself.

The algorithm is not screaming at your child because it is malicious. It is screaming because screaming works, and because no one at YouTube is paid to care what happens after the screen goes dark.

That distinction will not matter to the child.

Have you had to manually vet what your kids watch, or found AI content in their feed? Tell me what you found.

The Slopfather keeps the records so you don’t have to swim in the slop alone. If this was useful, share it with someone who needs it.

Subscribe now

The Slopfather covers AI-generated content and the systems that distribute it. If this made you feel something, pass it along.

Merriam-Webster named “slop” its word of the year for 2025.

That is the dictionary telling you something the platforms will not. The content flooding your feed is not a glitch. It is not a phase. It is a system. It has supply chains. It has economics. It has specific techniques designed to make you feel something before your brain catches up to the fact that what you are looking at was generated in four seconds by a server farm in a country you will never visit.

And nobody is giving it an award.

Until now.

My name is Stevan W. Pierce Jr….some people call me The Slopfather!

I been in technology for over 27 years. Fifteen of those years were in cybersecurity with a sociology and counseling degree which has helped me study how people exploit the gap between what something looks like and what it actually is. Phishing campaigns look like your bank. Malware looks like a software update. AI slop looks like a video your cousin filmed at the park.

Same playbook. New tools.

What I kept noticing, as the feeds filled up with impossible hands and suspiciously emotional veterans and fish watching a crucified shrimp in quiet contemplation, was that the people who needed to understand this the most were having zero fun learning about it. The serious tools were built for enterprise. The explainers were written for people who already knew what they were looking for. The coverage was thorough and mostly unreadable.

Nobody was making it funny. Nobody was making it a game. Nobody was handing out trophies.

So I did.

What Is The Sloppies?

The Sloppies is a monthly awards platform for the worst AI-generated content on the internet. Think the Razzies, except the honorees are not bad movies. They are images of Jesus rendered as crustaceans, LinkedIn posts written by a robot impersonating someone’s inspirational aunt, and military jets that cannot decide which aircraft model they want to be for more than eight consecutive seconds.

We run on a 31-day cycle, depending on the month. The first three weeks are open nominations. Anyone can submit. You find slop in your feed, you send it to us, and we evaluate it against our categories. The fourth week is community voting. The final days are the ceremony.

We just finished Month One. February 2026. The first-ever Sloppies awards have been given. https://www.thesloppies.com/winners

History was made. Poorly, and with too many fingers.

The Categories

We have ten of them. Each one targets a specific failure mode.

Finger Crimes is exactly what it sounds like. AI has been counting to five incorrectly since 2022. It has not improved. The nominees in this category represent the full spectrum of manual mathematics: the seven-fingered handshake, the fist with opinions, the hand that gestures confidently in a direction no human hand has pointed.

LinkedIn Slop is a category unto itself because LinkedIn is a biome. Somewhere above 50 percent of long-form posts on the platform are now AI-generated. The nominees here are the ones that achieved perfection: motivational quotes attributed to no one, professional headshots where the background makes architectural decisions mid-render, carousel posts that begin with “I have a confession” and end with a call to follow for more content.

Shrimp Jesus is our legacy category. It is named after the image that started this whole cultural conversation: an AI-generated Jesus Christ rendered with the body of a shrimp, distributed across Facebook with the caption “Type AMEN if you believe.” That image collected hundreds of thousands of interactions. It spawned imitators. It became a genre. The Shrimp Jesus Memorial Award honors the finest work in AI-generated religious engagement bait. There is a lot of competition.

Physics Doesn’t Work That Way is for content that treats the laws of thermodynamics as a polite suggestion. Smoke that grows from “small outlet fire” to “we are all going to die” in five seconds. Missiles that travel in directions unrelated to the aircraft that fired them. Buildings that cast shadows at angles no sun has ever occupied.

Alphabet Soup covers AI text rendering. AI can generate a convincing street sign in any font you want, as long as you do not need the letters on that sign to form words. The nominees in this category are monuments to the gap between looking like language and being language.

We also have Uncanny Valley, Historical Crimes, Too Many Teeth, Fakefluencer Slop, AI Slop Art, and a category for corporate AI content deployed with full boardroom approval and zero self-awareness.

Every month, one piece of content from each category receives a Sloppy. The community votes. The Slopfather presides.

What Is Coming

Month Two is already underway. March nominations are open now. The bar has been raised by the February winners, and the February winners set a high bar.

Later this spring, the first Sloppies Trading Cards arrive. The award winners become collectible cards. The Shrimp Jesus gets a card. The seven-fingered handshake gets a card. The LinkedIn motivational post gets a card. I am genuinely enthusiastic about this in a way that my security colleagues find concerning.

Beyond that: more monthly cycles, more categories as new slop genres emerge, and a community of people who have learned to slow down and look twice before hitting share. The educational content and the entertainment are not separate tracks. They are the same track.

Digital literacy is the mission. The trophies are the Trojan horse.

Why This Matters

Forty-one percent of Facebook posts are now estimated to be AI-generated. One in three videos in a YouTube recommendation feed is synthetic. The 2026 Winter Olympics opening ceremony used AI-generated visuals and received a response from viewers that I will diplomatically describe as a verdict.

The people most targeted by this content are not foolish. They are operating without a framework. Once you have the vocabulary, once you know what Shrimp Jesus is and why it works and what the algorithm is doing when it serves it to you, you see it everywhere. You are not smarter. You are just informed.

That is the whole point.

The Sloppies is not here to shame anyone. It is here to make the learning unavoidable by making it impossible to look away. Humor is the delivery mechanism. The education is the payload.

The internet has a slop problem.

Now it has a ceremony.

You’ve seen slop today. You know you have. What category does it belong in - and have you nominated it yet?

The Slopfather keeps the records so you don’t have to swim in the slop alone. If this was useful, share it with someone who needs it.

Subscribe now

Nominations for the March 2026 Sloppies are open now at thesloppies.com. If you found something spectacular in your feed, we want it. The Slopfather is watching. The Slopfather is always watching.

Next issue: a breakdown of the February winners, the specific tells that got each one nominated, and what each category reveals about where AI generation is still failing. Subscribe so you do not miss it.