AI Fatigue

AI Fatigue Browser Extension

A phone playing a night street video. AI FAKE
A laptop playing a video of a person speaking. AI FAKE
A tablet playing a video of a fire. AI FAKE

Every video and photo is scored only once.

Built for Chrome and Brave. It reads a video or photo already on screen on X, Facebook, Instagram, TikTok, YouTube, and Telegram. Coming in the future for Edge and Firefox.

Install

Get it in Chrome

Zero-Retention Guarantee. Your feed is never uploaded. One frame or one photo is scored. The picture is thrown away. Only the number stays.

  1. Open Chrome on a computer.
  2. Open the Chrome Web Store and search for AI Fatigue.
  3. Click Add to Chrome, then confirm.
  4. Open X, Facebook, Instagram, TikTok, YouTube, or Telegram. A video or photo that already has a score shows that score.

Brave uses that same Chrome Web Store listing. The button there says Add to Brave. There is no Brave store to submit to.

The listing is not public yet, so this page has no install link. When Chrome approves it, the link goes here.

A browser extension from AI Fatigue Productions, invented by Mike Franco. It scores a video or photo once. The next person who meets that post sees the score.

An older woman at her kitchen table, reading a tablet.

01

What AI Fatigue is

Today's digital warriors use our web extension to help weed out deepfakes before they can go viral. That is why a video or photo is scored only once.

AI Fatigue is a deepfake check for the feed. The first person to check a clip or photo spends one score. The extension sends one frame of the video, or the photo itself. The score comes back as a number for how generated that frame looks. The picture is thrown away. The number is kept.

The next person who comes across that same clip does not pay for another check. Their extension recognizes the video and shows the score already on file. Watching the feed does not upload the video.

A score is not a verdict that the video is a lie. A high number means the frame looks generated. A comedy sketch that is obviously AI generated is not the target of this tool, unless the user wants to block those types of videos as well. The default is to allow AI-created humor. The full cover, a blur and the words “Flagged deceptive video,” is only for a clip that 5 different paid members have confirmed as an unlabeled fake, not for every generated frame. “View anyway” still plays it. Nothing is deleted, and the account that posted it is not reported.

The extension does not ask for a Facebook, X, Instagram, TikTok, YouTube, or Telegram password. It only sees a video or photo already on the screen. It runs on X, Facebook, Instagram, TikTok, YouTube, and Telegram. No ads.

02

Why it is needed

The platforms are not labeling these videos and photos, so someone else has to.

A generated clip can sit in the same feed as a video a person actually filmed. The unlabeled ones are the problem. A sketch that tells you it is a sketch does not need a cover.

These are examples, not reports of a particular day.

  1. A phone video of a fire downtown that never happened, forwarded with the words “send this to your family before it is taken down.”
  2. A grandparent who sees a child, or a public figure, saying words they never said.
  3. A familiar face asking for money in a generated clip.
  4. A news-style clip with no station and no date.

None of those clips has to be clever. They only have to look ordinary long enough to be forwarded. Once one person has checked the clip, the score should be waiting for the next person. That is the product.

This one is a real clip. The post says a mountain sank into the ground. The frame at 0:06 scored 99. The label reads AI fake detected at 0:06 = 99. The score stays on the clip after that moment.

A landscape clip labeled AI fake detected at 0:06 = 99.

The mountain clip on X. The frame at 0:06 scored 99.

An older couple looking at a laptop together.

03

Score the part that looks fake

A mixed clip is real footage with a generated piece cut in. Do not check a random second. Play until the part that looks wrong is on the screen, and score that frame.

When a video is tagged, the tag does not say the video is fake. It names the moment. The words are AI fake detected at 1:29 = 99. The real minutes are not tagged.

A mountain goat clip on X is the example. The person mixed real parts with a generated part. The frame at 1:29 scored 99, so that is the line on the clip. The rest of the clip was not scored.

The clip stays on X. Anyone can watch it there. That does not make the file public domain, so this site does not keep a copy. The post itself is the example. Go to 1:29.

04

Best time to tag — for digital warriors

The detection looks only at the pixels of the single frame that is on your screen right now. It does not watch the whole video. That is why timing matters.

Play the clip until the moment that looks wrong is fully visible, then score it. A face that suddenly looks too smooth. A background that does not match the lighting. Hands that look off. A scene that feels generated. That is the frame to check.

Do not score the first second just because the video started. Do not score a blank or dark frame. Do not score ordinary color grading, filters, or speed changes — those are not treated as AI-generated. Score the part you already have reason to doubt.

When the stakes are highest

Evening hours (6:00 PM – 10:00 PM local time). Traffic peaks when people are tired and more likely to share emotionally. Shock clips, fake local emergencies, and emotional deepfakes are often pushed then. Check viral clips in your evening feed before you share or forward them.

Right after major breaking news. Within minutes of real events (weather, politics, public safety), synthetic clips appear — AI B-roll, old footage called live, or fabricated audio. If a breaking clip has no clear timestamp or real news desk, check it before you let it shape what you believe or send.

Before you forward to family. Older relatives and close family chats are common targets because trust is high and verification is low. Use AI Fatigue as the last step before you hit send on any dramatic voice note, urgent request, or sensational public-figure clip.

Once that frame is scored, the number (even a zero) stays on the clip for everyone else. No one else has to spend a scan on the same video. That is how the community saves tokens.

05

What the number means

The number is a 0–100 score on one frame, not on the whole video. For a photo, it is a score on the photo itself.

When you score, the extension finds the largest video or photo on the screen. For a video, it reads the exact frame that is visible right now. For a photo, it sends the full-size photo, not a screenshot. That single frame is sent to an AI detection algorithm that checks three things:

  • How AI-generated the frame looks
  • How likely a face has been swapped
  • Whether it looks like a photo of a screen

The main number you see is the AI-generated score, normalized to 0–100.

  • 0–49 means the frame does not strongly look AI-generated
  • 50–100 means the frame looks AI-generated

When the score is 50 or higher and the playback time is known, the label becomes AI fake detected at 1:29 = 99. A zero still shows the time, so everyone knows that moment was already checked. Otherwise it shows the number and the time.

Photos

Photos work the same way, with one difference: there is no moment to name, so a photo label reads AI fake detected = 92 with no time. The whole photo is checked, at full size. Profile pictures, emoji, and other small images are skipped. A photo in a post with a video gets its own score, so the two never mix.

Only one frame is checked. The rest of the video is never scored. A high number means that frame looks generated. It is not a verdict that the entire video is a lie or that the person who posted it is deceptive. The picture is thrown away after the check. Only the number, the time, and a local fingerprint of the video are kept.

The full cover that says “Flagged deceptive video” only appears after 5 different paid members confirm the clip is fake. A high score alone does not cover it. How a clip gets covered.

06

Who it hurts

The people most exposed are older viewers, and anyone who still takes a video at face value. They are not failing a test. For a long time, a moving picture in a family chat was a thing that happened. The feed no longer keeps that promise, and it does not say so on the clip.

What they lose is ordinary. The pleasure of scrolling without having to interrogate every face. The ease of forwarding something that looks urgent. When the clip lands in a family chat, someone else has to explain that the fire, the voice, or the request for money was never a record.

AI Fatigue does not stop a person from forwarding. It puts the score on the clip before the next person has to decide whether to believe it.

An older man on a sofa, watching a video on his phone.

07

What you see for free

A video covered by a blur, with the words Flagged deceptive video and a View anyway button.

The extension is free to install. There are no ads. If a video or photo was already scored, you see that score. You see when it was found and how many members tagged it. If 5 paid members confirmed it as an unlabeled fake, you see this cover, and you can still play it.

You do not spend a score by scrolling. Nothing is uploaded just because it passed through your feed. The extension asks only whether that video or photo is already on file.

A clip nobody has checked yet has no score. Free use does not run a new check or tag anything. A paid plan does.

08

How a clip gets covered, and why one person cannot game it

Everything here works the same for photos as for videos. A photo just has no timestamp.

Nobody can install the extension and start covering videos or photos for everyone else. It takes paid members, and it takes agreement.

Free

You see every score already on file, the moment in the video it was found, and how many members tagged the clip. You cannot scan and you cannot tag. Nothing a free install does changes what anyone else sees.

A score alone never covers a video

A paid member scans one video frame or one photo. The number and the time are shared, so the next person sees “AI fake detected at 1:29 = 99.” That is information, not a verdict. The video still plays.

Fake: 5 paid members

A clip is covered as a fake only after 5 different paid members have each scanned it themselves, each got 50 or higher, and each tagged it fake. Until then you see the score, the time, and a count like “2 of 5 say fake.”

  • Someone else’s high score does not let you tag. Your own scan has to be 50 or higher.
  • One member counts once, no matter how many times they tag.
  • The scans can be from the same moment or different ones. A clip that is real footage with one generated part only needs that part.
  • A member can take their tag back.

AI humor: 10 paid members

Some AI clips are obviously AI and worth watching, like a clever animated parody. Paid members can tag a clip “AI humor.” After 10 different members agree, it gets that label. AI humor plays for everyone by default. Members who do not want to see it can turn on “Block AI humor.” If more members call a clip fake than humor, it is treated as fake.

Why it works this way

One person with a grudge cannot cover a real video, and one person cannot wave a fake through. Every scan and every tag is logged, so a bad actor’s tags can be found and removed.

09

Plans

A video labeled AI fake detected at 1:42 = 99.
Free Paid
See scores already on file Yes Yes
Cover on clips 5 members confirmed fake Yes Yes
Run a new scan on an unchecked video or photo No Yes (uses one scan)
Tag clips fake or AI humor No Yes
Block AI humor too No Yes
$0
See the AI % and the moment it was found on every clip the community scanned. Clips 5 paid members confirmed fake are covered. Tap “View anyway” to watch, going in knowing it is probably fake.
$10 a year
A year of membership for the price of ten months. Block AI humor along with the fakes, and get 20 scans a month to scan and tag videos and photos yourself.
$1 a month
20 scans a month to scan and tag videos and photos yourself. Block AI humor too.
$5 a month
150 scans a month to scan and tag videos and photos yourself. Block AI humor too.
$10 a month
400 scans a month to scan and tag videos and photos yourself. Block AI humor too.
$20 a month
1,000 scans a month to scan and tag videos and photos yourself. Block AI humor too.

A scan is one frame of one video, or one photo. The number stays on the post for everyone else.

Paid plans open when the extension is in the Chrome Web Store. Nothing is for sale yet, so nobody pays for something they cannot use.

Join the waitlist

Send one email and you are on the list. When the extension is approved, you get one email back with the install link and the plans.

10

The community

A crowd standing together under a banner that reads AI-Fake Digital Warriors.

The free install shows a score the community already paid to check. A paid member is the person who checks a new clip. One frame is scored, the picture is not kept, and the number waits for the next person.

That is the open part. Anyone can see a score that already exists. The code is not published, and the site does not ask you to read it.

See today’s Hall of Fame — November 17, 2026

11

FAQ

Does it check every video and photo as I scroll?
No. A member spends one check on one frame or one photo. Anything nobody has checked has no score.
What happens when I hit an unchecked video or photo?
Free users see no score. Paid users can spend one scan on one frame or photo. The number then appears for everyone else.
Does it work on photos?
Yes. Scroll the photo into view and score it. The whole photo is checked, and the label has no timestamp. The same rules apply: 5 paid members have to confirm it before a photo is covered.
Is a high number proof the video is a lie?
No. It means that one frame looks generated. The rest of the video is not cleared or condemned.
Where do I run the check?
On the moment that looks fake. If that frame scores high, the tag names the time. A mountain goat clip on X is one. The frame at 1:29 scored 99, so the line is AI fake detected at 1:29 = 99. The rest was not scored.
Where does it run?
Chrome and Brave, on X, Facebook, Instagram, TikTok, YouTube, and Telegram. Coming in the future for Edge and Firefox.

Why the platforms are not labeling these videos

How to look at a clip on X before you believe it

12

Contact

Support questions go to the support page and to support@aifatigue.us.

Other mail goes to info@aifatigue.us.

AI Fatigue is a product of AI Fatigue Productions, invented by Mike Franco. The site is https://aifatigue.us. The Chrome Web Store support address is https://aifatigue.us/support.