scam safety
How to Tell if an Ad Is AI Generated: Labels, Tells, and 6 Checks
Google and Meta now label some AI-made ads, but scam ads usually carry no label. Here is where to look, what the FBI says to watch for, and 6 checks.
By DealsDMV Team · · 6 min read
There are two quick ways to tell if an ad is AI generated, and neither is perfect. First, open the ad's three-dot menu: Google and Meta both now show an AI note there for some ads. Second, look for the visual slips the FBI lists, like warped hands, odd teeth or eyes, and shadows that fall the wrong way. But here is the part most guides skip: a missing label proves nothing. Both companies say their labels mostly cover ads made with their own AI tools, or ads whose advertiser chose to disclose it. A scammer using an outside tool simply doesn't tick the box.
So the more useful question is not "was this made by AI?" It is "is what this ad tells me true?" That is a question you can actually answer, and the checks below show how.
An AI label tells you how an ad was made. Only a check tells you whether the deal is real.
Why does it matter now?
Because shopping ads on social media are where a lot of money disappears. In its April 2026 Data Spotlight, the Federal Trade Commission reported that people lost $2.1 billion in 2025 to scams that started on social media, about eight times the 2020 figure. Nearly 30% of people who reported losing money to any scam in 2025 said it began on social media.
Shopping scams were the most reported kind. More than 40% of people who lost money to a social media scam said it started when they ordered something they had seen in an ad. Most said the item never arrived; others got counterfeits or something very different from the picture. The FTC adds that most scams are never reported, so the true losses are likely higher.
AI makes these ads cheaper and more convincing. The FBI warned in December 2024 that criminals use AI image tools to show celebrities or social media personalities promoting counterfeit products or goods that are never delivered, and that AI helps them clean up the spelling and grammar mistakes that used to give scams away. With holiday ads ramping up in the weeks before Black Friday, it is a good time to sharpen your eye.
Where do Google and Meta show the AI label?
Here is what each company's own help page says, in plain terms.
| Platform | Where to look | What it covers | What it misses |
|---|---|---|---|
| Google (Search, YouTube, and others) | Tap the three-dot menu on the ad, then "How this ad was made" in My Ad Center | Ads with AI-made or AI-edited assets that the advertiser labelled, or that Google labelled automatically for some ads built with its own tools | Ads where the advertiser did not disclose AI use; Google says not all AI-created content needs a label |
| Meta (Facebook, Instagram) | The three-dot menu at the top right of the ad; or an "AI info" label beside "Sponsored" if the image shows a realistic AI person | Ad images fully made or materially edited with Meta's own AI creative features | Ads made with outside AI tools, which Meta says will not get AI info; the feature is still being rolled out |
Two things stand out. Google says disclosure rules differ by place, naming the European Union, India and New York as examples of places that require it, so you may see labels on some ads and not others. And Meta states plainly that ads made with third-party AI tools won't carry its label. A scam ad built elsewhere and uploaded as a finished video can look exactly like an unlabelled, honest ad.
That makes the label useful in one direction only. If you see it, you know AI was involved. If you don't, keep checking.
What visual tells still help?
The FBI's list of things to look for in images and video is a good starting point:
- Hands or feet that look distorted
- Teeth or eyes that look unreal
- A face that is blurry, smeared or irregular
- Glasses or jewelry that don't sit right
- Shadows that point the wrong way
- Odd watermarks
- Lag, a voice that doesn't quite match the mouth, or movement that looks unnatural
Use these as a reason to slow down, not as a pass or fail test. Tools improve every year, and a well-made fake can show none of these slips. On the other hand, a real ad can look polished and still be honest. Many genuine shops use AI for backgrounds or cleanup, and the FTC's own guidance says its rule on reviews and testimonials has no blanket ban on AI avatars in marketing.
What the FTC does draw a line around is a fake endorsement. In its answers on the rule, it says using a celebrity avatar without that person's permission to praise a product breaks the rule if shoppers would reasonably believe the celebrity really gave the testimonial. In other words, AI itself isn't the problem. Lying with it is.
The 6 checks before you buy from an ad
These come from the FTC's and FBI's advice, put in the order that saves you the most time.
- Ask who is speaking. If a famous face is selling something, search their name with the product and words like "scam" or "fake," as the FTC suggests. A real partnership usually shows up on the person's own verified accounts too.
- Leave the ad and go to the source. Type the brand's web address yourself instead of tapping the ad's link. If the "70% off" deal isn't on the brand's own site, treat the ad as fake. The FTC's 2026 report found many scam ads led to unfamiliar sites, or to sites pretending to be well-known brands.
- Check the shop, not the video. Search the store's name plus "scam" or "complaint." Our 7 free checks for a shopping website walk through the rest.
- Spot the countdown. The FTC warns that "limited time" pressure is there to stop you from doing research. A real sale is still there in ten minutes.
- Open the three-dot menu. Read "How this ad was made" on Google or "AI info" on Meta. A label is a clue; no label is not a clean bill of health.
- Pay the safe way. The FBI says never to send money, gift cards or cryptocurrency to people you know only online. If a seller will only take those, walk away.
If the ad is selling a health product, the FTC adds one more step: ask your doctor. Supplements aren't reviewed for safety or effectiveness by the government before they go on sale, and a celebrity's face doesn't change that.
What should you do if you already bought?
Act quickly, and keep everything: screenshots of the ad, the site's address, your order email and your payment record.
- Report the ad on the platform. Google's My Ad Center has a "Report this ad" option in the same three-dot menu.
- Report it to the FTC at ReportFraud.ftc.gov. The FTC uses these reports to investigate and to warn others.
- If money was lost to fraud, the FBI asks for a report at ic3.gov, with the date, payment type, amount and how you were contacted.
- Call your card company or bank about the charge as soon as you spot the problem.
If an item simply never shows up, our guide to the FTC's 30-day shipping rule explains what sellers owe you.
The bottom line
You can't always tell if an ad is AI generated, and that's fine, because it was never the real test. The real test is whether the person in the ad endorsed it, whether the shop exists, and whether the deal is on the brand's own site. Labels from Google and Meta help when they appear; their absence means nothing. Slow down, leave the ad, and check the source.
At DealsDMV, every listing lives on our own site, so you never have to buy from an ad's link. Browse Buy Now or the live auctions directly, and see the FAQ for how payment and pickup work.
Know someone who shops from social media ads? Send them this before the holiday rush.
Sources
- FTC Data Spotlight: Reported losses to scams on social media eight times higher than in 2020 (April 2026) (ftc.gov)
- FTC Consumer Alert: Did a celebrity really endorse THAT? Maybe not (April 2024) (consumer.ftc.gov)
- FTC: The Consumer Reviews and Testimonials Rule, Questions and Answers (November 2024) (ftc.gov)
- FBI IC3 Public Service Announcement: Criminals Use Generative AI to Facilitate Financial Fraud (December 2024) (ic3.gov)
- Google My Ad Center Help: AI Transparency in ads (support.google.com)
- Meta Help Centre: How AI-generated images in ads are identified and labelled on Meta (meta.com)
More to read
electronics
Are Refurbished iPads Worth It? Check the Model Year First
A refurbished iPad can be a smart gift, but this fall's update cut off five older models. Here is how to check the model and the seller before you pay.
By DealsDMV Team ·
scam safety
How to Tell if a Pokémon Card Is Fake: 7 Checks, No Ripping
Pokémon's own support team names a light test, the price and flea markets. Seven checks for cards and packs, none of which harm a real card.
By DealsDMV Team ·
scam safety
How to Tell if LEGO Is Fake: 7 Checks, Box to Brick
Seven checks that tell real LEGO from a clone or a counterfeit, from the listing and the box to the studs, the minifigures and LEGO's own product lookup.
By DealsDMV Team ·