so this one landed yesterday and it's kind of a big deal. google just opened up its synthid detector to everyone — a free website at synthid.com where you upload a photo, video or audio file and it tells you whether it was made with ai. no special access, no media credentials, just you and the file.

until now this thing was locked behind a waitlist of journalists, media people and researchers — google handed out test access at google i/o last year and basically kept it in that circle. techcrunch reported the public launch on october 7 2026, and it's live worldwide in english. if you've been wondering how to tell whether some viral photo is real or ai, this is google's official answer.

here's the thing though. it doesn't work the way a lot of people think it does. and the limits matter more than the feature itself. so let's go through what actually launched, how to use it, what it can and can't catch — and then the fun part: what to do about the ai photos that are all over your feed anyway, because they're not going anywhere.

What just launched

synthid is google's watermarking system for ai-generated content. it's not a visible watermark — nothing stamped on the image. it's an invisible signal embedded in the pixels (or the audio waveform) when the content gets generated. the detector at synthid.com reads that signal and reports back.

the scale here is honestly wild. since synthid launched in 2023, google says 180 billion-plus images and videos and 240,000 years of audio have been watermarked with it. and verification is already happening at massive scale — roughly a million verification requests a day are going through google search, the gemini app and chrome, where these checks are built in. the website just makes the same check available to everyone directly.

so what does it actually catch? anything carrying a synthid watermark. that includes content from google's own models — nano banana, veo, lyria — and google tools with generation features like gemini, flow, producerai and vids. plus partners: openai, nvidia and kakao have signed on too. apple is expected to add support soon, according to reports from techcrunch, sqmagazine.co.uk, ubergizmo and ndtvprofit.

How to check a photo yourself

it's genuinely simple. takes about a minute.

step 1 — go to synthid.com

open it in your browser. you'll need to sign in (google account), which is the one mildly annoying part — it's mentioned in coverage from macrumors as a requirement. not a huge deal, but heads up.

step 2 — upload the file

drag and drop or browse to the photo you're suspicious about. it supports basically every format you'd realistically have: images in jpg/jpeg/png/bmp/webp/avif/heic/heif/tiff/tif/gif, video in mp4/mov/webm, and audio in wav/mp3/ogg/flac/aac/m4a. so that weird photo your cousin forwarded, the mp4 your group chat won't stop sharing — it takes all of those.

step 3 — read the result

the site tells you whether it found a synthid watermark. if it did, it can tell you the content came from one of the supported generators. if it didn't find one, it doesn't mean the file is definitely human-made — and that brings us to the part everyone should read carefully.

What it can and can't tell you (the honest limits)

ok this is the section that matters, because the headlines are making this sound like a magic ai detector and it is not that. gizmodo and macrumors both made the point clearly: synthid detector is NOT a general ai detector.

here's what that means in practice:

it only catches watermarked media. if a photo or video was generated with synthid watermarking on — google's tools, openai, nvidia, kakao — the detector can flag it. but there are a lot of ai image generators out there that don't use synthid, and plenty of ai photos in the wild have no watermark at all. a clean result doesn't prove anything.

microsoft and meta are doing their own thing. they run their own watermarking standards, so content from their tools isn't necessarily in synthid's world either. the detector is a piece of the puzzle, not the whole puzzle.

watermarks can be damaged. heavy editing, aggressive recompression, cropping and re-exporting — anything that messes with the file enough — can weaken or strip the invisible signal. a screenshot of a screenshot of an ai photo might not carry enough signal to be detected.

so how should you actually use it? think of it like this: if it says "yes, this has a synthid watermark," that's a solid answer — the file came from an ai generator. if it says "no watermark found," that tells you almost nothing on its own. combine it with your eyes (weird hands, melted backgrounds, too-perfect skin — the usual tells), check the source, and don't treat a negative result as proof of real.

Why this matters right now

because your feed is drowning in ai photos and half of them look completely real. the trends are coming faster than people can keep up with, and the results are getting harder and harder to eyeball.

just in the last few weeks on this blog we've covered three of them: the hug my younger self trend — those emotional photos of people posed next to their childhood selves — the pumpkin-head halloween trend that's about to explode as halloween gets closer, and the viral '80s ai yearbook trend that turned everyone's feed into a 1986 high school photo album. all three look shockingly real at a glance. all three fooled a lot of people. now there's finally a free tool where you can upload one and check.

and it's not just fun trends. the same tech makes fake news photos, scam images and doctored screenshots. your family group chat is probably full of forwards that would make this tool earn its keep. having synthid.com in your pocket — on your phone browser, free — means you can settle the "is this real" debate in about thirty seconds instead of arguing about it for a day.

After you check: fixing ai photos free on your phone

here's the part nobody's talking about yet. checking whether a photo is ai is step one. step two — and this is where it gets fun — is that the ai photos people make are usually broken in small ways, and fixing them is what makes a trend post actually go viral instead of looking janky.

think about it. you generate a photo with one of these tools. it comes back looking amazing except the hands are melted, or the face is a little warped, or the background turned into a smear of watercolor nonsense. that's the standard ai photo experience. the people whose versions blow up are the ones who clean that up before posting.

that's exactly what picfix pro is for. it's a free ai photo editor for android and it handles the classic ai-generated glitches — warped faces, messed-up hands, smeared backgrounds, off lighting. ai face touch-up, background deblur and sharpening, one-tap object removal for stuff the generator hallucinated into the frame, lighting and color balance so the whole thing looks like one real photo.

so the workflow is kind of perfect now. see a viral ai photo trend, check suspicious ones with synthid.com, make your own version, then fix the glitches free on your phone before you post. picfix pro is free on google play — grab it at https://play.google.com/store/apps/details?id=com.dhanuk.photodoctorpro and the cleanup step takes about two minutes.

and if you're thinking "but wait, if the detector can't catch everything, won't ai photos keep flooding everything?" — yeah, probably. that's the point. the volume isn't going down. the winners in this era are the people who can make ai photos that actually look good, not the people who can spot the bad ones. a free photo editor on your phone is the cheapest possible way to be in the first group.

The short version

google opened synthid.com to everyone on october 7 2026. upload a photo, video or audio file and it checks for an invisible synthid watermark — it catches content from google's models (nano banana, veo, lyria), google tools (gemini, flow, producerai, vids) and partners openai, nvidia and kakao, with apple expected to follow. it's free, it's fast, and it's not a general ai detector — a negative result doesn't mean a photo is real. microsoft and meta use their own watermarking standards.

the real play here: detection is only half the story. everyone is making ai photos now — the pumpkin-head halloween trend, the '80s yearbook trend, the hug-my-younger-self trend, whatever blows up next week. the ones that go viral are the clean ones. so check the suspicious ones with synthid detector, then fix your own with picfix pro — free on google play. that's the whole game right now.

Related reading

more ai photo trend guides from this blog:

Related reading