Reverse image search + AI check

The two questions that verify almost any image are "where did this come from?" and "how was it made?" Reverse image search answers the first, an AI detector answers the second, and together they're far stronger than either alone. This guide shows how to run both and, more importantly, how to read their answers together.

Why two steps? Because the two most common ways an image misleads people are completely different problems. One is a real photo used dishonestly, an old picture passed off as breaking news. The other is a synthetic image passed off as a real photo. A detector is blind to the first; a reverse search is blind to the second. Run both and you cover both failure modes.

Step 1: find the origin

Run the image through Google Lens, TinEye, or Bing Visual Search. Uploading the file directly works better than pasting a link, since it searches the actual pixels rather than whatever page currently hosts them. You're looking for the earliest and most authoritative appearance: a news agency, an original social post, a stock photo listing, an official source. Where the tool allows it, sort results by date so the oldest instance rises to the top. Each engine indexes the web differently, so if one comes back empty, try another before concluding the image is nowhere.

What the origin tells you

An old photo attached to a new event is the most common form of image misinformation, and it requires no AI at all. If the "breaking news" picture first appeared years ago, the case is closed no matter how real it looks, because it is real, just not of what's claimed. If the image appears nowhere despite looking like a professional photograph, that absence is itself suspicious: genuine photos of genuine events tend to leave trails across news sites and social media. And if it traces to a stock library or an unrelated real person, you've likely found a stolen or repurposed image.

Step 2: check how it was made

When the search is inconclusive, no clear origin, no obvious reuse, move to the detector. Upload the image and use the crop tool to focus on the region that matters: a face, a hand, the specific detail being claimed as evidence. A focused crop gives a cleaner score than a full screenshot with borders and interface around it, both because it removes irrelevant pixels and because it preserves resolution where you need it. This is the step that catches fully synthetic images that never had a real origin to find.

Reading the two results together

The power is in the combination. No trail plus a high AI score means almost certainly generated, a fake with no past because it was never photographed. A clear older origin plus a low AI score means a real photo being recycled dishonestly, the misinformation is in the caption, not the pixels. A real origin but a high score on one cropped region can indicate a genuine photo that's been AI-edited in that spot, a subtler and increasingly common case. And when both checks come back ambiguous, that is a result too: treat the image as unverified and don't pass it on as fact.

A worked example

Say a dramatic "photo from today's protest" is circulating. Step one: reverse search it. If Lens surfaces the same image on a news site from three years ago, you're done, it's real but misdated, and no AI check is needed. If instead it appears nowhere, step two: run it through the detector, cropping to a face in the crowd and a patch of background. A high score on both, plus melting text on a protest sign you can see with your own eyes, and you can confidently call it generated. Same two tools, different paths depending on what step one returns.

Building the habit

The whole process takes a minute or two once it's routine, and it's worth making routine before you share anything emotionally charged, that's exactly the category that's most often manipulated and most damaging to spread. Find the origin, then check the construction, then read the two answers together. Neither tool is perfect alone, but the pair covers the gaps in each, and the discipline of checking before sharing is the single most useful habit for not being fooled.