Check Image AI or Not: A Step-by-Step Guide to Verifying Any Photo

A friend forwards you a striking photo in a group chat. "Is this real?" they ask. You stare at it for ten seconds and shrug. It could be. It might not be. You genuinely can't tell.

That moment of uncertainty is the whole problem. In 2026, AI-generated images have reached a level of realism where unaided human judgment fails roughly 40% of the time. The good news: you don't have to rely on your gut alone. There's a reliable, repeatable process you can follow to check whether an image was generated by AI or not — and it takes under five minutes.

This guide walks through that process step by step.

Step 1: Get the Best Version of the Image You Can

Before you do anything else, make sure you're working with the highest-quality version available. This matters more than people realize.

Social media platforms compress images aggressively. Each round of JPEG compression strips away high-frequency pixel data — and that's exactly what AI detectors rely on. An AI image that's been screenshotted from Instagram, re-uploaded, and screenshotted again is far harder to detect than the original file.

What to do:

• If the image came from social media, try to find the original post or the account that first shared it
• Download the full-resolution version rather than taking a screenshot
• Avoid forwarding through messaging apps (they re-compress) before checking
• If only a screenshot is available, note that detection accuracy will be lower

Step 2: Run an Automated AI or Not Check

This is your first hard signal. Upload the image to AIorNot — it's free, requires no account, and gives you a confidence score in under three seconds.

What you'll get:

• A confidence score from 0-100% indicating the likelihood the image was AI-generated
• A breakdown of which detection signals triggered (texture anomalies, edge artifacts, frequency distribution mismatch)

How to read the score:

• 85-100% AI: Strong evidence. The image is very likely AI-generated.
• 60-84% AI: Moderate evidence. Worth cross-checking.
• 40-59% AI: Inconclusive. Use additional verification methods.
• Below 40% AI: Likely a real photograph.

Don't stop here unless the stakes are low. A single automated check is a strong signal, but it's one signal. The complete guide to detecting AI-generated images covers why multi-layer verification outperforms any single method.

Step 3: Do a Visual Inspection (2 Minutes)

While the automated tool works on pixel-level statistics, your eyes catch different things — semantic inconsistencies that algorithms don't always flag. Here's what to look for:

Hands and Fingers

Still the most reliable quick check. AI models produce extra fingers, merged digits, or fingers that bend at impossible angles. Not every AI image has bad hands, but when you see it, it's a dead giveaway.

Text in the Image

Look at signs, book covers, screens, clothing labels, and license plates. AI generators struggle with text — it often looks right at a glance but dissolves into nonsense characters when you actually read it.

Lighting and Shadows

Do the shadows match the light source? If the subject is lit from the left but their shadow falls to the left too, something's wrong. AI models sometimes get the physics of light subtly incorrect.

Background Details

While the subject might look perfect, the background often reveals the truth. Look for:

• Objects that merge into each other
• Perspective lines that don't converge
• Repeating patterns where there shouldn't be any
• Reflections that don't match the environment

The 7 practical checks guide covers each visual indicator in detail with specific examples.

Step 4: Check the Image Metadata

Real photographs typically carry EXIF metadata — the camera model, settings, timestamp, GPS location, and sometimes the software used for editing. AI-generated images usually lack this metadata or have fabricated entries.

What to look for:

• Camera information: Real photos have camera model, lens, aperture, ISO, shutter speed. AI images often have none of this.
• Software tags: If the image was processed through Midjourney, DALL·E, or Stable Diffusion, the software tag may reveal this.
• Creation date: Does the timestamp make sense given where you found the image?
• C2PA manifest: Newer cameras from Sony, Leica, and Nikon embed cryptographic provenance data. AI generators like OpenAI (as of May 2026) embed C2PA manifests labeling content as AI-generated.

You can check metadata using the Image Content Analyzer, which extracts and displays embedded information alongside content analysis.

Caveat: Metadata can be stripped or faked. Its absence doesn't prove an image is AI-generated — many real photos have their metadata removed during upload to social media. But its presence (especially C2PA provenance) is strong evidence of authenticity.

Step 5: Do a Reverse Image Search

If the automated check and visual inspection are inconclusive, a reverse image search can provide context. Upload the image to Google Images, TinEye, or Bing Visual Search.

What you're looking for:

• Does the image appear elsewhere? If you find it on a stock photo site or a photographer's portfolio, it's likely real. If it only appears on AI art galleries or social media, that's suspicious.
• Is it part of a series? AI-generated images are often part of a batch — similar images with slight variations may appear.
• When was it first posted? The original posting date and context can be revealing.

Step 6: Consider the Context

Step back and think about the image in context:

• Who shared it? A verified news organization, a professional photographer, or an anonymous social media account?
• Does the content make sense? Would the person depicted plausibly do or say what's shown? Does the scene match what you know about the location or event?
• Is it too convenient? AI-generated images are often created to serve a narrative. If an image perfectly supports a controversial claim and appears from an unverified source, skepticism is warranted.

Our 3-second quick guide covers the fastest context checks you can do at a glance.

Step 7: Make a Decision

Combine all your signals:

If 3+ signals point the same direction, you can be reasonably confident in your conclusion.

If signals are mixed, seek additional verification. Contact the source directly, check with fact-checking organizations, or use multiple detection tools. Our technical guide to how AI detection works explains why mixed signals occur and how to weigh them.

Common Mistakes When You Check If an Image Was Generated by AI or Not

Relying on a single method. Any one approach — automated, visual, metadata, reverse search — has blind spots. Stacking methods gives you a far more reliable answer.

Checking a compressed screenshot. If you only have a low-quality screenshot, your detection accuracy drops significantly. Always seek the best version available.

Over-trusting visual perfection. AI images can look flawless. The absence of obvious flaws doesn't mean an image is real. Conversely, real photos can have artifacts from editing, compression, or poor photography.

Forgetting that context matters. A technically flawless detection result doesn't replace common sense about where an image came from and whether its content is plausible.

Frequently Asked Questions