Top 6 Methods to Identify AI-Generated Images

AI Detection
Surfer AI - Best All-in-one Assistant

- Your article will be ready in less than 20 minutes and it will be 10 times cheaper than using a dedicated writer.
- Create ready-to-rank articles in minutes with Surfer AI.
- Research, write, and optimize across industries with the click of a button.

If I gave you two photos to pick from, how specific are you that you can recognize which is AI? It is fairly difficult. I even failed our very own test (but we will not speak about that).

But which is the issue with pictures nowadays.

With the release of DALL-E 3 and Midjourney V6 possibly all around the corner, several of the common AI picture generators are getting to be smarter and much more effective. Technological innovation has type of caught up with people and reached a stage the place it can generate pictures that make you query who manufactured it.

With troubles this kind of as deepfakes getting to be more and more rampant, how can you defend by yourself from falling victim to AI pictures?

Do not fret. I received your back. Right here are 6 effortless and some of the most correct strategies of spotting AI-produced pictures.

How Can You Inform If It is AI?

Appear For Identifying Info

The quickest and simplest way to spot AI-produced pictures is by seeking for watermarks. You can normally discover this in the bottom side of the picture. This may well be a dead giveaway even though.


This does not perform all the time. For a single, Midjourney and the new DALL-E three model will not output pictures with watermarks. Consumers can also get rid of these watermarks by receiving a paid subscription.

What you can do rather is to download the picture. There are occasions when uploaders couldn’t be bothered to adjust an image’s authentic file title. The issue is, common AI picture generators frequently have a file title pattern that they comply with, which can make them effortless to spot if you have familiarized by yourself.

  • DALL-E three: “DALL·E [Generation Date] [Generation Time] – [Original Prompt]”
  • Adobe Firefly 2: “Firefly [Original Prompt] [Prompt UID]”
  • Midjourney: “[Discord Username]_[Original Prompt Separated By Underscores]_[Prompt UID]”

Here is an illustration of every single:

  • DALL-E three: DALL·E 2023-ten-23 ten.47.46 – a girl with striking blue eyes
  • Adobe Firefly two: Firefly a girl with striking blue eyes 71649
  • Midjourney: discorduser_a_girl_with_striking_blue_eyes_eb1cf9-3a-4b7-8a2e

Google’s New “About This Image” Attribute

Google just rolled out a new function referred to as “About This Image” that lets you verify the origin of an picture. To use this, pick any picture from Google Photographs and click the 3 vertical dots following to the near button. You ought to see this selection now.

Google About This ImageGoogle About This Image

You will be directed to a separate tab containing each and every world wide web outcome that involves the picture. From there, you ought to now have a much better concept of the image’s supply and whether or not or not it came from AI.

Google About This ImageGoogle About This Image

Sadly, this is only offered for Google Photographs. For third-get together pictures, you can use the constantly dependable Google Lens rather.

Google LensGoogle Lens

Examine For Glitches

We lately published an article about identifying which amongst two pictures are AI and human. The following picture is an excerpt from that report. Can you spot which is not actual?

AI Vs. HumanAI Vs. Human

You are receiving a gold star if you mentioned the a single on the left. But how can you inform specifically? Effectively, it is less complicated just to present you.

What Makes The Image AI?What Makes The Image AI?

Recognize how the edges of some petals get distorted? Some even mix collectively. This is frequent in AI pictures. You can also see that the blurring of the foreground and the background is not steady, especially on the daisies on the bottom left.

This symptom is most obvious in faces. Get this picture from Midjourney as an illustration:


You can count on a single hand how several of these faces are not distorted or totally warped, and they are normally the ones on the foreground. This phenomenon is not just constrained to faces as well. Occasionally, even other physique elements can be impacted by rendering troubles, generally manifesting as added fingers or missing limbs. There are occasions when the final results are nightmarish and borderline physique horror pictures even though.


Generative designs even now also register letters as shapes and couldn’t comprehend their which means. This final results in negative text generation, one thing that no AI model can resolve at their existing stage.

Text GenerationText Generation

Target On The Supporting Information

When you daydream, do you commit a whole lot of time considering of the background information? What do the individuals seem like? What language do they talk? What’s the climate like?

The identical goes for AI. They do not dwell on the supporting information. Alternatively, they concentrate their computing electrical power in the foreground, which is also why you do not see considerably encounter distortion in the major topics of the picture earlier.

To show this, allow me give you an illustration:

Find the background inconsistenciesFind the background inconsistencies

Most individuals would quickly glance in the middle of the picture. Nonetheless, when you seem closely, you will commence to recognize that some factors are not including up.

These are all just hidden in plain sight. Here’s yet another illustration featuring automobiles on the sidewalk, distorted background buildings, and missing heads.

More background inconsistenciesMore background inconsistencies

Apply Some Logic

What is incorrect with this picture?

Logic Errors in AI Images 1Logic Errors in AI Images 1

If the watermelons are within the property, how could they get moist from the rain? Oh, and who the hell cuts watermelons like that?!

That is the issue with AI picture generators: they lack frequent sense. Here’s yet another a single:

Logic Errors in AI Images 2Logic Errors in AI Images 2

There shouldn’t actually be a chair atop yet another chair. The shadow of the table’s centerpiece also is not steady with the placement of the lights.

Logic Errors in AI Images 3Logic Errors in AI Images 3

Why is the train in the passenger waiting region rather of the tracks?

AI lacks the nuance that we have. They do not see the planet for what it is, but they see it as an aggregate of factors. They do not have an precise concept of how they match collectively, but they know some objects and variations belong collectively. 

Here’s a quick checklist of logic mistakes that is frequent in AI pictures:

  • Further or missing fingers.
  • Extreme repetition of factors.
  • Inconsistent lighting and shadows.
  • Lack of flaws and imperfections in a encounter.

Use AI Picture Detectors (I did some testing as well!)

Lastly, you could also use detectors to discover out if an picture came from AI or not. There are some that I feel deserve a mention, beginning with Optic’s AI Or Not: a world wide web application that identifies the origin of an picture. This is the most common selection, with media shops this kind of as New York Times and the Wall Street Journal highlighting its effectiveness.

Optic's AI Or Not Landing PageOptic's AI Or Not Landing Page

If you like statistics, you are going to adore Hive Moderation. It is a comprehensive picture suite which involves attributes this kind of as visual moderation, demographic attributes, and reverse picture search. And get this: it does not just give you a probability score for AI, it even outputs a self-confidence score for every single generative model.

Hive Moderation Landing PageHive Moderation Landing Page

An additional instrument you can attempt is Illuminarty. Like AI Or Not, this one’s quite easy. The largest big difference is that you can get an AI probability score even with their totally free strategy, which is not the situation for AI Or Not.

Illuminarty Landing PageIlluminarty Landing Page

The most crucial factor of these detectors is their accuracy. So, I went ahead and did my very own testing. 1 huge advantage of creating about Midjourney, DALL-E, and Firefly for months is that I have my very own assortment of AI pictures, some of which I utilized right here.

Optic’s AI Or Not



Check #one: Midjourney

Most likely To Be AI

a hundred% AI

one.three% AI

Check #two: Midjourney

Most likely To Be AI

a hundred% AI

13.six% AI

Check #three: Midjourney

Most likely To Be AI

a hundred% AI

three.four% AI

Check #four: Midjourney

Most likely To Be AI

a hundred% AI

94.two% AI

Check #five: Midjourney

Most likely To Be AI

a hundred% AI AI

Check #six: DALL-E three

Most likely To Be AI

a hundred% AI

95.four% AI

Check #seven: DALL-E three

Most likely To Be AI

96.9% AI AI

Check #eight: DALL-E three

Most likely To Be AI

a hundred% AI AI

Check #9: DALL-E three

Most likely To Be AI

a hundred% AI AI

Check #ten: DALL-E three

Most likely To Be AI

a hundred% AI

98.two% AI

Check #eleven: Firefly two

Most likely To Be AI

a hundred% AI

83.four% AI

Check #twelve: Firefly two

Most likely To Be AI

99.9% AI AI

Check #13: Firefly two

Most likely To Be AI

a hundred% AI

95.six% AI

Check #14: Firefly two


99.9% AI

51.four% AI

Check #15: Firefly two


91.% AI

twenty.four% AI


 13 out of 15



I also wished to see how powerful these resources are at detecting human pictures. It is an powerful way of figuring out if you are at chance of false positives when employing them.

Optic’s AI Or Not



Check #one: Human

Most likely Human

99.six% Human Human

Check #two: Human

Most likely Human Human

86.four% Human

Check #three: Human

Most likely Human

a hundred% Human

93.five% Human

Check #four: Human

Most likely Human

a hundred% Human Human

Check #five: Human

Most likely Human

78.four% Human

eight.9% Human


 5 out of five



You can confirm these if you want to. For transparency, here’s a Google Drive folder containing all pictures I utilized for these exams.

From these final results, I’d say that Hive is the most powerful AI picture detector in the industry. Of program, this is just a little sample dimension so be diligent in studying which a single you will believe in.

In A Nutshell

So, what’s the most powerful way of spotting AI pictures?

You can constantly seem for watermarks, but they are not constantly there. Some pictures will not seem employing Google’s “About This Picture.” Seeking for inconsistencies, distortions, and logic mistakes is a bit like a game of Where’s Waldo and could consider some time.

I suggest going straight to AI picture detectors if there are no watermarks. Not only is this the quickest, but it is also the most correct.

Nonetheless, the most crucial issue is to hold a sharp eye. After you get utilized to seeing AI pictures, I guarantee you that identifying them will be as effortless as pie.