Founders, Sellers and Families: Check if Image Is AI Generated Before You Trust or Forward It

There was a time when a photo was proof. If someone sent you a picture of a factory, a flooded road, a damaged parcel or a politician at a wedding, you believed it, because making a fake took skill and hours. That time is over. Today, anyone with a phone can create a realistic image in under a minute, and most of us cannot tell the difference by looking.

So the useful skill now is not “spot the fake by eye”. It is knowing how to check if image is AI generated with a short routine that uses free tools, a little patience and some common sense. That is what this guide gives you.

I will walk through five steps in the order I use them: looking for a label or watermark, running a SynthID detector check, reading the metadata and content credentials, doing a reverse image search, and only then looking for visual tells. I will also explain how far to trust any AI image detector, what to do with a fake image WhatsApp forward, and how to prove your own photos are real. Whether you are a founder checking a supplier, a seller checking a competitor’s claim, or a parent checking a family group forward, the same routine works.

The Underpass Photo in Our Building Group

It was 7:40 on a Thursday morning. I was polishing my office shoes on the balcony when the building’s WhatsApp group started buzzing. Someone had forwarded a photo of the underpass on my route to office, fully under water, with an auto half-sunk and people standing on the divider.

Within five minutes there were thirty replies. “Take the long route.” “Work from home today.” “Who will pay if the car gets damaged?” I almost called my manager to say I would be late. Something about the photo felt too neat. The water was very still for a flood. The auto’s number plate was blurred in a strange, smooth way.

I spent four minutes checking it. The same photo had appeared months earlier with a caption from a different city. It was not even our underpass. I took my usual bus, crossed a perfectly dry road, and reached office on time. When I posted a short reply in the group with what I had found, two people thanked me and one person was annoyed. That is roughly how it goes with fake image WhatsApp forwards, and it is why I wrote this routine down.

Why It Is Now Hard to Check if Image Is AI Generated

Before the steps, it helps to understand why this has become difficult, because it explains why no single tool is enough.

AI images look real now

Early AI images had obvious problems: melted faces, extra fingers, garbled text. Many current tools have fixed most of these. If you want to see how far image tools have come, my comparison of Nano Banana Pro vs ChatGPT image accuracy shows how realistic product and people images have become.

Real photos get edited heavily

Phone cameras already use software to brighten faces, smooth skin and sharpen details. Editing apps add filters and AI tools that remove objects or change backgrounds. So a “real” photo may have AI touches, and an AI image may have been edited by a person afterwards. The line between real and synthetic is blurry.

Screenshots and forwards destroy clues

Every time an image is screenshotted, compressed or forwarded, it loses information. Watermarks can weaken, metadata disappears, and quality drops. By the time a fake image WhatsApp forward reaches you, most technical clues may be gone. That is why the routine includes steps that work even on a poor copy.

Detectors are not magic

Many websites promise to Audit with one click. Some are useful. None is perfect. I will explain later why you should treat an AI image detector as one opinion, not the final answer.

Step 1 to Check if Image Is AI Generated: Look for a Label

The easiest way to Inspect is to see whether someone has already told you.

Platform labels

Instagram, Facebook, YouTube and other platforms now show labels on content that creators declare as AI, or that the platform detects as AI. In India, this has become more important since the amendment to the IT Rules came into force on 20 February 2026, which requires platforms to label synthetically generated information prominently. My guide to the AI labelling rules India set explains what platforms must do.

If the post shows an “AI info”, “Made with AI” or “altered or synthetic content” label, you have your first answer.

Visible watermarks and captions

Some AI tools add a small visible mark or logo in a corner. Some creators write “AI image” in the caption or on the image. Look at the corners, the bottom edge and the caption.

What a missing label means

A missing label does not prove an image is real. Many people do not declare AI content, and screenshots remove platform labels. So if Step 1 gives you a label, good. If not, keep going.

Step 2: SynthID Detector and Other Watermark Checks

The second way to check if image is AI generated uses invisible watermarks. This is where the SynthID detector comes in.

What SynthID is

SynthID is Google’s system for adding an invisible watermark to content made with Google’s AI tools. According to a report on Techlomedia, Google says these watermarks are invisible to the eye and cannot be removed without damaging the image, and SynthID has been applied to more than 20 billion pieces of AI content since 2023.

How to use the SynthID detector in the Gemini app

In November 2025, Google added a SynthID detector check to the Gemini app. The same report describes the steps:

1. Open the Gemini app.
2. Upload the image.
3. Ask a direct question, such as “Was this created with Google AI?”
4. Gemini checks for a SynthID watermark and explains the result.

In December 2025, ETV Bharat reported that the check was extended to video, for files up to 100 MB and 90 seconds, in countries where the Gemini app is supported.

What the SynthID detector cannot do

This is the most important part. The SynthID detector only finds Google’s own watermark. If an image was made with another company’s tool, the check will not find SynthID, and that tells you nothing about whether it is AI. ETV Bharat also notes research suggesting such watermarks may not be foolproof.

So use the SynthID detector like this:

• SynthID found: strong sign the image was made or edited with Google AI.
• SynthID not found: the image may be real, or made with a different tool. Continue to the next steps.

Other watermark systems

Other AI companies use their own markers, and many are joining the content credentials standard explained in Step 3. As more tools add markers, it will become easier to check if image is AI generated at the source. For now, treat each watermark check as covering only its own maker.

Step 3: Metadata and Content Credentials

Every digital photo carries hidden information called metadata. Some AI tools and cameras also add content credentials. Both can help you check if image is AI generated, if they survive.

What metadata shows

A photo from a phone camera usually stores the camera model, the date and time, and sometimes the location. An AI image often has no camera information at all, or names the software that made it. On a computer, you can see basic metadata by right-clicking the file and opening its properties. On a phone, many gallery apps show details with an “info” button.

Content credentials

Content credentials are a standard backed by many technology and media companies. They work like a tamper-evident label attached to the file, recording how it was created and edited, including whether AI was used. You can upload a file to the free Content Credentials Verify tool to see if it carries this information.

If the file has content credentials showing an AI tool, you have a clear answer. If it shows a camera and normal edits, that is a good sign the image is real.

Why WhatsApp images usually have nothing

Here is the catch for most Indians. WhatsApp and many social apps compress images and strip much of the metadata when you send them. So a fake image WhatsApp forward will often show no camera details, no content credentials and no AI tool. Empty metadata does not mean fake. It only means the clues were removed on the way.

Ask for the original

If you need to check if image is AI generated for something important, like a supplier’s factory photo or a damage claim, ask for the original file, sent as a document instead of a photo. On WhatsApp, sending as a document keeps the original file. If someone refuses to share the original for a business decision, that refusal is information too.

Step 4: Reverse Image Search

Reverse image search is my favourite way to check if image is AI generated or misused, because it works even on low-quality forwards. It is how I solved the underpass photo in four minutes.

What it does

A reverse image search looks for the same or similar images across the internet. If the “new” image appeared years ago, in a different city, with a different caption, you know the story attached to it is false, whether or not it is AI.

How to do a reverse image search

• Google Lens: In the Google app or Chrome on your phone, use Google Lens on the image, or upload it to Google Images on a computer. Look at the “exact matches” and the oldest dates you can find.
• Other search engines: Bing and other services also offer image search. Trying two engines finds more matches.
• Crop and search again: If the full image finds nothing, crop the most unusual part, like a sign, a building or a face, and search again.

What to look for in the results

1. Older copies. The earliest date you can find is often the real origin.
2. Different captions. A flood photo from another city, a protest photo from another year.
3. AI art sites. If the image appears on an AI art gallery or a prompt-sharing page, you have your answer.
4. Stock photo sites. Sometimes “news” images are stock photos or AI stock images.

Reverse image search does not always tell you if image is AI generated. But it very often tells you whether the story around it is true, and for most decisions, that is what matters.

Step 5: Visual Tells, the Last Way to Check if Image Is AI Generated

Only now, after four other ways to check, do I look closely at the image itself. Visual tells are useful, but they are getting weaker every month, so they come last.

Hands and fingers

Look at hands. Extra or missing fingers, fingers merging, odd nails or rings that melt into skin are still common in some AI images.

Text and signs

Zoom in on any writing: shop boards, number plates, labels, newspapers. AI tools often create text that looks like letters from a distance but turns into nonsense up close. Indian scripts like Telugu, Hindi and Tamil are often garbled in AI images.

Light and shadows

Check whether shadows fall in the same direction as the light source, and whether reflections in water, glass or eyes match the scene.

Backgrounds and edges

AI images often have backgrounds that do not make sense when you look closely: railings that end mid-air, windows that do not line up, crowds of faces that blur into each other.

Too perfect

Skin with no pores, water with no ripples in a “flood”, food that looks better than any real dish, and lighting that looks like a film studio in a village. Perfection itself is a clue.

Why tells are fading

Every month, AI tools fix more of these problems. So a clean image with no tells is not proof of reality. Use visual tells to raise suspicion, not to give a final verdict. When you check, the visual step should confirm what the earlier steps suggested.

AI Image Detector Tools: How Far to Trust Them

There are many websites that promise to find in one click. They are worth knowing about, but you need to understand their limits.

How they work

An AI image detector is itself an AI model, trained on large sets of real and AI images. It looks for patterns and gives a score, like “85% likely AI”.

Why they make mistakes

• False positives: real photos, especially heavily edited or low-quality ones, can be marked as AI.
• False negatives: new AI tools the detector has not seen can pass as real.
• Compression: WhatsApp and social media compression can confuse detectors.
• Different detectors disagree: the same image can get very different scores on two sites.

How I use them

I use an AI image detector as one more opinion, after the first four steps. If a detector says “likely AI” and the reverse image search finds the image on an AI art site, I am confident. If a detector says “likely AI” but the photo has camera metadata and the supplier sent the original file, I trust the evidence over the score.

Never accuse a person, a supplier or a seller of faking an image based on a detector score alone. That is unfair to them and risky for you.

Fake Image WhatsApp Forwards: What to Do

Most of us need to know because of family and community groups. Here is a calm way to handle a fake image WhatsApp forward.

Do not forward first

The most useful thing you can do is pause. If a photo makes you angry, scared or excited, that is exactly when to check before sharing it.

Ask for the source

Reply politely: “Where is this from?” Often the person who forwarded it does not know. That question alone slows the spread.

Check quickly

Use reverse image search first. It is the fastest way to Inspect or simply taken from another place.

Reply with care

If you find it is fake or misused, reply with a short, kind message and the link or screenshot that shows the original. Avoid making the sender feel foolish. People are more likely to stop forwarding when they are not embarrassed.

Report when needed

Platforms let you report misleading content. For serious cases like fake news about violence, health or elections, use the report options in the app.

When It Matters for Your Business

For founders and sellers, the stakes are money, not just embarrassment. Here are situations where you should always Inspect before acting.

Supplier and factory photos

A new supplier sends photos of a big factory, stacked stock and a smart office. Before paying an advance, ask for a short live video call, recent photos sent as documents, and the factory address so you can check it on maps. If you are borrowing to fund a big order, my guide on business ideas that need a bank loan explains why careful checks before big payments matter so much.

Damage and return claims

Customers sometimes send photos of damaged products for refunds. Most are genuine. If a claim looks unusual, ask for a short video or a photo taken with a specific item in the frame, like today’s newspaper or a handwritten note with the order number.

Competitor claims

If a competitor’s ads show results that look too good, do not copy them. Many “before and after” images now use AI.

Your own marketing

If you use AI images in your marketing, label them. Customers who later discover an unlabelled AI image feel cheated. A small business lives on trust, a point I make in word of mouth vs ads for a first-year business.

Proving Your Own Photos Are Real

The other side of this problem: what if someone tries to Verify on your real photo, and the answer comes back wrong? It is happening more often, especially to sellers with very clean product photos.

Keep the originals

Save the original camera files, not just the edited versions. Keep them in a dated folder.

Use content credentials where available

Some newer phones, cameras and editing tools can attach content credentials to photos. If your device or app offers it, turn it on. It gives anyone who doubts your image a way to check its history.

Show the process

For product photos, a short behind-the-scenes clip of the shoot, or a photo of the setup, is powerful proof. It also makes good content.

Respond calmly

If someone says your image looks AI, share the original or the behind-the-scenes clip without arguing. Most people accept clear proof.

For small brands, this habit pays off twice: it protects you from accusations, and it gives you honest marketing material. If you run a shop like the one in my small-town pharmacy lessons, real photos of your real counter build more trust than any AI scene.

Check if Image Is AI Generated on Your Phone in Two Minutes

Most people will never open a laptop for this. So here is how to Verify using only a phone, in about two minutes.

Minute one: the fast checks.

1. Long-press the image and save it. If it came on WhatsApp, ask for it as a document if the decision matters.
2. Look at the caption and corners for a label or watermark.
3. Open the Gemini app, upload the image and ask, “Was this created with Google AI?” This is the quickest SynthID detector check on a phone.

Minute two: the search.

1. Open Google Lens from the Google app, select the image, and look at the matches. Scroll to find the oldest copy.
2. Zoom in on hands, text and signs. Indian scripts are especially useful, because AI tools often garble them.

That is it. Two minutes to Audit or misused is a small price before you forward a scary photo to fifty relatives or pay an advance to a new supplier.

Teaching Family and Staff to Check if Image Is AI Generated

The routine only protects you if the people around you use it too. In a small business, the person who receives a supplier photo or a customer complaint is often a staff member, not the owner.

For staff

Make it a rule: before any payment, refund or order based on a photo, the staff member must Audit AI generated with the phone routine and note the result in the order sheet. It takes two minutes and removes a lot of risk. Handing this kind of small check to your team is exactly the kind of delegation I describe in when a small business owner stops doing everything.

For family

Older family members forward most of the fake images, not because they are careless, but because they trust the person who sent it. Sit with them once and show them Google Lens on one real forward. Seeing the same photo from a different city years earlier is usually enough to change the habit.

For yourself

Make it automatic. If an image makes you feel a strong emotion, pause and Examine if image is AI generated before you react. Strong emotion is the most reliable warning sign of all.

A note on money decisions

Fake images cost the most when money is involved: advances to suppliers, refunds for “damaged” goods, investments based on photos of a busy shop. In the first months of a business, cash is tight, and one wrong payment hurts badly, as I explain in money in the first six months of a small business. A two-minute check is cheap insurance.

A Quick Routine You Can Save

Here is the whole method to Examine if image is AI generated on one card. Save it in your phone notes.

1. Label? Look for an AI label, watermark or caption.
2. SynthID? Ask Gemini “Was this created with Google AI?”
3. Metadata? Check file details or content credentials. Ask for the original as a document.
4. Reverse search? Google Lens or another image search. Find the oldest copy.
5. Visual tells? Hands, text, shadows, backgrounds, too perfect.
6. Detector? Use an AI image detector as one more hint, never the final word.
7. Decide. For forwards: do not share without a source. For business: do not pay without proof.

This routine takes two to five minutes. After a week of practice, it becomes a habit. It is the same habit I bring to AI tools in general, whether it is learning Claude Code, reading about new models like the JEV AI model, or following the GPT-5.1 update: trust, but check. And if you want a broader view of where AI is heading for small businesses, my notes on the AI agents market in 2026 and the Grok AI agents guide for solo founders are good next reads.

Questions Readers Ask Me About Spotting AI Images