How to Do a Reverse Image Search Using Google Lens, TinEye, and Bing on Mobile and Desktop

You are trying to answer the wrong question. You see a picture and ask: is this real? A reverse image search cannot tell you. It traces ancestry. It answers one question: where has this image appeared before? On a phone, Google Lens does that job. On a desktop, TinEye is faster and sorts by the oldest match. Bing Visual Search sits in between, useful for finding product pages and cropped versions. The goal is provenance, not a guess at synthetic origin. An image that looks like a genuine photo of a flooded street in July might be a genuine photo of that same street in January, recycled to fit a different story. The search does not tell you the flood is fake. It tells you the image is old. That distinction is the entire point.

The failure mode you are trying to avoid is the AI realism trap. You see a photorealistic shot of a politician shaking hands with a foreign leader, and your first reaction is to ask whether the picture is real. That is the wrong question. Generative AI now produces photorealistic synthetic media from a text prompt. The picture can look flawless and still be fake. Reverse image search does not care about pixels. It does not scan for artifacts. It builds a fingerprint of the picture and checks it against a database of known material. If the same visual exists somewhere else on the web, the tool finds it. If it does not exist anywhere, you have learned something else: the image is either very new, very obscure, or generated. The tool cannot tell you which. What it can do is show you whether the same photo was used last year in a different country with a different story attached. That is the single most useful piece of information you can get.

Start With The App You Already Have

For a phone workflow, you do not need to install anything new. The tool is embedded in both the Google app and the Google Photos app. On an Android phone, open Google Photos, select the image, and tap the Lens icon at the bottom of the screen. On an iPhone, the Google app works the same way, as does the Google Photos app, which is free on the App Store. If the image is on your screen, not in your camera roll, use the screenshot method: take a screenshot, crop it to isolate the relevant part, then run Lens on the crop. The crop is important. A screenshot of a news article contains both the picture and the text around it, and Lens treats the whole frame as input. Cropping removes the noise and improves the match.

Read The Output As A Lead, Not A Verdict

Once Lens has the image, it runs it through computer vision neural networks. The output screen shows you visually similar pictures, text extractions if the image contains text, and sometimes product listings if the image contains an object. Tap through to the matches that say 'visually similar' rather than 'text in image'. The similar-images panel is where you find the earliest match, or at least the earliest one Google has indexed. Google does not disclose the size of its index, so you cannot know what it has missed. That is a limitation, not a flaw. Treat a Google Lens result as a lead, not a verdict. The mobile interface is fast. It is not exhaustive.

You Are Checking Context, Not Pixels

Let us be direct about what 'how to check if image is real' means in practice. You are not checking if the image is real. You are checking if the image is real in context. The technique is lateral reading: leave the page that contains the image and open new tabs to check the source. The reverse image search is the first tab you open. Start with a Google Lens search on your phone to see what comes back. Then switch to a desktop. The desktop tools are better for this specific job. The phone is good for a quick check while you are scrolling. The desktop is where you can sort by date, filter by domain, and compare versions side by side.

The out-of-context image is a specific and well-documented failure in media literacy. It is the cheapfake that requires no AI at all: someone takes a real photo from a real event, strips the date and place, and attaches a different date and place. The image is authentic. The claim is false. A reverse image search catches this because it finds the original context. A tool that scans for AI artifacts will not catch it. There is nothing to catch. The image was never manipulated. You need provenance, not a pixel analysis.

Work Backwards To The First Appearance

For out of context photo detection, work backwards from the image to its first appearance. The tool that does this best is TinEye. TinEye uses perceptual hashing. It creates a digital fingerprint of the image and compares that fingerprint against its database of 72.1 billion images. This is different from a pixel-by-pixel match. Perceptual hashing finds images that have been cropped, resized, or slightly edited. The TinEye output page has a key feature that Google Lens lacks: a 'Sort by Oldest' button. Click it. The oldest appearance is what you want. The oldest appearance is almost always the original context. A photo of a protest in Jakarta in 2019 shows up in 2019 when you sort by oldest. If someone is using that photo to claim a protest in 2024, the date mismatch is visible immediately.

Use The Compare Tool And The Browser Extension

TinEye also has a 'Compare' feature that shows you the original image side by side with the version you uploaded. Use it when the crop is tight. The compare view reveals exactly what was removed or added. Install the browser extension for Chrome, Firefox, or Edge. It adds a right-click option to search any image on the web. That right-click shortcut is the fastest way to check an image while you are reading an article. One click, and the output opens in a new tab. The extension does not upload the image to TinEye's servers in a way that exposes your identity. It sends the URL or the file to the service you have selected.

Black Dot Research, a Singapore-based fact-checking organisation, recommends reverse image search as standard practice for verifying viral images. The POFMA Office's own fact-checking articles from 2023 to 2024 use the same technique. This is not esoteric. It is what fact-checkers do, and it is what you should do. The IMDA's Digital for Life resources include guidance on verifying online content and point to the same method. You are using the same workflow as professional verification shops.

Three Ways To Feed TinEye An Image

On a desktop session, you have three input methods. Drag and drop an image file into the search box. Paste a URL into the field and press enter. Or right-click an image anywhere on the web if you have installed the browser extension. The URL paste method is the one most people skip. That is a mistake. If you are looking at an image on a news site, you do not need to download it. Right-click the image, copy the image address, paste it into TinEye. This preserves the original resolution and avoids any compression the site might have applied.

TinEye accepts JPEG, PNG, GIF, BMP, TIFF, and WebP. The maximum upload file size is 20 MB, which is generous for any practical purpose. The output list shows the number of times the image has been found on the web, the oldest appearance date, and the domain where each match lives. That domain filter is crucial. An image appearing on a site you have never heard of in a language you do not read is a red flag. An image appearing on a reputable news site with a date from five years ago is your answer. TinEye does not do facial recognition. It will not tell you who is in the photo. It will tell you where the photo has been.

Find The Product, Then Find The Page

Bing has been doing reverse image search since 2018, and it has one advantage over both Google Lens and TinEye: it excels at finding the product in the image. If your picture contains a specific handbag, a pair of shoes, or a piece of furniture, Bing often finds a product page where you can buy it. This is not useful for news verification. It is extremely useful for a different kind of research. The method is the same: upload the image, paste the URL, or drag and drop the file. Bing's maximum upload size is 10 MB, half that of the others, so compress a very large file first.

Bing's output includes visually similar pictures, but the real value is the 'Pages that include this image' section. This section tells you which web pages have embedded the image, not just which pages have similar ones. This is closer to TinEye's approach than to Google's. Bing also does text extraction. If your image contains a sign, a receipt, or a document, copy the text directly out of the image. The deep learning object detection in Bing is especially good at identifying landmarks and text. For a quick check of whether an image came from a specific event, Bing is a solid second pass. Use it after you have checked with Google Lens and before you move to TinEye for the oldest appearance.

Reverse image search works best on still images, but a lot of disinformation is disseminated as video. To check a video, extract a keyframe. Play the video, find a clear frame that is not motion-blurred, and take a screenshot. On a phone, pause, screenshot, crop. On a desktop, use the print screen button or the snipping tool on Windows, or command-shift-4 on a Mac. Then run that keyframe through any of the three tools. A video is just a sequence of still images. The frame that gets shared as a thumbnail is often the most striking, and it is the one picked up by the image index. Do not neglect this. A cheapfake that re-captions an old video clip gets caught the same way an out-of-context photo gets caught: by finding the earliest appearance of the keyframe.

Zero Results Proves Nothing

Here is the failure case. If your reverse image search returns zero results, you have not proven the image is real. You have proven the image is not indexed. The image could be too new, or it could be generated. A deepfake of a person who does not exist returns zero results on a reverse image search. The image has never existed before. That does not mean it is real. It means it is new. This is where you need a different tool, and you need to be sceptical of any tool that claims to detect AI-generated content. The real-world false positive rate of AI-content detectors is high enough that no responsible fact-checker treats them as definitive. They are a lead, not a verdict. The reverse image search is a lead too, but it is a much better one. It gives you a dated, citable source: the earliest appearance.

The Two-Minute Sequence

The single best move you can make is to use all three tools in sequence. Start with Google Lens on your phone to get a quick read. Move to Bing Visual Search to find product pages or to see if the image is being discussed on forums. Finish with TinEye on the desktop to sort by oldest and to check if the image has been altered. This sequence takes two minutes. It costs nothing. Black Dot Research and the POFMA Office use the same technique. It does not require you to understand how the algorithms work. It requires you to understand the question you are asking. The question is not 'is this image real'. The question is 'where has this image been before'.

This method is for the curious traveller of information: the person who sees a viral image in a group chat and wants to know, before they share it, whether it is what it claims to be. It is for the student writing a paper who needs to verify a source. It is for the parent who has been sent a warning about a local danger and wants to check if the photo is actually from Singapore or from a flood in Jakarta five years ago. It is for the person who wants to understand the distinction between content moderation platform rules and legal action under Singapore law, and who knows that a POFMA direction is not the same as a fact-check article.

It is not for the person who wants a quick yes-or-no verdict with zero effort. That person exists, and there are tools that pretend to serve them. Those tools are not reliable. It is not for the person who believes that because a message comes from a trusted friend, the underlying image must be true. That is source confusion, and no tool can fix it. It is not for the person who has already decided the answer and is looking for confirmation. The method works only if you are willing to follow the evidence, even when it contradicts your prior belief. If you are not willing to do that, no image search will help you. If you are, this two-minute habit will change the way you see the internet. The tools change. The principle is durable: check where it came from before you decide what it means.