Visual search is searching with an image instead of typed words. You point your camera at something, or upload a photo or screenshot, and the search engine works out what is in the picture: a product, a plant, a landmark, a line of text. It then shows you information, similar items or places to buy. It is related to reverse image search, but the goal is different: visual search asks "what is this?", while reverse image search asks "where else does this exact picture appear?"
Key takeaways
- Visual search identifies what is in an image; reverse image search finds copies of an image. Most modern tools do both.
- It works by detecting objects in the photo, turning each into a numeric "embedding," and finding the closest matches in a huge index.
- Google says Lens handles nearly 20 billion visual searches a month, and about 20% of them are shopping-related (October 2024).
- The main tools are Google Lens, Bing Visual Search, Pinterest Lens, Amazon Lens and Apple's Visual Look Up.
- Images you search are sent to the provider's servers. Check history settings, and don't upload documents, faces of strangers or anything private.
Visual search, defined
In a traditional search, you have to describe what you want in words. That's hard when you don't know the name of the thing: a chair you saw in a café, a weed in your garden, a sign in a language you can't read. Visual search removes that step. The image itself becomes the query.
There are three common ways to start a visual search:
- Live camera: point your phone at an object and tap the shutter in an app such as Google Lens, Pinterest Lens or Amazon Lens.
- An existing image: upload a photo or screenshot from your gallery or computer.
- Something on your screen: features like Circle to Search on Android and Visual Intelligence on recent iPhones let you select part of whatever is on screen.
Increasingly, visual search is also multimodal: you can add words to an image. Photograph a patterned shirt and add "in green," or select a dish and ask "is this vegetarian?" Google calls this adding to your search; other tools have similar "ask about this image" options.
Visual search vs. reverse image search
The two terms are often used interchangeably, and the same app can do both. The difference is in the question being answered and in what counts as a good result.
| Visual search | Reverse image search | |
|---|---|---|
| Question | What is this, and where can I learn more or buy it? | Where else does this exact image appear online? |
| A good result | The right product, species, landmark or translated text | Pages that host the same photo, ideally the original source |
| Matching style | Similar content (any photo of the same kind of lamp) | Same image, even if resized, cropped or edited |
| Typical uses | Shopping, identifying plants and places, translation, homework help | Checking profile photos, finding an image's source, finding stolen photos |
| Typical tools | Google Lens, Pinterest Lens, Amazon Lens, Bing, Apple Visual Look Up | TinEye, Google Lens (exact matches), Bing, Yandex |
This matters in practice. If you upload a dating-profile photo to Google Lens, its visual-search side may show you people who simply look similar, which is not evidence of anything. For tracing a specific photo, see our complete guide to reverse image search.
How visual search works
Every provider has its own system, and the details are proprietary. But the broad approach used across the industry follows a recognizable pipeline.
Step 1: Object detection
A photo of a living room might contain a sofa, a lamp, a rug and a plant. Before searching, the system runs an object detection model that draws boxes (or precise outlines) around the distinct things in the picture. That is why Lens and similar tools show dots or a selection box on parts of your image, and why you can drag the box to focus on one item. Specialized models run alongside: optical character recognition (OCR) to read text, barcode and QR readers, and classifiers tuned for plants, animals or landmarks.
Step 2: Embeddings
Each detected region is passed through a neural network trained on very large collections of images. The network's output is an embedding: a long list of numbers (a vector) that summarizes what the region looks like, including shape, texture, color and pattern, and, in more advanced models, what it is. Two photos of the same model of sneaker, taken in different light from different angles, produce vectors that sit close together. A photo of a boot lands further away.
Some newer models are trained on images and text together, so a picture of a red dress and the words "red dress" map to nearby points. That shared space is what makes "image plus words" searches possible.
Step 3: Nearest-neighbor search
The provider has already computed embeddings for billions of indexed images: product photos, web images, map and landmark photos. Finding the closest vectors to yours is a nearest-neighbor search. At that scale, systems use approximate methods that trade a tiny amount of accuracy for answers in a fraction of a second.
Because matching compares meaning-like features rather than pixels, visual search still works when your photo is cropped, rotated, badly lit or shows the object from a new angle. That's also why it returns "similar" results rather than only identical ones. Engines built for exact-copy matching, such as TinEye, rely more on compact fingerprints of the image itself, which is why they excel at finding duplicates and edited copies.
Step 4: Ranking and presentation
Raw matches are then ranked using other signals: the text on the pages where matching images appear, product catalog data such as price and availability, your language and location, and the type of query detected (a plant gets a species card; a shoe gets shopping results). Many tools now add an AI-generated summary on top, produced by a model that can read both the image and the matched pages.
What people use it for
Shopping
This is the biggest commercial use. Google said in October 2024 that about 20% of Lens searches are shopping-related. Photograph a jacket, a lamp or a pair of shoes and you'll get matching or similar products with prices from multiple stores. Amazon Lens and Pinterest Lens are built around this. Tip: include a plain background and the whole item for better matches, and check the seller carefully: visual results can include lookalikes and knock-offs.
Identifying plants, animals, landmarks and art
Visual search is good at common species, dog and cat breeds, famous buildings and well-known artworks. It's a great starting point for curiosity, gardening or travel. It is not a safety tool: never rely on an app's identification to decide whether a wild mushroom or plant is safe to eat or touch.
Translating and copying text
Point Google Lens at a menu, sign or label and it overlays a translation in place. Circle to Search on Android can translate everything on screen. The same OCR lets you copy text from a photo, such as a Wi-Fi password on a router sticker or a quote from a printed page, and paste it elsewhere.
Homework and learning
Lens and Bing can recognize a photographed math problem or diagram and point to explanations or similar worked examples. Used well, that's a study aid; copying answers wholesale doesn't help anyone learn, and AI-generated explanations can contain mistakes, so check them.
Checking images and people
Many people reach visual search tools while trying to check whether a photo is genuine or a profile is real. That's really a reverse image search job. See our catfish check guide.
The main visual search tools
| Tool | Where you find it | Best at |
|---|---|---|
| Google Lens | Google app, Chrome, Google Photos, many Android cameras; Circle to Search on supported Android phones | All-round identification, shopping, translation, text, homework |
| Bing Visual Search | bing.com > Images > camera icon; Microsoft Edge | Similar images, products, pages containing the image |
| Pinterest Lens | Camera icon in the Pinterest app's search bar; Lens icon on the website | Style, home decor, recipes, outfit ideas |
| Amazon Lens | Lens icon in the Amazon Shopping app's search bar | Finding products on Amazon by photo, upload or barcode |
| Apple Visual Look Up / Visual Intelligence | Photos app Info button; screenshots and Camera Control on supported iPhones | Plants, pets, landmarks, art; quick on-device lookups |
Google Lens is the most capable general-purpose option and is free. Google said in October 2024 that Lens handles nearly 20 billion visual searches a month. Bing Visual Search works on desktop and mobile: open Bing Images, tap the camera icon, and take a photo or upload one. Pinterest Lens works in the app (tap search, then the camera icon) and on the website (click the Lens icon in the search bar). It returns Pins rather than web pages, which suits inspiration more than fact-finding. Amazon Lens offers camera search, photo upload and barcode scanning, and lets you circle one product in a busy photo. Results are limited to Amazon's catalog. Apple's features are covered in our guide to searching images on iPhone and Android.
If your goal is tracing a photo rather than identifying what's in it, compare engines in our roundup of the best reverse image search tools.
Privacy considerations
Unless a feature explicitly runs on your device, a visual search sends your image to the provider's servers. Keep these points in mind:
- History settings. Google's optional Visual Search History saves images you search with Lens and Circle to Search to your Web & App Activity. Google says it's off by default and that saved images may be used to improve its visual search technology. You can check it under Google Account > Data & privacy > Web & App Activity > Include Visual Search History, and delete saved items at myactivity.google.com.
- Service improvement. Microsoft states that photos you provide to Bing Visual Search may be used to improve Bing's image processing services.
- Hidden metadata. Photos straight from your camera can contain location and device details. Learn what's stored and how to remove it in our guide to photo metadata (EXIF).
- Sensitive content. Don't upload ID cards, bank documents, medical paperwork, photos of children, or intimate images to any search tool.
- Other people's faces. General visual search tools won't reliably put a name to a stranger's face. Dedicated face search engines exist, but they raise serious privacy and consent questions; see our explainer on face search engines before using one.
Limitations and how to get better results
Visual search is impressive but far from perfect.
- Confident mistakes. Lookalike species, similar products and generic buildings get confused, and AI summaries can state a wrong answer fluently. Microsoft's own help page notes that AI might still make mistakes.
- Commercial bias. Many tools lean toward shopping results because that's where matches are easiest to monetize. You may need to scroll past products to find information.
- Coverage gaps. Rare items, local products, handmade goods and content behind logins (private social accounts, most dating apps) are poorly indexed or not indexed at all.
- Image quality. Blurry, dark or tiny images produce weak embeddings and weak results.
- It can't confirm authenticity. Finding similar images doesn't tell you whether a photo is real, edited or AI-generated.
Frequently asked questions
Is Google Lens visual search or reverse image search?
Both. By default Lens leans toward visual search: identifying objects, shopping and similar images. It can also show pages containing the same image, which is reverse image search.
Is visual search free?
Yes. Google Lens, Bing Visual Search, Pinterest Lens, Amazon Lens and Apple's Visual Look Up are free to use. They are funded largely through shopping and advertising.
Can visual search identify a person from a photo?
General tools like Google Lens and Apple's Visual Look Up aren't built to name strangers from their faces. They may surface pages where the same photo appears, which is reverse image search. Specialized face search engines do attempt face matching, with significant privacy concerns.
Does visual search work offline?
Mostly no. Searching the web and product catalogs requires a connection. Some on-device features, such as basic text recognition on phones, can work without one, but full results need the internet.
Sources & further reading
- Google: Google Lens now offers shopping product details (October 2024)
- Google Search Help: Manage your Visual Search History
- Microsoft Support: Using Bing Visual Search
- Pinterest Help: Pinterest Lens
- About Amazon: How to use Amazon Lens
- Apple Support: Look up what's in a photo with your iPhone or iPad
Spotted an error or something out of date? Tell us — we review corrections promptly. Read how we research and update guides in our editorial policy.