TinEye Reverse Image Search

By Bill Hartzer · Updated 27 October 2026

What TinEye is, and why it is not a web search engine in the usual sense

TinEye is a reverse image search engine: instead of typing words, the user supplies a picture — by upload, by URL, by drag-and-drop or through a browser extension — and TinEye returns the web pages on which that same picture, or an edited copy of it, appears. It has been operated by Idée, Inc. of Toronto since its launch in 2008.

On the question this site asks of every engine — whose index answers the query — TinEye's answer is unambiguous. It crawls the web itself and answers from a collection it built. It does not re-serve Google's, Bing's or anyone else's results. TinEye's own help center says so directly: the company is "constantly crawling the web, adding to our collection of billions of images."

So why is it not classed as an own-index web engine? Because what it indexes is not the web in the sense the other verdicts on this site use. TinEye does not store page text. It does not answer keyword queries. It does not rank documents by their relevance to a question. It stores compact digital signatures of images and compares a submitted image against them. That makes it a specialist, non-web index: genuinely independent, genuinely crawler-built, and limited to pictures.

The distinction matters because readers often lump TinEye together with Google Lens. They are built for opposite jobs. Google Lens is designed to recognize what is in a picture — a landmark, a plant, a product. TinEye is designed to find copies of the same picture, including copies that have been resized, cropped, recolored or lightly edited. It generally does not return merely similar images, it does not identify objects, and it does not perform facial recognition.

How the index is built and searched

TinEye's collection grows by continuous crawling, and the crawl is not uniform. TinEye states that stock-photo and curated collection sites — it names Wikipedia, Flickr and NASA as examples — are crawled more often, and that images from those sources are labeled "stock" or "collection" in the results. That labeling is a small but useful signal: it tells the user when a match is likely the licensed original rather than a reuse.

TinEye also states, in the same help article (last updated 13 March 2025), that "We do not typically add social media images" to its index. That is the single most important practical limit on the engine. A picture that has circulated mainly on social platforms may return few or no matches, not because it is rare, but because TinEye has chosen not to collect the places where it lives.

The matching mechanism, in outline, is this. Each indexed image is reduced to a compact digital signature, sometimes called a fingerprint — a short numerical summary of the image's visual content that is far smaller than the image itself. A submitted image is reduced the same way, and its signature is compared against the signatures in the index. The comparison tolerates scaling, cropping and many kinds of editing. TinEye does not publish the algorithm, and whether it is a perceptual hash, a local-feature scheme or a learned embedding has not been disclosed; anything more specific than "compact signature" is guesswork.

Results show the number of matches, a preview, and the URL of each matching page. They can be sorted by best match, most changed, biggest image, newest or oldest. Accepted formats are JPEG, PNG, WebP, GIF, BMP and TIFF, with a 20 MB maximum upload.

How big the index is — and how sure anyone can be

TinEye is one of very few search engines that publishes its own index count, displaying it on its home page and in its results. The dated figures available for this page trace a steep curve:

23 September 2011 — 2,013,406,246 images, per the TinEye blog.
June 2014 — more than 5 billion images.
September 2025 — more than 77.6 billion images, as shown in TinEye's search results and reported on Wikipedia.

The September 2025 figure is the most recent one this page can support. A secondary review dated 30 September 2026 reported "more than 85 billion images", but that is a third party's reading, and TinEye's own counter was not independently re-checked for this page in October 2026: the tineye.com application could not be rendered from the research environment. Treat the 77.6 billion figure as dated, and expect the live counter to be higher.

Two pieces of context keep these numbers honest. First, growth from about 2 billion images in 2011 to roughly 78 billion in 2025 is the real story — a nearly fortyfold expansion for a privately held Canadian company. Second, even tens of billions is still a small share of the images on the web; Wikipedia noted that the 2014 figure was "relatively small" against the web as a whole.

The comparison with the large engines is harder than it looks. Google and Yandex publish no image-index count. Microsoft has published one only indirectly: its 2018 research paper on Bing visual search described "tens of billions of images in the index". Those are 2018 figures from a research paper, not a current product statistic, and they do not make the two indexes directly comparable — Bing's system is built to find visually similar images and objects, not only copies.

Who runs it, and how it makes money

TinEye belongs to Idée, Inc., founded in Toronto in 1999 by Leila Boujnane and Paul Bloore. The company is privately held; no public ownership filings were found, and ownership beyond the founders has not been established.

The consumer search engine is free, with optional registration that stores search history, and it is intended for non-commercial use. The business sits behind it. Under the TinEye brand, Idée sells image-matching technology to organizations, and structurally the line-up is:

TinEye API — programmatic search of TinEye's web index, used for verification, moderation of user-generated content, copyright compliance and fraud detection. It is sold in prepaid bundles of searches rather than by subscription.
MatchEngine — duplicate and modified-copy detection inside a customer's own image collection, not TinEye's web index.
MobileEngine — mobile image recognition linking physical items to digital content.
MulticolorEngine — color extraction and color-based search.
WineEngine — recognition of wine, beer and spirits labels.

A monitoring product, TinEye Alerts, has also been offered commercially; its current form could not be confirmed. The structural point is clear without it: the free search engine is the shop window, and the business is licensing image recognition. The consumer engine demonstrates the index; the paying customers are organizations that need image matching at volume.

History

Idée was founded in 1999, nine years before TinEye appeared. The engine launched in invitation-only beta on 6 May 2008 and opened to the public in August 2008. TinEye describes itself as the "first image search engine on the web to use image identification technology" — a company claim, reported as such, that has not been independently adjudicated.

The index passed 2 billion images in September 2011 and 5 billion by June 2014. Interactive API documentation followed in October 2019, alongside changes to the Safari extension; an Edge extension was announced in March 2021, and an auto top-up feature for prepaid API bundles was added the same month. The blog remained active into 2026, with posts dated April and May. As of 27 October 2026, TinEye's help center, its commercial services site and its blog were all live.

The open-source metasearch software SearXNG ships a TinEye engine, and its documentation notes that it scrapes the site rather than using the official API — a small illustration of how few independent reverse-image indexes exist for other tools to draw on.

Who it suits, and who it does not

It suits anyone asking where a specific picture has been used. TinEye was designed for copyright owners and brand marketers tracing unauthorized uses, and it is widely used by fact-checkers and journalists tracing a picture's earliest appearance online. Its matching is strict, which is a strength for this job: a list of exact and edited copies is more useful for provenance work than a page of look-alikes. The absence of facial recognition is a deliberate design choice that sets it apart from face-search services such as PimEyes.

It is a poor choice in four situations, each grounded in how the index is built:

Identifying what is in a picture. TinEye does not recognize objects, places, plants or products. That is Google Lens's job, or Bing's visual search.
Finding visually similar images. It generally returns copies, not alternatives in the same style.
Tracing images that live on social media. By TinEye's own statement, social-media images are typically not added to the index.
Identifying a person from a face. It will not, by design.

For verification work, the sensible practice is to treat TinEye as one engine among several rather than as a final answer. An empty TinEye result means TinEye has no copy — not that no copy exists.

Common misconceptions

"Oldest result" does not mean "original". Sorting by oldest is why journalists use TinEye to estimate when an image first appeared online, and it is useful. But the date reflects the earliest time TinEye crawled a copy, not the date the image was created or first published. An image that circulated for years before TinEye found it will show a misleadingly recent "oldest" date.

TinEye is not a front end for Google or Bing. Its index is its own, built by its own crawling. That is why its results differ from Google's and Bing's, sometimes sharply.

A bigger counter is not a more complete answer. Tens of billions of images is a large collection and still a sample of the web. Coverage depends on what TinEye chooses to crawl, and it has chosen to leave out most social media.

TinEye is not a face-search tool. It does not perform facial recognition, and expecting it to identify a person from a photograph misunderstands what it was built to do. TinEye's own description of its index sets out what it collects and what it leaves out.

TinEye logo history

TinEye has used 2 documented logos. Each is shown with the period it can be dated to from a published source; where only a first sighting is known, the label says "by" rather than guessing a start date.

by Oct 2010
Earlier TinEye logo

In use by January 2025

Logos are trademarks of their owners and are shown only to identify the search engine this page discusses. Files are reproduced from Wikimedia Commons and the services' own sites. See the history of search engine logos.

Frequently asked questions

Does TinEye have its own index?

Yes. TinEye crawls the web itself and answers searches from its own collection of image signatures; it does not re-serve Google's or Bing's results. The index holds images only — no page text — so TinEye is a specialist image index rather than a web search engine in the usual sense. Its own counter showed more than 77.6 billion images in September 2025.

How many images does TinEye have?

More than 77.6 billion as of September 2025, according to TinEye's own counter as reported on Wikipedia. A secondary review dated 30 September 2026 reported more than 85 billion, but the live counter was not independently re-checked for this page in October 2026. Earlier dated figures are 2,013,406,246 images in September 2011 and more than 5 billion in June 2014.

Is TinEye better than Google Lens?

They do different jobs, so neither is better in general. TinEye finds copies of the same picture, including cropped, resized and edited ones, which suits tracing where an image has been used. Google Lens recognizes what is in a picture — objects, places, products — and returns visually similar images. TinEye does not recognize objects or faces at all, by design.

Does TinEye do facial recognition?

No. TinEye does not perform facial recognition, and it is not designed to identify people from photographs. It matches whole images against copies of the same image, which is a different task. That is a deliberate design choice and a clear contrast with face-search services such as PimEyes, which are built specifically to match faces.

Why does TinEye find no results for my image?

Usually because TinEye has never crawled a copy of it. TinEye states that it does not typically add social-media images to its index, so pictures that circulate mainly on social platforms often return nothing. It also returns copies rather than similar images, so a picture with no exact or edited duplicates on the crawled web will show no matches. An empty result does not prove an image is original.

Can TinEye tell me when a photo was first posted?

Only approximately. Sorting results by oldest shows the earliest copy TinEye has in its index, which is why fact-checkers use it to estimate an image's first appearance. But that date reflects when TinEye crawled the copy, not when the image was created or first published, so it can be years later than the true first appearance.

Who owns TinEye, and is it free?

TinEye is owned by Idée, Inc., a privately held company founded in Toronto in 1999 by Leila Boujnane and Paul Bloore. Consumer searching is free for non-commercial use, with optional registration that stores search history. The company earns its revenue from commercial products: the TinEye API, sold in prepaid search bundles, and image-matching tools such as MatchEngine and MulticolorEngine.

TinEye