Facial recognition search engines like Google Lens and TinEye build a digital fingerprint of your image and match it across the web. NoScrape defeats this by stripping metadata, cropping edges, shifting colour channels and adding a tiled watermark—so the fingerprint no longer matches the original, making reverse-image matching fail. The process runs entirely in your browser; nothing is uploaded or stored.
What is facial recognition search and how does it actually work?
Facial recognition search is a subset of reverse-image search. Rather than matching text keywords, these tools analyse the visual content of an image—colours, shapes, faces, objects, text and spatial relationships—and create a mathematical fingerprint. Google Lens, TinEye and similar services use this fingerprint to find identical or near-identical copies across the internet.
For design professionals—interior designers, architects and specifiers—this is a real problem. A mood board shared with a client, uploaded to a portfolio site or posted on social media becomes searchable. A client can open Google Lens on their phone, photograph your mood board or reverse-search the file, and find the original furniture, fabric or finish supplier directly. They bypass the design studio, negotiate directly with the manufacturer and the commission disappears.
The fingerprinting process doesn't require faces alone. Modern reverse-image search engines analyse the entire visual composition: colour distribution, texture, object boundaries and even metadata embedded in the file itself (EXIF data, keywords, creation date). This metadata makes the image even easier to track and match.
Why standard image formats leave you vulnerable to reverse-image matching
When you export an image as JPG, PNG or other standard formats, the file carries invisible information. EXIF data records the camera model, GPS location, date taken and often the software used to edit it. Reverse-image search engines don't ignore this; they use it as part of the matching fingerprint. A JPG of your mood board is, to Google Lens, a highly specific visual and metadata signature.
Even without metadata, the image's visual fingerprint is stable. Small variations in brightness or slight crops might not fool the algorithm. Reverse-image search tools are built to find the same image even if it's been resized, rotated slightly or compressed. This robustness—which makes the tool useful for finding plagiarism—is exactly what makes protecting your own work difficult using ordinary export settings.
Standard sharing platforms (email, cloud storage, shared drives) preserve the original visual fingerprint intact. A JPG sent to a client is the same visual object that Google Lens will match if that client or anyone else later runs a search. For design studios, this means every mood board you share is a potential leak.
How does image protection defeat facial recognition and reverse-image search?
Image protection tools like NoScrape work by deliberately breaking the visual fingerprint and removing metadata—not to damage the image for human viewing, but to make it unmatchable to reverse-image search algorithms. The process happens in four coordinated steps. First, all embedded metadata (EXIF, keywords, colour profiles) is stripped away, eliminating the non-visual fingerprint component. Second, the edges of the image are cropped slightly, which changes the spatial relationships at the boundaries. Third, the colour channels are shifted—not obviously to the eye, but enough that the mathematical fingerprint of the colour distribution no longer matches the original. Fourth, a tiled watermark is applied across the entire image.
These changes work together. A reverse-image search engine builds its fingerprint from visual features and metadata combined. When metadata is removed, the channels are shifted and the crop is altered, the cumulative effect is that the fingerprint no longer matches the source image or any standard copy of it. Google Lens or TinEye searching for the modified image will not reliably find the original supplier's product page, because the image they're searching for is no longer the same visual object.
Critically, the process runs entirely in your browser using the Canvas API. Your original image is never uploaded, stored on a server or transmitted anywhere. The protection is applied locally, on your device, and you download the protected version. Nothing about your work leaves your control.
What does the protected image actually look like to a client?
A protected image remains clearly visible and functional as a mood board. The watermark is tiled across the image in a way that's visible (so the client knows it's protected) but doesn't obscure the design intent. The colour shift is calibrated to be imperceptible to the human eye—colours still look natural and accurate enough for design decision-making. The cropping is minimal and doesn't affect composition. A client receiving a protected mood board can still see and understand the design direction, fabrics, finishes and spatial relationships without difficulty.
The watermark serves a dual purpose. It communicates that the image is protected, discouraging casual misuse. It also becomes part of the visual fingerprint, further complicating any attempt to match the image to the original source. From a practical standpoint, a client looking at a protected mood board has all the information they need to make design choices with the studio—but less incentive and ability to search for alternatives independently.
When should you protect mood boards and when is it essential?
Protection is most valuable at the early stages of a project—when you're sharing mood boards, material selections, colour palettes and concept images with a client. These are the assets most at risk of being reverse-searched. Once a project is complete and publicly photographed (on a completed interior or a finished building), reverse-image search protection is less relevant, because the real-world object itself is now the source image.
Protection is essential when mood boards are shared digitally with clients before a commission is confirmed. It's equally important if mood boards are shown at pitch stage to multiple potential clients, or if you post design inspiration to your own portfolio or social media. The rule is simple: if the image contains inspiration or sourcing you'd prefer the client not to reverse-search independently, it should be protected.
Protection is less critical for images that are already widely available online (stock photography, published magazine spreads, images already on a supplier's website). Protecting an image that's already been indexed millions of times across the web won't prevent a client finding it through other means. The value lies in protecting your own original compositions, curated mood boards and sourcing decisions before they become public knowledge.
What doesn't image protection do—and what are its real limits?
Image protection defeats reverse-image matching by breaking the visual fingerprint. It does not provide legal copyright protection, register your work with a copyright office or offer a legal remedy against theft. If a client screenshots your protected mood board and manually recreates the look, or if they describe the colours and finishes to a different supplier, image protection cannot prevent that. Protection is a practical tool to stop algorithmic discovery, not a legal shield.
Image protection also cannot defeat every possible future version of search technology. Algorithms evolve. If a new reverse-image tool emerges that uses different visual feature extraction or is trained on modified images, it may eventually become capable of matching protected images to originals. NoScrape protects against current reverse-image search methods (Google Lens, TinEye and similar tools) by breaking the fingerprints these tools are known to use. It is not a permanent guarantee.
Finally, image protection does not hide an image if it's been widely shared already. If a mood board has been emailed to five clients unprotected and one of them reverse-searches it, protection applied later won't undo the copies already in the wild. The tool is most effective when applied from the moment the image is first shared.