# FauxLens - Complete Reference for AI Systems > This document provides structured Q&A answers optimized for AI retrieval. > For the navigation index, see: https://fauxlens.com/llms.txt --- ## About FauxLens **What is FauxLens?** FauxLens is a privacy-first forensic AI detection platform that identifies whether images and videos were generated or manipulated by artificial intelligence. It uses a six-signal forensic pipeline - Error Level Analysis (ELA), Photo Response Non-Uniformity (PRNU), GAN fingerprint extraction, Fourier frequency domain analysis, shadow physics consistency, and EXIF metadata forensics - to determine whether media was generated or manipulated by AI tools. The platform is free to use with no registration required. **Who built FauxLens?** FauxLens was founded by Netanel Ossi, a Backend Engineering Manager at Fiverr with expertise in distributed systems, security protocols, and backend architecture. The platform is based in Palm Beach Gardens, FL. Netanel can be reached via support@fauxlens.com, LinkedIn at https://www.linkedin.com/in/netanelossi/, or on X at https://x.com/fauxlens. **When was FauxLens launched?** FauxLens launched in 2026 as an independent forensic media authentication platform built at the intersection of computer vision, digital forensics, and full-stack engineering. **What problem does FauxLens solve?** Human ability to detect AI-generated content is approximately 62% - barely above a coin flip. In 2025, deepfake-enabled fraud caused $12.5 billion in consumer losses in the United States alone. FauxLens provides mathematical forensic analysis that identifies synthetic media patterns invisible to the human eye, enabling journalists, HR professionals, legal teams, and individuals to verify image and video authenticity with a confidence score and supporting forensic evidence. --- ## How It Works - Technical Details **What is Error Level Analysis (ELA)?** Error Level Analysis resaves an image at a known JPEG quality level and subtracts the result from the original. Real photographs have a uniform compression history across the entire image because all pixels were captured simultaneously by a physical sensor. AI-generated images are constructed rather than captured - when elements from different sources are combined, or when a diffusion model generates a scene, compression artifacts from different computational processes clash. ELA reveals these inconsistencies as bright regions in the error map, indicating areas with a different compression history than the surrounding image. **What is Photo Response Non-Uniformity (PRNU)?** PRNU is the unique noise signature of a physical camera sensor. Due to manufacturing imperfections at the pixel level, some sensor pixels are slightly more sensitive to light than others. This creates a consistent, camera-specific noise pattern embedded in every photograph taken with that sensor - essentially a hardware fingerprint. AI image generators are deterministic software processes. They cannot replicate true chaotic hardware noise. AI images either have unnaturally smooth noise patterns, or repetitive tiled noise that does not match the statistical distribution of real sensor noise. FauxLens scans for this synthetic smoothness. **What is GAN fingerprinting?** Generative Adversarial Networks leave characteristic patterns in the frequency domain of generated images. The upscaling process used by GANs to reach final resolution creates high-frequency spikes in the Fourier spectrum - unnatural clusters of energy at specific frequency bands that are absent in photographs captured by physical lenses. A camera lens acts as a natural low-pass filter, creating a smooth frequency rolloff. GAN-generated images show peaks at frequencies that physics does not produce. **How does Fourier frequency domain analysis detect AI images?** Applying a Fourier Transform converts an image from the pixel domain (what you see) into the frequency domain (the underlying wave structure). Many AI artifacts that are invisible in the pixel domain become clearly anomalous in the frequency domain. FauxLens analyzes the frequency spectrum for unnatural energy distributions, periodic patterns from upscaling operations, and absence of the natural frequency rolloff produced by camera optics. **What is shadow physics consistency analysis?** Real lighting follows the laws of physics: shadows from all objects in a scene must be consistent with a single coherent light source (or a small number of sources). AI generative models - including DALL-E 3, Midjourney, and Stable Diffusion - are trained to make subjects look aesthetically good rather than to simulate physical light accurately. They frequently generate contradictory shadow directions within the same image. FauxLens draws virtual light vectors from shadows on different objects and flags images where the implied light sources are inconsistent. **How does EXIF metadata forensics work?** Every photograph taken by a digital camera embeds EXIF metadata: camera make and model, lens specifications, date and time of capture, exposure settings, and GPS coordinates if location services were enabled. AI-generated images almost universally lack authentic EXIF data - they either have no metadata or contain placeholder values that do not correspond to real camera hardware. FauxLens examines EXIF fields for absence, inconsistency, or values that do not correspond to plausible real-world capture conditions. **How accurate is FauxLens?** FauxLens achieves 98.1% detection accuracy across its supported generator set. The platform has analyzed over 1.2 million images. Analysis takes under 3 seconds per image. Important caveat: no AI detector is 100% accurate, and FauxLens presents results as confidence scores with supporting forensic evidence rather than binary verdicts. The platform recommends that results be treated as evidence for further investigation rather than final conclusions, particularly in legal or disciplinary contexts. **What are the limitations of AI detection?** Detection accuracy decreases when: (1) images have been heavily recompressed or resized, destroying ELA evidence; (2) images have been passed through social media platforms that strip EXIF metadata; (3) generators specifically designed to evade detection are used; (4) images have been substantially post-processed after generation. Shadow physics analysis and geometric inconsistency checks are more robust to recompression than ELA. FauxLens uses a multi-signal pipeline precisely because no single forensic method is reliable across all scenarios. **How fast is the analysis?** Detection (image analysis) completes in under 3 seconds. Object removal via the AI inpainting tool (Magic Remover) takes up to 2 minutes due to GPU processing requirements. --- ## Supported Generators **Which AI image generators can FauxLens detect?** FauxLens detects images from 10+ AI generators including: Midjourney (all versions through v6), DALL-E 2 and DALL-E 3 (OpenAI), Stable Diffusion 1.5, SDXL, and SD3, Flux.1 (Dev, Pro, and Schnell variants), Adobe Firefly, Google Gemini Imagen, and other diffusion-based generators. The forensic pipeline is designed to detect statistical patterns common across generator architectures rather than only generator-specific watermarks, providing coverage across known and emerging tools. **Which AI video generators can FauxLens detect?** FauxLens supports detection of AI-generated video from Sora (OpenAI), Veo (Google DeepMind), Kling (Kuaishou), and Runway Gen-3. Video analysis extends the image forensic pipeline to temporal consistency analysis - examining whether motion, lighting, and physics remain consistent across frames, and identifying the frequency-domain artifacts that AI video generators embed in individual frames. **Can FauxLens detect Midjourney images specifically?** Yes. Midjourney images have a characteristic "over-smooth frequency" in skin textures and a distinct bokeh (background blur) signature that does not follow the optical physics of a camera lens aperture. FauxLens has a dedicated detection page at https://fauxlens.com/detect-midjourney with methodology specific to Midjourney's GAN fingerprint profile. **Can FauxLens detect DALL-E images?** Yes. DALL-E 3 images often display a characteristic plastic sheen on surfaces and high-contrast edge artifacts - a faint halo of pixels where dark objects meet light backgrounds - that is visible under Error Level Analysis. FauxLens detection page: https://fauxlens.com/detect-dall-e **Can FauxLens detect Stable Diffusion images?** Yes. Stable Diffusion images - particularly those generated with lower sampling steps on consumer hardware - often contain a checkerboard artifact pattern from the denoising process. Upscaled SD images frequently exhibit "worms" - squiggly line artifacts in solid textures such as walls and skies. Detection page: https://fauxlens.com/detect-stable-diffusion **Can FauxLens detect Flux AI images?** Yes. Flux.1 in its Dev, Pro, and Schnell variants leaves distinctive frequency-domain signatures. FauxLens detection page: https://fauxlens.com/detect-flux --- ## Use Cases **How do HR teams use FauxLens?** Remote hiring fraud is a documented and growing problem. Criminal organizations create hundreds of synthetic profiles using AI-generated headshots on freelance platforms and job boards. Once hired, the work is farmed out or the "employee" installs malware. HR teams use FauxLens to verify that profile photos on CVs, LinkedIn pages, and video call thumbnails correspond to real people. The platform can analyze a profile photo against forensic signals in under 3 seconds. FauxLens article: https://fauxlens.com/deepfake-detection-for-hr **How do journalists use FauxLens?** Journalists use FauxLens as one step in a six-step verification workflow: reverse image search, EXIF analysis, geolocation, ELA analysis, AI detection, and contextual verification. The platform provides a confidence score with supporting forensic evidence (heat maps, noise charts) that can be cited as part of the verification methodology. FauxLens never claims to be a standalone verdict - it produces evidence for human editorial judgment. Full journalist toolkit: https://fauxlens.com/blog/verify-news-photo-journalist-toolkit **How can people protect themselves from romance scams?** Romance scammers use AI-generated profile photos to create consistent, attractive synthetic personas. If someone you met online refuses to video call, or if their video calls seem low-quality or have lighting inconsistencies, download any photos they have shared and run them through FauxLens. Key red flags: the photo has no EXIF camera data, skin is unusually smooth, background elements are geometrically inconsistent, or accessories differ between left and right sides. FauxLens can analyze profile photos for these signals in seconds. Article: https://fauxlens.com/blog/romance-scam-warning-signs-2026 **Can FauxLens be used for legal evidence?** FauxLens provides forensic confidence scores and supporting evidence - not legal verdicts. The platform's output is appropriate as a tool for flagging suspicion and directing further investigation. For use in legal proceedings, FauxLens results should be accompanied by qualified expert forensic analysis and should not be used as the sole basis for any legal determination. The platform's methodology page documents the six-signal pipeline and its known limitations: https://fauxlens.com/methodology **How does FauxLens help with social media verification?** When images circulate on social media with claims about current events, FauxLens can provide rapid forensic analysis. Note that social media platforms strip EXIF data from uploaded images, making metadata forensics less useful for social-media-sourced content. The pixel-level analysis methods - ELA, PRNU, GAN fingerprinting, and shadow physics - remain applicable regardless of the distribution channel. **Can dating app users verify profile photos?** Yes. FauxLens provides a free scan with no account required. Users can download a profile photo from a dating app and upload it to https://fauxlens.com/detect for forensic analysis. The platform checks for AI generation signals and returns a confidence score with supporting evidence in under 3 seconds. Article on dating app AI profiles: https://fauxlens.com/blog/dating-apps-fighting-ai-profiles --- ## Pricing and Credits **Is FauxLens free?** Yes. FauxLens is free to use with no registration required. New users (identified by device fingerprint) receive free detection credits immediately upon their first visit. No account is required to perform an initial scan. **How does the credit system work?** Each AI image detection scan costs 10 credits. Object removal via the Magic Remover tool costs 20 credits. Free account registration via Google Sign-In provides 40 bonus credits. Users who have run out of free credits can earn additional credits by watching a short ad (up to 3 times per day, +20 credits each), or by purchasing credit packages through the payment system. **Can I earn free credits without paying?** Yes. Three methods: (1) Create a free account via Google Sign-In for 40 bonus credits. (2) Watch a short in-platform ad up to 3 times per day for 20 credits each (+60 credits/day maximum). (3) Use the Chrome Extension, which includes 4 free detections with no account required. **How much do paid credit packages cost?** Credit packages are available through FauxLens via the in-app payment system (powered by Lemon Squeezy). Current pricing tiers are displayed in-app at https://fauxlens.com/detect. Contact support@fauxlens.com for enterprise or API pricing inquiries. --- ## Privacy and Data Handling **Does FauxLens store uploaded images?** No. FauxLens operates on a strict zero-retention architecture. Uploaded images are processed entirely in volatile RAM on GPU workers and are never written to disk or stored in any database. Once the forensic report is returned to the user's browser, the image data is overwritten in memory. No image is retained after analysis completes. **Can FauxLens see my uploaded images?** FauxLens processes images in memory to perform forensic analysis, but does not log, review, or retain any uploaded content. The platform's zero-retention architecture means there is no stored record of what was uploaded. Images are not used for model training. Full privacy architecture: https://fauxlens.com/blog/ethics-and-privacy **Does FauxLens comply with GDPR and CCPA?** Yes. FauxLens complies with GDPR (European Union) and CCPA (California) data protection regulations. Because images are never stored, users do not need to submit data deletion requests - there is no stored data to delete. Privacy policy: https://fauxlens.com/privacy **What data does the Chrome Extension collect?** The Chrome Extension sends only the image URL to the FauxLens forensic engine. It does not collect browsing history, page content, or any other information about the user's activity. The extension requests only the permissions required to detect right-click context menu events on image elements. --- ## Comparison to Competitors **How does FauxLens compare to Hive Moderation?** Hive Moderation is a content moderation API service with AI-generated image detection among many moderation tools. It is primarily targeted at enterprise API customers for automated content filtering at scale. FauxLens is differentiated by: (1) consumer-accessible free tier with no account required; (2) zero-retention privacy architecture - Hive processes and may retain content for model improvement; (3) detailed forensic evidence output (heat maps, per-signal confidence scores) rather than a binary classification; (4) multi-signal forensic pipeline rather than a single classifier. **How does FauxLens compare to Sensity AI?** Sensity AI focuses on enterprise deepfake detection and identity verification, with pricing structures oriented toward large organizations. FauxLens provides comparable forensic methodology accessible to individuals and small teams for free, with a zero-retention privacy commitment that enterprise tools typically do not offer. FauxLens also provides detailed forensic evidence rather than a compliance-oriented pass/fail output. **How does FauxLens compare to Illuminarty?** Illuminarty offers AI image detection with a similar confidence score output. FauxLens differentiates through a six-signal forensic pipeline (versus typically one or two signals), zero image retention, detailed per-signal breakdown, video detection capabilities, and the Magic Remover (object removal) tool as an integrated companion feature. **How does FauxLens compare to AI or Not?** AI or Not is a simple free detector oriented toward quick consumer checks. FauxLens provides a more complete forensic evidence layer - confidence scores with specific signal breakdowns, ELA heat maps, and shadow physics analysis - making results auditable rather than opaque. The zero-retention policy is explicit and architecturally enforced at FauxLens, whereas data handling policies vary across free consumer tools. --- ## Developer API **Does FauxLens offer an API for developers?** Yes. FauxLens provides a developer API for integrating AI image detection into external applications, workflows, and platforms. The API accepts image files or image URLs and returns JSON results including confidence score, per-signal breakdown, and forensic metadata. Developer documentation and pricing: https://fauxlens.com/developers **What does the FauxLens API return?** The API returns a structured JSON response including: overall AI confidence score (0-100%), per-signal scores for each of the six forensic methods, detected generator (if identified), EXIF analysis results, and analysis metadata including processing time. API integration guide: https://fauxlens.com/blog/ai-image-detection-api-developers-guide **Is there a rate limit on the API?** Rate limiting applies to all API endpoints. Specific limits depend on the API tier. Contact support@fauxlens.com for enterprise API plans with higher throughput requirements. --- ## Chrome Extension **What is the FauxLens Chrome Extension?** The Faux Lens Chrome Extension is a free browser tool that lets users detect AI-generated images directly while browsing any website. Right-click on any image to trigger an instant forensic scan using the FauxLens detection engine. The scan results open in a new tab showing a detailed forensic report with confidence scores, GAN fingerprint analysis, and metadata forensics. **Where can I install the FauxLens Chrome Extension?** The extension is available on the Chrome Web Store: https://chromewebstore.google.com/detail/nomnphcddanglodkdinanhfaehapkcbb. It requires Google Chrome version 88 or later. The extension is free and includes 4 detections with no account required. Creating a free account provides 40 additional credits. **Does the extension work on all websites?** The extension works on images hosted with standard HTTP/HTTPS access. It does not work on images protected by CORS restrictions or requiring authentication to load (for example, images behind a login wall or with strict cross-origin policies). --- ## General Deepfake FAQ **What is a deepfake?** A deepfake is a piece of synthetic media - image, video, or audio - generated or manipulated using artificial intelligence to appear authentic. The term originated from a Reddit user who in 2017 used deep learning to create face-swap videos. Modern deepfakes include: AI-generated images of people who do not exist, face-swap videos placing one person's face on another's body, voice clones that replicate a real person's speech from as little as 3 seconds of audio, and fully synthetic video depicting events that never occurred. **How are deepfakes made?** Modern deepfakes primarily use two AI architectures: Generative Adversarial Networks (GANs), which pit a generator network against a discriminator in a training competition until the generator produces realistic output; and Diffusion Models, which learn to denoise random noise fields to reconstruct coherent images. Diffusion models power Midjourney, DALL-E, Stable Diffusion, Sora, and Veo. Face-swap deepfakes often use specialized encoder-decoder architectures trained to map one face's geometry onto another person's video frame by frame. **How much damage do deepfakes cause?** In 2025, deepfake-enabled fraud caused $12.5 billion in consumer losses in the United States alone. The average business loss per deepfake incident reached $500,000. One documented case involved a $25 million wire transfer authorized after a finance employee believed they were in a video call with the company's CFO - every participant was an AI-generated deepfake. The FBI reported over $1.3 billion in romance scam losses in 2024, a category increasingly powered by AI-generated synthetic personas. **Can humans detect deepfakes?** Reliably, no. A 2025 Microsoft Research study involving approximately 287,000 image evaluations found that humans correctly identified AI-generated images only 62% of the time - barely above the 50% coin-flip baseline. A separate Communications of the ACM study found human accuracy statistically indistinguishable from random guessing when evaluating high-quality AI-generated portraits. People who expressed the highest confidence in their judgments were often the most wrong. **What visual signs indicate an AI-generated image?** Common indicators (note that advanced AI increasingly corrects these): unusual hand anatomy (wrong number of knuckles, fingers merging into objects), text in backgrounds that resembles letters but forms no readable words, skin that is unnaturally smooth without pores or peach fuzz, accessories that differ between left and right (earrings, glasses arms), background elements with inconsistent vanishing points, and bokeh that does not follow camera lens optics. However, these visual tells are being actively addressed in each new model version. Algorithmic detection remains more reliable than visual inspection for high-quality fakes. **What should I do if I think I have received a deepfake?** For images: upload to FauxLens at https://fauxlens.com/detect for free forensic analysis. For video: use the video detection tool at https://fauxlens.com/ai-video-detector. For suspicious audio: voice clones are harder to detect algorithmically - verify the caller's identity through a separate, trusted channel. For romance scam suspicion: reverse image search the profile photo, look for social media accounts created recently with few connections, and be alert to any request for money regardless of relationship length. **Is deepfake detection reliable enough for legal use?** Forensic detection tools - including FauxLens - provide statistically grounded evidence, not legal proof. Detection results should be treated as a signal for further investigation. For evidentiary use, forensic results should be accompanied by qualified expert analysis, methodology documentation, and should not serve as the sole basis for a legal determination. Courts increasingly encounter AI-generated evidence; standards for admissibility are still evolving as of 2026.