Blog

Research & Updates

How we test our detectors, what we find in the wild, and what's new in Sightova.

Latest · Product

Introducing AI Video Detection: Check Any Video Frame by Frame

Sightova now detects AI-generated video. Upload a clip or paste a TikTok, Instagram, X or YouTube Shorts link, and our video model scores frames across the whole clip and returns a verdict with a timeline. Here is how it works, what a result looks like, and how it scores on public benchmarks.

November 14, 2026 · 6 min readRead article

Product

Our New Model Catches the Small AI Edits Behind Fake Damage Claims

On 9 November 2026 we put a new detection model into production, built for the photos people send as evidence. We broke a camera, a toaster, a lamp and a clock with ChatGPT and Gemini, stripped every watermark from some of them, and scanned them all. Every edit was flagged, and every real photo passed.

November 9, 2026 · 6 min read

Product

New in Sightova: Age Estimation, an Extension You Turn On in Settings

Sightova can now estimate the apparent age of each person in an image. It's off by default: switch it on under Settings → Extensions and every AI image check shows an estimated age and a likely range, in the dashboard and through the API.

October 29, 2026 · 4 min read

Research

AI Images in the Wild: 131 Documented Cases and What We Found Scanning 4,364 Images

How AI-generated pictures slip into the news, stock libraries and listings, how they get caught, and what happened when we ran our own checks over 4,364 public images, including rental photos that started carrying AI watermarks in 2026.

October 17, 2026 · 12 min read

Industry

The Dent That Wasn't There: AI-Edited Insurance Claim Photos

We asked GPT-Image 2.5 to dent the door of a real car photo. Sightova flagged the edit through OpenAI's invisible watermark and signed credentials. The pixel model pointed at the dent, but too weakly to decide on its own.

October 15, 2026 · 6 min read

Industry

"It Arrived Broken": How AI-Edited Photos Fuel Refund Fraud

We asked an AI editor to break a real teapot and a real mug in their photos. Sightova flagged both edits, but the invisible watermark and signed credentials carried the verdict, not the pixel model. Here is what that means for damage claims.

October 15, 2026 · 6 min read

Industry

Rental Scams With AI-Generated Homes: Why Listing Photos Aren't Proof Anymore

We generated a complete rental listing for a townhouse that doesn't exist. No reverse image search can find its photos, because they are new. Here is what Sightova's three checks found, and what renters and listing sites can do.

October 15, 2026 · 6 min read

Industry

Products That Don't Exist: AI-Generated Shops, Listings and Review Photos

A fake web shop used to need stolen product photos. Now one image model makes the catalogue, the lifestyle shots and the customer review photos. We made five with GPT-Image 2.5 and ran them through Sightova's checks.

October 15, 2026 · 6 min read

Industry

Inflated Invoices and Fake Medical Bills: One Added Line, Totals That Still Add Up

We added one line to a medical invoice with GPT-Image 2.5, and the edit fixed the subtotal and the total too, so the maths still checks out. Here is what Sightova's checks found, and what to ask for when the file alone can't settle it.

October 15, 2026 · 6 min read

Technology

Face Swaps Without a Real Victim: How Deepfakes Are Caught

We used GPT-Image 2.5 to put one synthetic man's face into another synthetic man's café photo, then scanned all three images. Here is what each of Sightova's checks found, and where face swaps get harder to catch.

October 15, 2026 · 6 min read

Research

We Tested Sightova Against GPT-Image 2.5 and Nano Banana Pro. Here's What It Found.

Thirteen fresh images from today's frontier image models, made after our October model shipped. Every one carried an invisible watermark and signed content credentials, and our model scored all of them between 94% and 96%.

October 15, 2026 · 6 min read

Research

One Number Changed: How AI Editors Forge Receipts, and How Sightova Catches the Edit

We asked an open-source AI image editor to change one amount on a real receipt. It took ten seconds and looked perfect. Here is what Sightova's document forensics model saw, down to the line that was rewritten.

October 15, 2026 · 6 min read

Product

NSFW Detection Without the False Alarms: Why Skin Isn't the Signal

Moderation that panics at swimwear, sportswear and wedding dresses punishes the users you want to keep. We ran five skin-heavy everyday photos through Sightova's NSFW model to show where they land.

October 15, 2026 · 5 min read

Launch

Launch

Introducing Sightova: A New Standard in AI Image Detection

Today we're launching Sightova — an AI image detection platform built from the ground up to identify synthetic media and show the evidence behind each result. Here's why we built it and what makes it different.

April 7, 2026 · 5 min read

Technology

Technology

Pixel Heatmap Analysis: How Sightova Shows Exactly Where AI Edited an Image

Binary 'real or fake' verdicts aren't always enough. Sightova's heatmap shows which regions of an image were most likely changed by AI — a critical tool for partial edits and deepfakes.

April 7, 2026 · 4 min read

Product

Product

Beyond Detection: How Sightova Combines Generator ID, Watermarks and Forensic Analysis in One Scan

A single AI image scan on Sightova combines our detection model, invisible-watermark checks, content credentials, generator identification and a manipulation heatmap. Here's how they fit together, and where NSFW checks come in.

April 7, 2026 · 4 min read

Research

Research

Watermark Forensics: How Sightova Detects Invisible AI Watermarks Hidden in Your Images

Google and OpenAI hide invisible watermarks in the pixels of the images their models generate — not in metadata, in the image itself. Every Sightova scan checks for them.

April 7, 2026 · 6 min read