DeepSee.io - Publisher Intelligence

Most allowlists we audit have the same origin story. Someone exported a top-domains report years ago. A few sites got added because somebody asked. Then the file got copied into every campaign since. The problem is that the sites on it haven’t stood still. A domain is just a name. The business behind it can change completely while the URL stays the same. Sites get sold, and the new owner has different plans. Ad units get added until the page is more ads than article. Traffic gets bought to fill them. Writers get let go and the archive fills up with AI content. We flag thousands of new risk signals across domains each month. So a list that was solid at launch will have bad entries in it a few months later if nobody’s checking. We wrote up the full process, from SSP inventory reports to supply path checks to keeping the list current. Link in comments.

New DeepSee.io - Publisher Intelligence Updates: October Edition This month is all about apps. Here's what's new 👇 🔗 Network Tab for Apps App profiles now show the advertising identifiers we've seen in an app, and which other apps share them. Select any identifier to see every app using it and how many of those are high risk. 🔞 Clearer Adult Content Flags An 18+ age rating and explicit content aren't the same risk. We've split the adult signal in two for mobile and CTV apps, in both the portal and the API, so you can keep age-gated apps while excluding explicit ones. 📊 XLSX Exports With a Glossary Lists can now be exported as Excel files, with a glossary sheet that defines every signal and column. Share an export with a colleague or client without a separate explainer. 📱 New Research: Served, Not Seen We classified 212,344 utility apps across Google Play, Apple and Amazon. On Google Play, 90.8% of apps with no viewing session at all declare advertising, against 71.7% for the rest of the store. The less likely anyone is to see the ad, the more likely the app is to carry one. Check out our detailed October update linked in the comments! We'll be back next month with more.

Who's watching the ads on a lock screen app when the phone is in your pocket? Nobody (that's the whole problem). We sorted Google Play utility apps by whether a human is plausibly looking at the screen when an ad renders. The apps with no viewing session at all (lock screens, sleep sounds, live wallpapers) declare ad monetization at an enormously high rate: 90.8%. The less likely anyone is to see the ad, the more likely the app is to carry one. Wallpapers are the best example. The ad surface literally sits behind your icons. You can't optimize a wallpaper into the foreground of a foreground session, despite what IVT reporting might say. Some inventory just doesn’t make sense for ads. Exclude the category, don't bid “smarter�� into it. New report: “Served, Not Seen,” link in comments.

An app can sit at 4.6 stars and still have a review feed full of unclosable ads and forced installs. Our September update digs into that gap. The new Reviews tab shows what users are actually complaining about, backed by compliance tests for Android and iOS covering predatory loan offers, subscription dark patterns, undisclosed tracking, and more. Also shipping: an MCP server, so you can point an AI agent straight at DeepSee data without writing lines of integration code. Link to the full write up in the comments.

It will take years to fully untangle the psychological harm that AI companion & AI character apps are unleashing on vulnerable people (children and adults alike). Before it gets worse, let's come together to demand accountsbility by app marketplaces, and the platforms helping monetize them with programmatic ads.

New DeepSee.io - Publisher Intelligence Platform Updates: July Edition We're back with another set of updates. Here's what's new this month 👇 💳 Manage Your Plan & Usage New self-serve Plans & Usage screen brings your plan, seats, and usage together in one place. Track lookups across web, apps, and CTV, monitor list exports and API credits, and manage everything directly from the portal. 🤖 New AI Content Compliance Tests We've added tests that measure how much of a site's text and imagery is AI-generated, flagging sites accordingly with new risk signals. 📱 Improved App Risk Signals Sharper detection for AI companion apps and utility-style apps, helping you spot low-quality, non brand safe and potentially fraudulent inventory faster. 📊 New Research: App & CTV Inventory Fresh research on what app and CTV inventory really looks like, and how to evaluate it. More on the way. 🧹 Bug Fixes & Polish A round of fixes and small improvements across the platform for a smoother day-to-day experience. Check out our detailed July update linked in the comments! We'll be back next month with more.

If you’re buying on the open web right now, you’re probably asking some version of this question: What am I actually buying? Brand-safe inventory is scaling while fraud and AI slop run rampant. But reporting still looks fine. So how are you supposed to decide where spend goes? That’s why we created The Observed Web, a new DeepSee.io newsletter that tells you what’s really happening inside digital media supply. Every entry features insights, mini-analyses, and news roundups about supply health and quality. First up, we examine how AI-generated health content: Passes suitability checks Mimics medical authority Absorbs spend in high-trust environments ...while offering little accountability behind the scenes. If you care about signal quality and inventory integrity, and about protecting your performance over time, subscribe to The Observed Web today (link in comments)

Our CEO Rachmiel (Rocky) Moss publishing some interesting research on instream vs outstream video advertising