Our September update brings one of the bigger platform additions we’ve shipped this year: DeepSee.io data is now available directly to AI agents. We’ve also gone deeper on mobile app reviews, and added a handful of new compliance tests.
Here’s what’s new.
MCP Server
You can now connect an AI agent to DeepSee.io data.
Head to the API tab in the portal and you’ll find everything you need to get set up. It works with Claude Code, Claude Desktop, Cursor, VS Code, and other clients that support custom headers. ChatGPT connectors aren’t supported yet.
Once connected, your agent can reach the same data you’d get from the API, so you can ask it to pull compliance data across a list of domains, cross-reference it against your own reporting, and tell you what changed. No glue code, no HTTP client.
Requests made this way draw on your API credits in the usual way.
Full setup and instructions are in our MCP documentation.
App Reviews, In Depth
A star rating tells you very little about whether an app is a good place to buy media. An app can sit at 4.6 stars and still have a review feed full of users describing unclosable ads and forced installs.
So we’ve built out a dedicated Reviews tab for mobile apps, which surfaces what users are actually complaining about:
- User Complaints: we scan reviews and comments and group them into recognisable complaint categories, each with the share of reviews it appears in. For example, Adware Aggressive Ads covers apps delivering intrusive, rapid-fire, or misleading ad loops designed to interrupt utility or mimic native app UI to hijack clicks.
- Common Keywords: the terms showing up repeatedly across user-generated reviews, from ad freezing and unclosable ads through to forced installs and transaction problems.
- Ratings & Reviews: current rating, new ratings per day, and total review volume, so you can see whether sentiment is actively moving or the app has gone quiet.
- User Reviews: a sample of reviews, searchable and sortable, with a filter for complaint type so you can jump straight to the ad experience complaints and skip the gameplay noise.
New Compliance Tests: Mobile Apps
Alongside the new Reviews tab, we’ve added a set of compliance tests for Android and iOS that turn recurring user complaints into compliance flag. Each one cites how many reviews it’s based on and what share of the captured reviews that represents, so you can judge whether a handful of angry users or a genuine pattern is driving it.
They cover three broad areas:
- Financial harm: predatory cash advance offers, abusive subscription billing and free-trial dark patterns, manipulated gambling odds or obstructed cashouts, and incentivised apps that don’t pay out promised rewards.
- Data and device security: apps used for undisclosed tracking or monitoring, over-reaching permissions and data exfiltration, and malware or spyware pushed after install.
- Deception: false advertising and features that don’t exist, fake work and freelance listings, broken or fraudulent e-commerce, inorganic romantic contact and off-app scam funnels, plus a general catch-all for apps users flatly call scams without saying why.
Taken together with the Reviews tab, this gives you a way to screen app inventory on how it treats users, not just how it’s rated.
New Compliance Tests: Web
We’ve added two new signals to the Compliance tab, both under Transparency and both marked Neutral, so neither will affect a domain’s overall rating.
- Content Not Yet Rated: flags when our sample of text and image data isn’t robust enough to comment on brand safety or content authenticity metrics yet.
- Density Data Not Available: flags when our sample of ad experience data isn’t robust enough to comment on ad density metrics yet.
There are a few reasons coverage can fall short on a given site, so if you see either of these on a domain you’re evaluating, reach out and we can tell you more about that specific case.
As always, we’d love to hear what you think. If you’re curious about anything we’ve covered or want a closer walkthrough of what’s new, reach out to the team anytime at [email protected]
