The prior art layer.
Feedback requested.

Priorly is computer vision infrastructure grounded in public IP registries — USPTO trademarks and patent figures. Two products, one shared technical foundation.

An API that sits at ad creative upload time. Computer vision extracts logo regions, wordmarks, and design elements from the uploaded image. Features are matched against embeddings of all 3.2M active USPTO trademark registrations using nearest-neighbor search. The API returns: flag status, USPTO registration number, trademark class, registrant name, and similarity score — in under two seconds.

The buyer is ad platforms (Meta, Google, DSPs) and agencies. Both have growing legal and compliance exposure from infringing creative running at machine speed. Existing brand safety tools don't touch trademark visual conflicts at all.

A figure-based prior art search tool for the applicant side. In October 2025, the USPTO launched DesignVision — an AI image search tool available only to examiners. Patent counsel and corporate IP teams are now being examined with a tool they cannot access themselves.

Priorly's patent product is the applicant-side equivalent: upload a patent figure, receive ranked visually similar patents across USPTO, EPO, and WIPO, with attention overlays and legal status for each match.

• Trademark infringement liability for ad platforms is increasing. Courts and regulators are scrutinizing ad delivery infrastructure more closely. Existing keyword-based tools don't catch visual conflicts.
• USPTO DesignVision (launched Oct 2025) created a real asymmetry: examiners have AI image search, applicants don't. Multiple IP firms have issued client alerts advising pre-filing visual clearance.
• The competitive gap is specific. On the trademark side: no product does proactive USPTO-grounded detection at upload time. On the patent side: no commercial product offers production-grade figure embedding search on the applicant side.
• The technical components exist. Vision models, vector embeddings, ANN search. What's been missing is someone who builds the pipeline against the public IP corpus specifically and wraps it in a clean API.

Your team processes ad creative at scale and manages the compliance and legal exposure that comes with it. I want to understand whether the visual trademark gap I've described is something your policy, legal, or ad ops teams have actually felt — or whether the current reactive workflow is considered sufficient. I'd also want to know where a tool like this would land: legal, policy, ad ops, or somewhere else.

Your work involves the kind of prior art searches where the figure layer matters — design patents, mechanical claims, FTO opinions for hardware or medtech companies. I want to understand whether the DesignVision asymmetry has registered as a real problem in your practice, whether pre-filing visual clearance is something clients ask about, and what the output would need to look like to be useful in a client deliverable.

Three specific questions. A reply or a 20-minute call is more than enough — I'm not asking for a commitment of any kind.

I'll share a working prototype and more detailed technical brief with anyone who wants to go deeper.