AI driven tracker detection

AI-driven tracker detection

NextDNS already has AI-driven threat detection for identifying previously unknown malicious domains. Could the same concept be applied to ads and trackers?

Current DNS blocking is largely list-based, so a new tracking domain is allowed until someone discovers it and adds it to a list. NextDNS could use ML to identify likely trackers based on behavioral patterns, prevalence across unrelated sites/devices, CNAME/infrastructure relationships, similarity to known trackers, and other signals.

Rather than blindly blocking anything the model flags, it could work as a discovery pipeline: unknown domain → ML detection → validation → classification → dynamic blocklist. This is somewhat analogous to Privacy Badger's original learning approach, but NextDNS could potentially do it at much greater scale using anonymized aggregate DNS patterns.

An optional "AI Tracker Detection" setting—with report-only, high-confidence blocking, and aggressive modes—could potentially catch new tracking infrastructure well before it reaches conventional blocklists.