AI Category Data for DNS Filtering Products

The Problem

The AI Category Gap in DNS Filtering

60–120 Categories Already Shipped Your DNS filtering product blocks malware, adult content, and gambling. These categories have been refined over a decade and cover the established web.

AI Tools — The Missing Category AI tools are the fastest-growing class of web applications in history. The category didn't exist at meaningful scale until 2023, and most vendors lack the classification infrastructure to populate it.

Your Customers Are Already Asking

School districts need to block AI chatbots that students use to generate homework

Enterprises need to prevent employees from pasting proprietary data into generative AI services

Healthcare organizations must ensure no unapproved AI tool processes patient information

Financial institutions need audit trails showing which AI tools traders and analysts access

Every use case requires the same foundation: a comprehensive, accurate, continuously-updated list of AI-tool domains mapped to functional categories your policy engine can act on.

Why Building In-House Takes Years

Massive Scope Tens of thousands of domains across dozens of functional categories — text generation, image synthesis, code assistants, voice cloning, autonomous agents, and more.

Constant Churn New tools launch daily. Existing tools rebrand, merge, or migrate to new domains. A pipeline accurate last month is already missing hundreds of entries.

We Already Built It Our pipeline processes 102 million domains daily with a continuously-verified taxonomy. Integrate ours and ship the AI-tools category within weeks, not years.

Market Demand Signals

78%

of enterprises want AI blocking

92%

of K-12 districts filter AI

#1

requested new category

0

competitors with 18 subcategories

Taxonomy

18-Category Taxonomy for Granular Policy Control

Single AI Category

The minimum viable product. Lets administrators block or allow all AI tools as a group — one toggle, no granularity.

18-Category Granularity

The competitive differentiator. Block AI chatbots in classrooms while allowing educational tutoring. Block code generators for finance while allowing them for engineering.

Stable Category Identifiers Categories are never renamed or restructured without a major version bump and a 90-day migration window. Your internal category mapping is durable.

~180 Subcategories Each of the 18 categories contains approximately 10 subcategories. Expose as much or as little granularity as your UX supports.

Flexible UX Exposure Some vendors surface all 18 categories in their admin console. Others expose a single "AI Tools" toggle and offer category-level control only via API or advanced settings.

Text & Language

Chatbots, writing assistants, summarizers, translation tools, grammar checkers, and content generators.

Image & Visual

Image generators, AI art platforms, photo editors, background removers, and visual design tools.

Code & Development

Code completion tools, AI pair programmers, automated testing platforms, and code review assistants.

Audio & Music

AI music generators, voice synthesizers, audio editing tools, podcast tools, and sound effect generators.

Video & Animation

AI video generators, deepfake tools, video editors, avatar creators, and animation platforms.

Search & Research

AI-powered search engines, research assistants, academic tools, and citation generators.

Business & Productivity

Meeting assistants, document automation, AI scheduling, CRM copilots, and workflow automation tools.

Data & Analytics

AI data analysis tools, visualization generators, predictive analytics, and business intelligence assistants.

Agents & Automation

Autonomous AI agents, browser automation, RPA with AI capabilities, and task orchestration platforms.

Education & Learning

AI tutoring platforms, homework helpers, study tools, language learning, and educational chatbots.

Marketing & Sales

AI copywriting tools, ad generators, email marketing assistants, SEO tools, and lead generation platforms.

Healthcare & Science

Medical AI assistants, drug discovery tools, diagnostic aids, clinical note generators, and bioinformatics platforms.

Models & Infrastructure

Model hosting platforms, inference APIs, training infrastructure, MLOps tools, and AI development frameworks.

Companions & Social

AI companion chatbots, virtual friends, AI dating assistants, social simulation, and character interaction platforms.

Design & Creative

AI-assisted design tools, logo generators, UI/UX assistants, 3D modeling, and creative suite integrations.

Security & Compliance

AI-powered security scanners, compliance automation, threat detection, and vulnerability assessment tools.

Translation & Localization

AI translation services, localization platforms, real-time interpretation, and multilingual content tools.

General & Multi-Purpose

Multi-modal AI platforms, general-purpose assistants, AI aggregators, and tools spanning multiple categories.

Delivered as a Stable JSON Tree Each category has a permanent string identifier (e.g., text-language, code-development) that never changes across feed versions.

Map Once, Never Break Your integration maps these identifiers to your internal category IDs once, and the mapping remains valid indefinitely.

Non-Breaking Updates When we add new subcategories — roughly quarterly as the AI landscape evolves — existing identifiers are never modified, so your mapping stays intact.

Classification

Domain Classification Pipeline

Every domain in the feed is processed through a multi-stage classification pipeline operating on our 102-million-domain corpus. The pipeline is designed to maximize precision above all else.

Precision-First Design A false positive blocks a legitimate website for every user on the network — unacceptable. A false negative (a missed AI tool) remains accessible until the next cycle catches it, typically within 24–48 hours.

1

Content Analysis Crawlers fetch the homepage, product pages, documentation, and metadata for every domain in the corpus. NLP classifiers identify domains describing AI-powered functionality — references to ML models, generative AI, NLP, computer vision, or other AI technologies.

2

Functional Verification Domains passing the initial screen are checked for concrete indicators of an active AI tool — API documentation, model selection interfaces, prompt input fields, inference endpoints, SaaS pricing patterns, and technology stack signatures. Domains passing both stages receive a category, subcategory, and confidence score.

3

48-Hour Probation Window Newly classified domains are flagged as "pending" and excluded from the production feed. Automated quality checks verify the classification isn't a false positive triggered by marketing copy, and confirm the domain resolves to an active service. Verified domains are promoted in the next delta update.

// Classified domain record schema (JSON)
{
  "domain":            "chatgpt.com",
  "primary_category":  "text-language",
  "subcategory":        "general-chatbots",
  "confidence":         0.99,
  "status":             "active",
  "first_seen":         "2024-01-15T00:00:00Z",
  "last_verified":      "2026-07-09T06:00:00Z",
  "risk_tier":          "high",
  "data_handling":      "cloud-processed",
  "languages":          ["en", "es", "fr", "de", "ja"],
  "related_domains":    ["openai.com", "chat.openai.com"],
  "tags":               ["genai", "llm", "user-data-upload"]
}

Risk Tiering by Data Handling

The risk_tier field reflects how the domain handles user data. This lets your customers build nuanced policies that flat blocklists cannot support.

Risk Tier	Data Handling	Example	Policy Recommendation
High	Accepts and cloud-processes arbitrary user input — text, files, images, or code	ChatGPT, Claude	Block for data exfiltration risk
Medium	Processes user data with limited scope (e.g., short text snippets only)	AI grammar checkers	Allow with monitoring
Low	Provides AI functionality without accepting sensitive user input	AI-powered search engines	Generally safe to allow

DNS Integration

RPZ Integration for DNS Resolvers

Response Policy Zones (RPZ) are the standard mechanism for injecting policy decisions into DNS resolution. If your product is built on a standard resolver, RPZ is the native integration path.

Supported Resolver Platforms

BIND — native RPZ support with automatic zone reload

Unbound — RPZ module with action-override and logging

PowerDNS — Lua-based RPZ primary with configurable refresh

CoreDNS — RPZ plugin with NXDOMAIN/NODATA/redirect actions

When a client queries a domain in the RPZ, your resolver returns a policy response — NXDOMAIN, NODATA, a redirect to a block page, or a passthrough with logging — instead of the actual DNS answer.

Zone File Delivery

HTTPS Download Standard RPZ zone file available for direct download. Conforms to the RPZ specification implemented by all major resolver software.

AXFR/IXFR Zone Transfer (Recommended) Your resolver subscribes to the RPZ zone and receives incremental updates as domains are added or removed. Changes apply in real time without reloading the full zone.

; RPZ zone file — AI tools category (excerpt)
; BIND, Unbound, PowerDNS, and CoreDNS all support this format

$ORIGIN ai-tools.rpz.aitoolsblocklist.com.
$TTL 300

; SOA and NS records (required for zone transfer)
@  IN  SOA  ns1.aitoolsblocklist.com. hostmaster.aitoolsblocklist.com. (
       2026070901  ; serial (YYYYMMDDNN)
       3600        ; refresh
       600         ; retry
       86400       ; expire
       60          ; minimum TTL
   )
@  IN  NS  ns1.aitoolsblocklist.com.

; Block AI chatbot domains (return NXDOMAIN)
chatgpt.com               CNAME  .
*.chatgpt.com             CNAME  .
claude.ai                 CNAME  .
*.claude.ai               CNAME  .
gemini.google.com         CNAME  .
perplexity.ai             CNAME  .
*.perplexity.ai           CNAME  .

; Redirect AI image tools to block page
midjourney.com            A      10.0.0.1
*.midjourney.com          A      10.0.0.1
leonardo.ai               A      10.0.0.1
*.leonardo.ai             A      10.0.0.1

; ... 16,024+ domains across 18 categories

Resolver Configuration Examples

Configuration snippets for the four most common resolver platforms. Your product's configuration interface generates these for end customers.

# BIND 9 — named.conf RPZ configuration
options {
    response-policy {
        zone "ai-tools.rpz.aitoolsblocklist.com"
            policy nxdomain
            min-update-interval 300;
    };
};

zone "ai-tools.rpz.aitoolsblocklist.com" {
    type secondary;
    primaries { 198.51.100.10; };
    file "rpz/ai-tools.db";
};

# ─────────────────────────────────────
# Unbound — unbound.conf RPZ module
rpz:
    name: "ai-tools.rpz.aitoolsblocklist.com"
    primary: 198.51.100.10
    zonefile: "/etc/unbound/rpz/ai-tools.zone"
    rpz-action-override: nxdomain
    rpz-log: yes
    rpz-log-name: "ai-tools-rpz"

# ─────────────────────────────────────
# PowerDNS Recursor — recursor.conf
lua-config-file=/etc/pdns-recursor/rpz.lua

# rpz.lua:
rpzPrimary("ai-tools.rpz.aitoolsblocklist.com", "198.51.100.10",
    { policyName="ai-tools", refresh=300 })

# ─────────────────────────────────────
# CoreDNS — Corefile RPZ plugin
. {
    rpz ai-tools.rpz.aitoolsblocklist.com {
        action nxdomain
        log
    }
    forward . 8.8.8.8 8.8.4.4
}

REST API Alternative For products using a proprietary resolver or custom fork, the REST API delivers the same domain data in JSON or CSV format. Your backend imports it into whatever data structure your resolver uses for policy lookups. RPZ is simply the most efficient path for standard open-source resolvers.

Confidence Scoring

Confidence Scores and Policy Thresholds

Every domain carries a confidence score between 0.0 and 1.0 representing the strength of evidence for its AI-tool classification.

Composite Measure, Not a Probability Derived from the number of classification signals detected, their consistency, and the domain's verification history across multiple crawl cycles.

High Score (0.98) Verified through multiple independent indicators — clear product documentation, active API endpoints, user-facing AI interfaces, and consistent classification across crawl cycles.

Lower Score (0.72) Fewer or less definitive indicators — perhaps the product is in beta, the domain is new, or it has limited web presence.

Threshold Tuning

The confidence score is the primary knob your customers use to tune coverage vs. accuracy.

We recommend exposing this as a configurable setting in your admin console with a default value of 0.80 — the sweet spot delivering comprehensive coverage with negligible false positive risk.

Conservative: 0.90+

Highest precision. Only domains with overwhelming evidence are blocked.

Recommended for financial trading floors, medical networks, and production infrastructure. Covers ~85% of known AI-tool domains.

Balanced: 0.80+ (Default)

Optimal trade-off with less than 0.1% false positive rate.

Recommended for most enterprise and education deployments. Covers ~94% of known AI-tool domains.

Aggressive: 0.65+

Maximum coverage including borderline domains with suggestive but inconclusive evidence.

Recommended for high-security environments that prefer to over-block and handle exceptions. Covers ~99% of known AI-tool domains.

Query-Time Filtering

Ingest Everything Your backend should ingest the entire feed — all domains regardless of confidence score — into your category database. The score is stored as a field on each record.

Filter at Query Time When your resolver evaluates a DNS query, it checks the stored confidence score against the administrator's configured threshold. No re-download needed when thresholds change.

Per-Policy Thresholds Different thresholds per policy — e.g., aggressive threshold for the student network, conservative threshold for the teacher network.

Speed

New Domain Detection and Classification Speed

The AI-tool landscape is not static. New tools launch daily — some on new domains, others as AI features added to existing platforms.

1

Daily Scanning (300K+ Domains) Our pipeline scans approximately 300,000 newly registered or newly active domains every 24 hours. Domains identified as potential AI tools enter classification immediately.

2

Content Analysis + Verification (24 Hours) Initial crawling and content analysis identifies and verifies AI tool functionality within the first 24 hours.

3

Probation + Production (48 Hours) The 48-hour probation window ensures classification accuracy before the domain reaches your customers' resolvers. Median time from first detection to feed inclusion: 72 hours total.

Accelerated Classification for High-Profile Launches

Major new AI products from well-known vendors, tools gaining viral traction, or AI services covered by tech press trigger accelerated classification.

Trend detection monitors domain registrations, social media mentions, news feeds, and app store listings. These domains enter the pipeline within hours — though the 48-hour probation window still applies. Result: 48–72 hours from public launch to feed inclusion.

Continuous Re-Verification

Every domain in the feed is re-verified at least once every 14 days

Lower-confidence domains are re-verified more frequently

Domains that have pivoted away from AI or become inactive are removed in the next delta update

Domains that add AI features after initial classification are detected and categorized

Detection Timeline

300K

Domains scanned daily

72hr

Median time to feed

48hr

High-profile launches

14 day

Full re-verification cycle

Licensing

Bulk Data Licensing for DNS Filtering Vendors

Our OEM license lets you embed AI-tools category data into your own product and redistribute it to all your end customers.

What the License Covers

Ingest our domain feed and store it in your category database

Serve policy decisions based on the data to every protected endpoint

No per-endpoint, per-query, or per-customer surcharge

Covers your entire installed base — resolvers, endpoints, and customer accounts

Licensing tiers are based on the features you need, not the volume you serve.

Standard

Full domain feed (18 categories)
Daily delta updates
REST API + RPZ zone transfer
JSON, CSV, RPZ formats
Email support
Unlimited redistribution

Premium

Everything in Standard
Real-time lookup API
Dedicated integration engineer
Custom category mapping
4-hour support SLA
Priority onboarding

Enterprise

Everything in Premium
Priority domain classification
Raw classification signals
Co-branded report templates
Custom SLA terms
Quarterly business reviews

Full Redistribution Rights

Product-Level License, Not Customer-Level Whether you serve 100 small businesses or 10,000 enterprise accounts, the license cost is the same. Your growth incentive is fully aligned with ours.

Aligned Incentives We succeed when your product succeeds. No surcharges that penalize your growth.

30-Day Technical Evaluation

1

Free Access — No Obligation Full API credentials, complete data dump, integration documentation, and access to our integration engineering team.

2

Integration (2–4 Weeks) Most vendors complete integration in 2–3 weeks via RPZ or 3–4 weeks via REST API.

3

Production Contract Choose the licensing tier that matches your needs. Guaranteed SLAs for feed uptime, update frequency, and support response times.

Ready to Add an AI Category to Your DNS Product?

Start with a free 30-day technical evaluation. Full API access, complete dataset, and a dedicated integration engineer — no commitment required.

Download Sample Data View Licensing

Request DNS Category Data Package

Tell us about your DNS filtering product and target market. We will provide API credentials, sample data, and integration documentation within 24 hours.

Related Resources

Palo Alto Integration FortiGate Integration REST API Documentation DNS Blocklist Guide Domain Feed Specifications Shadow AI Detection K-12 AI Filtering AI Data Loss Prevention Pricing Plans Contact Us

AI Category Data Built forDNS Filtering Products