AI Tools Blocklist for Schools

The difference

What changes when AI blocking is in place

Two years ago, an IT director could name the AI sites students used on one hand. Today a new essay writer, homework solver or character-chat app can appear on Monday and be circulating through a middle school by Friday. Blocking these tools one URL at a time is a race the district always loses.

Without AI Blocking

Staff discover new AI tools only after students are already using them
Blocking one URL at a time — a race the district always loses
No visibility into which AI categories students are accessing
Student data pasted into ungoverned AI tools is gone for good
Essay writers, paraphrasers and solvers undermine academic integrity
Deepfake, voice-cloning and companion tools enable harassment
AI image generators with weak guardrails can produce harmful content on demand
Every AI pilot requires a manual exception — a steady stream of urgent tickets

With AI Blocking

Continuously maintained dataset: every known AI-tool domain, sorted into 18 categories and 165+ subcategories, refreshed every day
Team shifts from hunting domains to making policy — deciding which categories serve instruction
Category-level reporting turns vague worry into numbers a district can act on
Blocking un-reviewed AI categories at the moment of the paste, not after the harm
Essay mills stay blocked regardless while AI pilots run smoothly
Deepfake and face-swap is a standing category, refreshed daily, blocked district-wide
Explainable blocks: a named subcategory added on a known date for every block
Open one subcategory for one group without manual exceptions

AI rules you can enforce

Put the AI policy into the network itself

The hosted CIPA filter has an AI / LLM Tools category, so a district can block AI sites by switch, or allow the ones it has approved. Per-school settings let a high school differ from an elementary building.

Always-allow lists open the approved AI tools for a school.
New AI sites are caught through the category data, with no spreadsheet to keep.
The decision is by domain, so the filter does not read prompts.

Policy switches including the AI / LLM Tools category

11,800+

AI tools covered in our AI category

Milliseconds

response time for each DNS lookup

categories, one of them for AI

Look at our hosted filter for AI tools Talk to us about AI tools Look up a category for AI tools

The policy fork

Block all AI, or selectively allow?

Both are defensible starting points, and both are easy to implement from the same list. The difference shows up months later, in classrooms and in your ticket queue.

Consideration	Selectively Allow by Category	Block All AI Outright
Classroom AI pilots & AP courses	Open one subcategory for one group	Every pilot needs a manual exception
Academic-integrity categories	Essay mills stay blocked regardless	Blocked
Safety categories (deepfake, companions)	Blocked, with reporting per category	Blocked
Teacher unblock requests	Rare — policy already matches instruction	A steady stream, each one urgent
Students learning AI literacy	Possible under supervision	Pushed to personal phones, unfiltered
Simplicity on day one	Requires a short policy discussion first	One rule, done

Many districts start with block-all in September and move to selective allowance by spring, once the curriculum office decides where AI belongs. Because the same 18-category structure supports both, that migration is a policy edit — not a new procurement.

What's inside

The AI categories a K-12 policy actually needs

Eighteen top-level categories cover the AI landscape. These are the ones that show up in school incident reports — each one independently blockable, so allowing a supervised chatbot never means allowing a face-swap site.

General Chatbots

Conversational assistants students reach for first. Some districts permit them for supervised use; the subcategory structure lets you allow specific kinds while keeping the rest closed.

Essay Writers & Paraphrasers

Tools that draft assignments or rewrite AI output to slip past detectors. The core of the academic-integrity problem — covered in depth on our AI cheating tools page.

Homework & Code Solvers

Photo-a-problem math apps and code generators that hand over finished answers. Useful in a few electives, corrosive everywhere homework is meant to be practice.

Image Generators

Text-to-image services with wildly uneven content controls. Fine for a supervised art unit on some campuses, a source of inappropriate imagery on others.

Deepfake & Face-Swap (200+)

More than 200 domains that place a real person's face into fabricated photos or video. In a school, these are harassment tools before they are anything else.

Voice Cloning (250+)

Over 250 services that mimic a voice from a short sample — a teacher's, a classmate's, a parent's. A category with almost no legitimate K-12 use case.

AI Companions & Character Chat (470+)

The largest safety category: 470+ companion and roleplay chat domains built for emotional attachment, with age controls that range from weak to absent. See our student-safety AI filtering page.

Writing & Productivity Aids

Grammar helpers, summarizers and study aids that sit in the gray zone. Subcategories separate the tutoring-style tools from the do-it-for-me tools so your policy can too.

…and Ten More Categories

Video generators, audio tools, developer platforms, AI search and other clusters — 18 categories and 165+ subcategories in all, so no AI tool lands in your network uncategorized.

One List, Four Policy Conversations

Filtering AI is not a verdict against the technology. Plenty of districts run supervised AI pilots and still block the majority of these domains, because four distinct risks sit behind unmanaged access — and only one of them is about cheating.

The quietest risk is data. When a student pastes an essay about their family, their health or their address into a random AI site, that text leaves the district's control entirely, landing with an operator who never signed a student-data agreement. No acceptable-use policy retrieves it afterward.

Board asks: "Are the face-swap sites from the news blocked here?"
Yes — deepfake and face-swap is a standing category, refreshed daily, blocked district-wide.

The AI category, entry by entry

AI sites and their labels

Real lookups from our data. All of these land in the AI / LLM Tools category, so a single switch handles them together.

Website	Categories in our data
chatgpt.com	AI / LLM Tools
claude.ai	AI / LLM Tools, Web-based Applications
perplexity.ai	AI / LLM Tools
character.ai	AI / LLM Tools, Web-based Applications
midjourney.com	AI / LLM Tools
huggingface.co	AI / LLM Tools

Try a lookup on AI tools

Why districts filter AI at all

A single "artificial intelligence" category forces an all-or-nothing decision that satisfies nobody. The curriculum office wants a writing assistant open for one course; the principal wants roleplay chat gone everywhere; a flat bucket cannot honor both requests at once.

Subcategory depth is what makes the list a policy instrument instead of a blunt switch. "Paraphrasing tools," "AI companions," "photo-solve math apps" and "voice synthesis" are separate levers, so every allow or block maps to a decision an educator actually made.

Academic integrity — essay writers, paraphrasers and solvers undermine the work assignments exist to build
Student safety — deepfake, voice-cloning and companion tools enable harassment and unhealthy attachment
Student-data privacy — personal information pasted into ungoverned tools is gone for good
CIPA and E-Rate exposure — AI sites that generate explicit content fall squarely inside what a filter must stop

Kept current, every day

Built from the firehose of new domain registrations

Roughly 300,000 domains are registered every single day. We screen that stream daily for AI tools, so the list reflects what launched this week — not last semester.

~300kNew Domains Screened Daily

16,328+AI-Tool Domains Listed

165+Subcategories for Precision

24hUpdate Cycle

Behind the list

How the blocklist stays ahead of new AI tools

A blocklist is only as good as its maintenance. Ours runs on a daily pipeline, not a quarterly review.

Screen New Registrations

Each day's ~300,000 newly registered domains are screened for AI-tool signals, alongside continuous monitoring of app directories and launch channels where new tools first surface.

Classify Into Categories

Confirmed AI tools are assigned to one of 18 categories and the precise subcategory that describes what the tool does — an essay writer is never filed as a generic chatbot.

Review High-Risk Entries

Deepfake, voice-cloning and companion domains get extra scrutiny, since these are the entries schools depend on most and the ones bad actors rename most often.

Publish Daily Update

Additions and recategorizations ship every day in every delivery format, so the copy your firewall pulled this morning already knows about the tools that launched yesterday.

Age-appropriate rules

One list, different rules for a nine-year-old and a senior

The hardest part of school AI policy is that the right answer changes with age. A code assistant that is genuinely instructional in a high-school programming class is an answer machine in seventh-grade math. A flat, district-wide rule is guaranteed to be wrong for somebody.

Elementary (K-5)

All 18 AI categories blocked. Younger students have no instructional need that outweighs the exposure, and teachers demo approved tools from their own accounts.

Middle School (6-8)

Safety and integrity categories blocked; a narrow set of study aids allowed in supervised settings. This is the band where companion-chat curiosity peaks, so the 470+ companion domains stay firmly closed.

High School (9-12)

Approved chatbots and writing aids open for coursework; essay mills, paraphrasers, deepfake, voice-cloning and companion categories blocked. Seniors in a CS elective may get code-assistant access their peers do not.

Apply categories per grade band, building or user group
Grant course-level exceptions without opening a category district-wide
Tighten or relax a band's posture in one policy edit as your AI stance matures

Because every domain on the list carries a category and subcategory, the same dataset can enforce three different postures across three grade bands without anyone maintaining three lists. The policy layer decides; the blocklist just keeps the map of the AI landscape accurate underneath it.

Deployment

Formats that fit the stack you already run

No rip-and-replace required. The blocklist ships in the formats school networks actually consume, and the same categorized data drives every one of them.

CSV

The full categorized list as a flat file — ideal for audits, board reviews, or importing into any platform with a custom-list feature.

DNS Blocklist / RPZ

Response Policy Zone feeds for BIND and compatible resolvers, so AI domains are answered at the DNS layer before a connection ever forms.

EDL for Firewalls

External dynamic lists your firewall polls on a schedule — the daily update lands without anyone touching a rulebase.

PAC / Hosts & API

Proxy auto-config and hosts-file builds for endpoint-level enforcement, plus an API for querying categories programmatically or syncing custom tooling.

Districts running 1:1 programs typically pair the DNS/RPZ or PAC delivery with device-level policy so the database follows Chromebooks home — our Chromebook & 1:1 device filtering page covers that setup, and our malware & phishing protection page shows the same database delivery formats carrying security categories.

Beyond blocking

Visibility your board and auditors will actually ask for

Half the value of a categorized list is what it lets you see. Every block resolves to a named category, which turns vague worry about "students using AI" into numbers a district can act on.

Category-Level Reporting

See attempts against essay writers versus companions versus image generators, by building and by week — evidence for the board, the audit file and your own policy reviews.

Explainable Blocks

When a parent or teacher asks why a site was stopped, the answer is a named subcategory added on a known date — not "it was on a list somewhere."

Early Warning

A spike in attempts against a single new domain usually means a tool is spreading through a student body. Your team hears about it from the report, not from a discipline incident.

Compliance

Where AI filtering meets CIPA, E-Rate and student data

CIPA & E-Rate Obligations

CIPA obligates E-Rate-funded schools to run a technology protection measure against obscene material, child sexual abuse material and content harmful to minors, and to monitor minors' online activity
Generative AI complicates that certification in a way static websites never did: an image generator with weak guardrails can produce harmful-to-minors content on demand, on a domain no traditional category list has flagged
A categorized AI tools blocklist turns that question into a records exercise — show which AI categories are blocked, when each domain was added, and that the list updates daily
The same explainable, category-level evidence you already produce for adult content and gambling — our plain-English CIPA guide walks through the full obligation

Student-Data Privacy

Student-data laws and district DPAs assume the district knows where student information flows
Ungoverned AI tools break that assumption silently: a thirteen-year-old pasting a personal narrative into a free chatbot has just exported student data to a vendor nobody vetted
Blocking un-reviewed AI categories is currently the only enforcement mechanism that operates at the moment of the paste, rather than after the harm
Districts that put the blocklist in place before writing their formal AI policy consistently report the easier rollout, because enforcement questions are already answered by the time the committee meets

None of this requires hostility to AI. It requires the same posture districts already take toward every other classroom technology: reviewed tools in, unreviewed tools out, and a list that keeps up with the pace at which "out" multiplies.

Need the written rules behind the list? Our AI tools policy for schools gives every AI tool a K-12 decision with its reason, and a board-ready policy to match.

Precision

Why 165+ subcategories, not one big "AI" bucket

When your acceptable-use policy names a practice, there is a subcategory that enforces it. "Paraphrasing tools," "AI companions," "photo-solve math apps" and "voice synthesis" are separate levers, so every allow or block maps to a decision an educator actually made.

The blocklist is part of our web filtering database for schools and is included in every subscription. Districts integrate it into the firewalls, DNS resolvers and filtering infrastructure they already run — via API, CSV, DNS/RPZ, PAC, hosts files or firewall EDL — in whichever format their stack prefers.

Policies apply per user group, so staff can be exempted from student-facing rules entirely or given a wider allowance. Most districts block student access to unreviewed categories while leaving teacher accounts open to approved tools, then revisit each term as the curriculum office approves more.

Maintaining an AI blocklist yourself or using the category

Compare the work of keeping up with AI sites.

Task	Manual AI blocklist	Hosted CIPA filter
Find new AI sites	Watch the news and add	Covered by the AI / LLM Tools category
Allow an approved tool	Remove from your list	Per-site always-allow entry
Different rules by school	Several copies of a list	Separate settings per school
Apply an update	Reload and test	Live in seconds
Show the board the policy	A spreadsheet	Policy record and reports
Students away from school	Not covered	Private encrypted link keeps the same rules

See the hosted CIPA filter for AI tools Request a pilot for AI tools

Questions

AI blocklist questions we hear from districts

How many AI tools does the blocklist actually cover?

The list currently tracks 16,328+ AI-tool domains, organized into 18 categories and more than 165 subcategories. Coverage grows daily because we screen roughly 300,000 newly registered domains every day for AI-tool signals. The subcategory depth is what makes selective policies possible — you block "essay writers," not just "AI."

Can we allow ChatGPT-style chatbots but block essay-writing sites?

Yes — that is the most common configuration we see. General chatbots and essay writers are separate categories, so a district can permit supervised chatbot access for certain grade bands while keeping essay mills, paraphrasers and homework solvers blocked for everyone. The academic-integrity page covers how districts draw that line.

What formats can we integrate it into?

CSV for import and audit, DNS blocklist/RPZ for resolvers, EDL for firewalls that poll external lists, PAC and hosts files for endpoint enforcement, and an API for custom integrations. All formats carry the same daily-updated data, so mixing them — RPZ on campus, PAC on take-home devices — keeps policy consistent.

How fast do brand-new AI tools get added?

New domain registrations are screened daily, and confirmed tools ship in the next daily update — typically within a day or two of a tool becoming reachable. That matters most for the high-churn categories like deepfake and companion chat, where operators launch replacement domains quickly after takedowns or press attention.

Does this replace our web filter?

No, it extends it. Your existing filter handles the broad web — adult content, gambling, malware — while the AI blocklist adds depth in the one area growing too fast for general categorization. It is included with our school web filtering database, and you integrate it into whatever filtering, DNS or firewall stack you already run via the delivery format that fits your infrastructure.

Will blocking AI categories interfere with teachers using AI?

What happens when a blocked tool gets approved for classroom use?

You move it, you don't fight the list. An allow rule for that domain or subcategory takes precedence for the groups you choose, while the daily updates keep flowing underneath. Districts usually route approvals through their existing instructional-technology review, then reflect the decision as a one-line policy change.

Is an AI blocklist required for CIPA compliance?

CIPA does not name AI tools specifically — it requires blocking obscene material, child sexual abuse material and content harmful to minors, plus monitoring and student education. But AI image and companion sites can generate exactly the content those rules target, so districts increasingly treat AI coverage as part of demonstrating their filter works as certified. It strengthens the compliance story rather than being a separate mandate.