Cloudflare AI Labyrinth: what it does and does it affect SEO
Cloudflare AI Labyrinth adds hidden nofollow links that trap bots ignoring crawl rules. Here is who it targets, how to turn it on or off and how to check its effect.
By Crawlito · 6 min read
In short
- AI Labyrinth injects invisible nofollow links into HTML; bots that follow them reach AI-generated decoy pages and are recorded as likely bad actors.
- Cloudflare documents that AI Labyrinth actions are not mitigations: it does not block or challenge the request, so your real pages still load.
- Cloudflare says bots that respect no-crawl instructions ignore the honeypot links, and that the decoy pages carry meta directives to keep them out of search indexes.
- Events appear in Security Analytics as AI Labyrinth Served and AI Labyrinth Crawls, which is how you check whether a crawler you care about followed a decoy link.
- Turning the feature off stops new link injection immediately, but links created earlier stay valid for a limited time and can keep generating events.
On this page
- What does AI Labyrinth actually do?
- Who does AI Labyrinth target?
- Does AI Labyrinth affect SEO or AI search visibility?
- How to turn AI Labyrinth on or off
- How to check whether AI Labyrinth is affecting crawlers
- When to use AI Labyrinth, and when not to
- Frequently asked questions
- Check what crawlers actually receive
Cloudflare AI Labyrinth is an opt-in bot protection feature that adds hidden links to your pages. Crawlers that ignore no-crawl instructions and follow those links end up in a maze of AI-generated decoy pages, and Cloudflare uses that behaviour to identify and fingerprint them. It does not block or challenge any request, and Cloudflare says it is designed to leave compliant crawlers and SEO unaffected.
For most site owners the useful questions are narrower: will it confuse Googlebot or AI search crawlers, where is the switch, and how do you tell whether it is doing anything. This post answers those from Cloudflare's own documentation, with the gaps marked.
What does AI Labyrinth actually do?
Cloudflare announced the feature in a blog post on 19 March 2025. The mechanism has three parts:
- Link injection. When the feature is on, Cloudflare adds honeypot links to HTML responses. The AI Labyrinth documentation says the links are invisible, carry
nofollow, and are only seen by bots. - Decoy pages. A bot that follows a link reaches a page generated with Workers AI. Cloudflare says the content is real and related to scientific facts, and is not taken from your site. Each decoy page includes meta directives intended to keep it out of search indexes.
- Fingerprinting. A human would not click an invisible link, so a client that follows several of them is treated as a likely bot. Cloudflare says it adds what it learns to its list of known bad actors and feeds it into its bot detection models, which benefits customers who block AI bots.
The point is to waste the time of scrapers that ignore crawling rules rather than to stop them at the door. A blocked scraper learns it was blocked; a scraper reading decoy pages does not.
Who does AI Labyrinth target?
Cloudflare's documentation says it targets AI crawlers that ignore crawling guidelines and scrape content without permission, and that bots which respect no-crawl instructions ignore the honeypot. The launch post adds that the links are shown to suspected AI scrapers while legitimate users and verified crawlers browse normally.
In practice that gives you two layers of protection for crawlers you want:
| Crawler | Expected interaction, per Cloudflare | What to verify |
|---|---|---|
| Googlebot, Bingbot | Verified crawlers browse normally; links carry nofollow | No AI Labyrinth Crawls events with their user agents |
| OAI-SearchBot, Claude-SearchBot, PerplexityBot | Compliant bots ignore the honeypot | Same check, filtered by user agent |
| Unverified scrapers that ignore robots.txt | The intended target | AI Labyrinth Crawls events from unknown clients |
| Human visitors | Links are invisible | No visible change to the page |
Cloudflare's documentation does not list exactly how it decides who is shown the links, and it does not describe the meta directives on decoy pages by name. Treat the table as the documented intent, and confirm it with your own analytics as described below.
Does AI Labyrinth affect SEO or AI search visibility?
According to Cloudflare, no. The documentation says the links do not affect SEO or your site's appearance, and the decoy pages are marked so search engines do not index them. Nothing about your real pages changes for human visitors: same URLs, same content, same status codes.
There are still three things worth checking on your own site, because they depend on your setup rather than on Cloudflare's design:
- Your audit or monitoring tools. If you run your own crawler, an uptime checker or an SEO tool through Cloudflare, check whether it reports unexpected new links. A tool that respects
nofollowshould not follow them. - Cached HTML. The feature modifies HTML responses. If you compare raw HTML across tools, a small difference in markup may come from injected links rather than from your application.
- Crawlers you want that are not verified. A newer or smaller AI search crawler that is not on Cloudflare's verified list could be classed as a suspected scraper. If it respects
nofollowit should still ignore the links; if your analytics show it following them, you will see it in the events.
AI Labyrinth is separate from the settings that actually block crawlers. A search crawler that cannot reach your pages is far more likely to be stopped by Bot Fight Mode, a WAF rule or the AI crawler access policies. Our Cloudflare AI crawler settings guide maps those settings one by one.
How to turn AI Labyrinth on or off
The documentation gives this path in the Cloudflare dashboard:
- Open your domain and go to Security, then Settings.
- Filter by Bot traffic.
- Open AI Labyrinth and switch it on or off.
The launch post says the feature is available on an opt-in basis to all customers, including the Free plan. Plan details can change, so check what your dashboard shows.
When you turn it off, Cloudflare says new link injection stops immediately, but links created earlier stay valid for a limited time and may keep generating events until they expire. If you switch it off and still see events for a while, that is expected.
How to check whether AI Labyrinth is affecting crawlers
Cloudflare records two event types under the AI Labyrinth service in Security Analytics and Security Events:
| Event | Meaning |
|---|---|
| AI Labyrinth Served | Honeypot links were injected into an HTML response |
| AI Labyrinth Crawls | A client followed a honeypot link |
The documentation is explicit that these actions are not mitigations, so a request logged as AI Labyrinth Crawls was served, not blocked.
To check a crawler you care about:
- Open Security Events and filter by the AI Labyrinth service.
- Filter to the AI Labyrinth Crawls action.
- Group or filter by user agent and look for the tokens you want to keep, such as
Googlebot,Bingbot,OAI-SearchBot,Claude-SearchBotorPerplexityBot. - For any hit, confirm whether the request really came from that operator. A scraper can claim any user agent; verifying real AI crawlers shows how to check with reverse DNS or published IP ranges.
A spoofed Googlebot user agent following decoy links is exactly what the feature is for. A verified Googlebot following them would be worth reporting to Cloudflare support.
You can also look at the injected markup yourself. Fetch a page and compare it with the HTML your origin sends:
curl -s https://www.example.com/guides/setup -o via-cloudflare.html
curl -s --resolve www.example.com:443:203.0.113.10 \
https://www.example.com/guides/setup -o from-origin.html
diff <(grep -o '<a [^>]*>' from-origin.html | sort) \
<(grep -o '<a [^>]*>' via-cloudflare.html | sort)Replace the IP address with your origin's. Cloudflare says the links are shown to suspected bots, so a normal request from your machine may show no difference at all. An empty diff does not mean the feature is off.
When to use AI Labyrinth, and when not to
AI Labyrinth suits sites that already allow the crawlers they want and are annoyed by scrapers that ignore robots.txt. It adds a cost for bad actors without adding friction for visitors, and it does not change any allow or block decision.
It is not a replacement for a policy. If you want to stop a named training crawler, a robots.txt rule or a block rule is the documented control; see how to allow AI search but block AI training. If the problem is server load, a rate limit for that crawler is more direct, as covered in rate limiting AI crawlers. And if you are deciding whether to charge crawlers instead, Cloudflare's pay per crawl works on a different principle entirely.
Frequently asked questions
Does AI Labyrinth block AI crawlers?
No. Cloudflare's documentation says AI Labyrinth actions are not mitigations and that it does not block or challenge the request. It wastes the time of scrapers that follow hidden links and records them for detection.
Can AI Labyrinth hurt my Google rankings?
Cloudflare says the links carry nofollow, do not affect SEO, and that decoy pages include meta directives to prevent indexing. You can confirm that Googlebot is not following the links by filtering AI Labyrinth Crawls events by user agent.
Is AI Labyrinth available on the Free plan?
Cloudflare's launch post in March 2025 said it is available on an opt-in basis to all customers, including the Free plan. Check your dashboard for the current state.
Why do I still see AI Labyrinth events after turning it off?
Cloudflare says links created before you turned the feature off stay valid for a limited time and may keep generating events until they expire.
Check what crawlers actually receive
AI Labyrinth rarely causes access problems, but the settings around it often do. The free check shows whether a page returns real content or a challenge to automated clients, and an audit reviews Cloudflare rules and robots.txt together across your site.
Filed under CDN and bot protection. Topics: cloudflare, bot-protection, ai-crawlers, waf