noai and noimageai meta tags: do AI crawlers respect them?
The noai and noimageai values are not part of any standard, and OpenAI, Anthropic and Perplexity do not document them. Here is what to use for a real opt-out.
By Crawlito · 7 min read
In short
- noai and noimageai are not defined in RFC 9309 or in Google's documented robots meta tag rules, and OpenAI's, Anthropic's and Perplexity's crawler pages do not mention them.
- DeviantArt introduced both values in November 2022 as a signal that content should not be used in third-party AI datasets, and its announcement names no AI company that honors them.
- OpenAI, Anthropic and Google document robots.txt tokens (GPTBot, ClaudeBot, Google-Extended) as the way to opt out of training, so those are the controls to rely on.
- Adding the tags may be read by tools that choose to honor them, but Google documents no effect from them on search visibility or snippets, and they do not change what a crawler can fetch.
- A robots.txt rule is followed by crawlers that honor it and a server rule is enforced by your server; a noai tag is a public statement of intent that nothing forces a crawler to follow.
On this page
No major AI operator documents support for the noai and noimageai values at the time of writing. They are not part of the robots exclusion standard or Google's documented robots meta tag rules, and the crawler pages from OpenAI, Anthropic and Perplexity do not mention them. You can add them as a statement of intent, but if you want an opt-out that an operator has actually described, use the named robots.txt tokens instead.
This guide covers where the two values came from, what the official documentation says, how to send them if you still want to, and what to put in place for a real opt-out. Facts come from the operators' own pages, RFC 9309 and the tools named below, read on 11 November 2026.
What are the noai and noimageai values?
They are two keywords meant to sit in the same places as noindex: a robots meta tag in the page head, or an X-Robots-Tag HTTP response header.
<meta name="robots" content="noai, noimageai">X-Robots-Tag: noai, noimageaiThe values come from DeviantArt. In a November 2022 announcement, it described the directives as signals that content should not be used as input to third-party AI datasets, and it published both the meta tag form and the HTTP header form. On its own platform, DeviantArt said it places a tag with noimageai on the page and sends a header with noai when an image is downloaded directly from its servers. It also said third parties that keep using its content to train machine-learning models must ensure their training data excludes anything carrying either directive, and that noai is available for use by all.
Note what the announcement does not contain: it names no AI company, dataset or model that agreed to honor the values, and it describes no way to check compliance. That is where the confusion can start. The directive is a request from one platform, not a rule that other parties have agreed to follow.
Are noai and noimageai part of any standard?
No. Two documents define what crawlers are expected to read, and neither includes them.
- RFC 9309. The Robots Exclusion Protocol defines the
/robots.txtfile: groups,User-agent,AllowandDisallowlines, and how crawlers handle errors. It does not mention meta tags,X-Robots-Tag,noindexornoai. It also says its rules are "not a form of access authorization", so even the real standard is a request that compliant crawlers follow, not a lock. - Google's robots meta tag rules. Google's robots meta tag documentation lists the rules it supports for the tag and header, including
noindex,nofollow,nosnippet,max-snippetandunavailable_after. Neithernoainornoimageaiis among them.
Google's list is a widely used written reference for meta robots values, and the two values are not in it. For how the file-level and page-level systems differ, see robots.txt vs meta robots.
What do the AI operators document instead?
Each of these operators documents robots.txt, not meta tags, as the way to control its crawlers. None of the pages below mention noai or noimageai.
| Operator | What it documents for opting out | Meta tag mentioned? |
|---|---|---|
| OpenAI | GPTBot is for content that may be used to train its models; disallowing it signals that your content should not be used for that. OAI-SearchBot is managed separately. | No |
| Anthropic | ClaudeBot collects content for model training; Anthropic says its bots honor robots.txt "do not crawl" signals, set per subdomain. | No |
| Google-Extended is a robots.txt product token for Gemini training and grounding. It has no separate user agent string and does not affect inclusion in Google Search. | noai not listed | |
| Perplexity | PerplexityBot is not used for foundation model training and is controlled through robots.txt. | No |
Sources: OpenAI crawler overview, Anthropic's crawler article, Google's common crawlers page and Perplexity's bots page.
Two cautions apply. First, absence from documentation is not proof that nothing reads the tags. An operator, or a third-party scraper, may well check for them without saying so. You just cannot rely on that. Second, the user-triggered fetchers are a separate matter. OpenAI says robots.txt rules "may not apply" to ChatGPT-User, and Perplexity says Perplexity-User generally ignores robots.txt. Neither page says how those agents treat meta robots values. The difference between training, search and user-triggered crawlers is covered in allowing AI search while blocking AI training.
Does anyone read the tags?
Some tools do. The open-source img2dataset downloader has a disallowed_header_directives option that drops any image whose HTTP response carries certain X-Robots-Tag directives. Its documented default list is noai, noimageai, noindex and noimageindex, and you can turn the check off by passing an empty list. That shows the header form is read by at least one open-source dataset tool, by default. It is a statement about that tool's behavior only: it says nothing about OpenAI, Anthropic, Google or any other operator, and it covers the HTTP header, not the meta tag.
How to add the tags if you still want to
The tags are cheap to add and may be read by tools that choose to honor them, as long as you do not treat them as your main control. The header is the better choice, because it also works for images and other files that have no HTML head, and a downloader that reads the header will see it on every response.
Meta tag: add it to the head of each page.
<meta name="robots" content="noai, noimageai">nginx: add the header in a server or location block.
add_header X-Robots-Tag "noai, noimageai";Apache (with mod_headers enabled):
Header set X-Robots-Tag "noai, noimageai"Google documents the Apache and nginx header patterns on its robots meta tag page. Check the result on a live URL rather than trusting the config:
curl -sI https://example.com/images/photo.jpg | grep -i x-robots-tag
curl -s https://example.com/ | grep -io '<meta[^>]*robots[^>]*>'On WordPress and similar platforms, check what a plugin actually sends before relying on it. A plugin checkbox that says it blocks AI may only add the meta tag. If you use the tag form, remember that a crawler can only read it after fetching the page, so a robots.txt Disallow on the same URL means the tag is never seen.
What the tags do not do
- They are not documented to change search visibility. Google's documented rules do not include them, so Google describes no effect on Google Search. They are not
noindex. - They do not control snippets or AI answers. In Google, the rule that stops content being used as a direct input to AI Overviews and AI Mode is
nosnippet, which also removes your snippet from ordinary results. See noindex, nosnippet and X-Robots-Tag. - They do not block fetching. A crawler that ignores them downloads the page exactly as before.
- They are not enforced. Nothing audits whether a scraper honors them. Raptive, which offers a checkbox for the tags in its ads plugin, says in its announcement that it has no assurance all bots will respect them and that they work mainly as a public statement of your policy.
What to use for a real opt-out
Use the controls the operators describe, then add server rules for anything that does not follow them.
1. Named robots.txt groups
Name each training crawler. A crawler obeys the group that names it, so these groups sit alongside your existing rules:
User-agent: GPTBot
Disallow: /
User-agent: ClaudeBot
Disallow: /
User-agent: Google-Extended
Disallow: /This addresses the training crawlers without touching search: OpenAI says its GPTBot and OAI-SearchBot settings are independent, and Google says Google-Extended does not impact inclusion in Google Search (it also covers Gemini grounding, not only training). Anthropic says its bots honor these signals, and that a rule naming ClaudeBot takes precedence over a general one. Add each crawler's token separately; the full syntax is in robots.txt for AI crawlers, and what Google-Extended covers is in Google-Extended explained.
2. Server rules for what robots.txt cannot stop
robots.txt is a request. For a crawler that ignores it, or one you do not want at all, refuse the request at your server or CDN using the user agent, optionally combined with the operator's published IP ranges, since a user agent string is easy to fake. Be careful with Anthropic's crawlers: Anthropic says blocking its IP addresses "may not work correctly or persistently guarantee an opt-out", because it impedes their ability to read your robots.txt file. Keep /robots.txt reachable and use the named group for them. Examples for nginx and Apache are in blocking AI crawlers with user agent rules.
3. Authentication for anything that must stay private
If content must not reach any AI tool, do not serve it to anonymous visitors. RFC 9309 itself says the protocol is not a substitute for content security measures.
Frequently asked questions
Does ChatGPT respect noai?
OpenAI's crawler documentation does not mention noai or noimageai. For training it documents GPTBot in robots.txt, so use a GPTBot group if you want to opt out.
Do noai tags hurt my SEO?
Nothing in Google's documented robots meta tag rules covers noai or noimageai, so Google does not document any effect from them. They are different from noindex, which does remove a page from results.
Should I use the meta tag or the X-Robots-Tag header?
The header, if you can set it. It covers images and other non-HTML files, and the one dataset tool documented above reads the header rather than the meta tag. Both can be used together.
If I add noai, can I skip robots.txt?
No. The robots.txt tokens are the opt-out that OpenAI, Anthropic and Google actually document, and the noai tags are not. Use robots.txt first and treat the tags as an extra signal.
Check what your site sends
The free check reads one page the way AI crawlers do, with no account needed. It reports the robots.txt rules for each named AI crawler and the robots meta tag and headers it finds, so you can see what a crawler would actually read. A full-site audit runs the same kind of checks across up to 500 pages.
Filed under Access and robots.txt. Topics: meta-robots, ai-crawlers, robots-txt, google-extended