Two lines in your website’s robots.txt file can tell AI companies not to train on it

If you run a website, you will soon be able to block AI training on it with one short entry in a file called robots.txt, instead of hunting down a separate rule for every AI company. Microsoft, OpenAI, Google and Anthropic are backing the shared signal, which was filed as a proposed internet standard on Oct. 30, 2026.

The upshot: small business owners, bloggers, churches, schools and creators get one plain switch that says “do not use my pages to train AI models”. It is a request, not a lock, and Microsoft says it does not change whether your pages appear in regular Bing search results.

How to block AI training on your website

Your robots.txt file is a short text file that sits at the root of your site, for example yoursite.com/robots.txt. Search engines and other bots read it to learn your rules. The new signal is a special name, __usage_ai-training, that you put where a bot’s name would normally go.

To ask that none of your site be used for AI training, add these two lines:

User-Agent: __usage_ai-training
Disallow: /

To keep one folder open to AI training, for example a press kit you want AI tools to learn from, add an Allow line under it:

User-Agent: __usage_ai-training
Disallow: /
Allow: /press-kit/

To block only one part of your site, such as paid articles, name that folder instead:

User-Agent: __usage_ai-training
Disallow: /members/

Most website builders and WordPress SEO plugins have a robots.txt editor in their settings. If you are not sure, ask your web host or whoever built your site. It is a two-minute change.

Three mistakes to avoid

1. Do not rely on the star rule. A line that starts with User-Agent: * does not count as this preference. You need the __usage_ai-training name for it to apply.
2. Do not expect it to stop bots from visiting. This signal tells companies how they may use your content. It is not a crawling rule, so bots can still load your pages.
3. Do not leave it blank and assume the answer is no. If your file has no __usage_ai-training section, you have not expressed any preference, and Microsoft treats that as allowed.

When AI companies will start honoring it

Microsoft gave the clearest dates. It says it will start honoring the new AI training preference on Feb. 1, 2027, as it crawls sites. After March 1, 2027, it will stop reading the older NoArchive and NoCache tags as AI training or AI answer settings, so site owners who used those tags should add the new ones before then. Microsoft also says your choice only takes effect after it crawls your page again.

OpenAI has said its support will arrive in the coming months, and Google has pointed to next year, according to reports citing both companies. Anthropic’s Kevin Kelley wrote the draft and filed it with the Internet Engineering Task Force, the group that looks after internet standards such as robots.txt. For now it is an individual draft that the IETF has not endorsed, and nothing in it forces any company to follow it.

Training is not the same as AI answers

There are two different ways AI uses your pages, and this new line only covers one of them.

• Training is when a company feeds your pages into a model so it learns from them. The new __usage_ai-training line covers this.
• Answers, which Microsoft calls grounding, is when an AI tool reads your page in the moment to answer someone’s question, often with a link back to you. Microsoft handles this with a separate tag on each page, not in robots.txt.

For Microsoft’s Copilot and Bing AI answers, that page tag looks like this:

<meta name="Bingbot" content="ai-grounding:disallow">

The downside: blocking AI answers can cut how often your business shows up when people ask an AI tool for a recommendation. Microsoft warns about this directly. For most small businesses, blocking training but allowing answers is the sensible middle ground, because answers can still send customers your way.

Does this hurt your search ranking?

Microsoft says these preferences do not change whether your pages can appear in regular Bing search results. The training signal is about how your content is used, not whether it is indexed. Microsoft’s announcement covers Bing only, so keep your normal search rules for every search engine as they are.

What to do this week

Open yoursite.com/robots.txt in a browser to see what is there now. Decide whether you want AI models trained on your pages. If not, add the two lines above, save, and check the file again in your browser. Set a reminder for February 2027 to confirm it is still there after any site updates.

While you are in your site settings, check whether your site is safe after the October PHP security fixes. For your own accounts, here is how to stop Google ads from using your Gmail and Photos, and what to know before you let ChatGPT use your apps.

Sources: the draft standard filed with the Internet Engineering Task Force and Microsoft’s Bing announcement, both dated Oct. 30, 2026.