> ## Content Index
> Fetch the complete content index at: https://www.2ndshift.gg/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI Training Opt-Out: One robots.txt Rule for Google, OpenAI & More
- URL: https://www.2ndshift.gg/ai-training-opt-out-robots-txt/
- Published: 2026-11-04T13:00:37.000Z
- Updated: 2026-11-07T15:22:47.000Z
- Description: Google, Microsoft, OpenAI and Anthropic will honor one robots.txt AI training opt-out. Here's the rule, the dates and what it won't do.
- Author: CJ Davis
- Tags: AI

**Google, Microsoft, OpenAI and Anthropic will all honor one robots.txt rule that says "don't train on my site."** They announced support on Friday, October 30\. If you run a site, blog, portfolio or game page, you'll eventually be able to opt out of AI training with a single group instead of chasing every crawler by name. It's not live everywhere yet, and it's a request, not a lock.

## How to opt out of AI training with robots.txt

The token is `__usage_ai-training`. It goes in your robots.txt like a user agent. This is the draft's own example, which opts out the whole site except a press kit:

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

A plain `User-agent: *` block doesn't count. If your file only has that, you've expressed no training preference, so you need this group spelled out.

## When each company starts honoring it

- **Microsoft:** February 1, 2027
- **Google:** sometime in 2027 (its crawler docs say early 2027)
- **OpenAI:** "in the coming months"
- **Anthropic:** work has started, no date
- **Meta, Amazon, Apple:** no commitment

## What it doesn't do

- **It doesn't block crawling.** It states a preference about use, and the draft has no enforcement.
- **It isn't an official standard.** It's an individual IETF draft from Anthropic's Kevin Kelley, with no IETF endorsement, and it expires May 3, 2027.
- **It doesn't touch old scrapes.** The preference follows the robots.txt in effect when content was fetched, and the draft doesn't address anything collected earlier.
- **It doesn't get you paid.** Microsoft says it creates no licensing or payment deal.
- **It doesn't cover live AI answers by default.** Pulling your page in to answer a question isn't training, and it won't change Google or Bing rankings.

**One trap for Google:** if you use the token without mentioning Google-Extended, Google treats that as a Google-Extended block. PPC Land reports that also pulls your content from Gemini grounding, so add a Google-Extended allow if you still want to show up in Gemini.

## The 2nd Shift take

This is real progress and also a little late. One rule beats a robots.txt full of bot names, and any creator who's tried to keep up with that list knows it. But the labs agreed to a polite sign on the door after the house was already cleaned out, and nothing here pays anybody.

Add it anyway when you're ready. It's three lines, it puts your position on record, and the big players said they'll respect it. Just don't confuse a preference with protection. Not sure where you fit as a creator in all this? Take the [CRAFTR Quiz](https://www.2ndshift.gg/quiz/).

Are you blocking AI training or leaving the door open? Bring it to the [2nd Shift Discord](https://discord.gg/SbzeqKvGVA?ref=2ndshift.gg).

**Keep reading:** [Epic will allow AI training on Unreal Engine outputs](https://www.2ndshift.gg/epic-unreal-engine-eula-ai-training/) and [AI clone fears hit indie game demos](https://www.2ndshift.gg/ai-clone-fears-indie-game-demos/).

*Sources:* [*The Next Web*](https://thenextweb.com/news/ai-training-opt-out-robots-txt-google-openai-microsoft-anthropic?ref=2ndshift.gg)*,* [*PPC Land*](https://ppc.land/google-microsoft-openai-and-anthropic-back-one-ai-training-opt-out/?ref=2ndshift.gg)*.*