# Opinion: A shared “no AI training” signal is progress, but silence still means yes

URL: https://ai-pulse.aglaganova.com/blog/opinion-a-shared-no-ai-training-signal-is-progress-but-silence-still-means-yes
Published: 2026-10-31
Author: Roman Aglaganov, Full-stack developer
Category: Opinion
Region: Global
Tags: copyright

> One line in robots.txt that all the big AI companies respect is better than a dozen. But the new signal still treats every site that says nothing as willing, and it does nothing about pay.

![Illustration: a black text-file outline with a single red line of text, surrounded by many blank grey file outlines](https://ai-pulse.aglaganova.com/api/media/file/opinion-training-signal-1.png)

This week OpenAI, Google, Microsoft and Anthropic said they would honour a shared robots.txt rule, “__usage_ai-training”, that lets a website say once whether its content may be used to train AI. It is a genuine improvement. Until now, opting out meant tracking down every company’s crawler names and blocking them one by one, and a site that missed one had opted in by accident.

## What it gets right

The draft is modest in a good way. It reuses robots.txt syntax that every web developer already knows, older software ignores it safely, and it separates two questions that have long been tangled: whether a bot may visit a page, and what the content may be used for. Microsoft’s version goes further, letting sites refuse training while still allowing their pages to be used, with links, in AI answers.

## What it leaves out

Three things. First, the default. Under the draft, a site with no such rule has expressed no preference, and Microsoft’s documentation is explicit that both training and grounding “default to Allow”. Millions of small sites will never hear of the new rule, so their silence will continue to be treated as consent.

Second, the past. Microsoft says its preferences cover future training. A publisher who adds the line today does not get its work out of models that were already trained on it.

Third, money. The signal is a yes or a no. As Microsoft puts it, “Neither choice guarantees inclusion, traffic, or payment.” Opting out of grounding even costs visibility in AI answers. For publishers whose search traffic is falling as AI summaries replace clicks, that leaves a choice between being used for free and being invisible.

## What should come next

The companies backing the signal should publish regular reports on how they honour it, so publishers can check rather than trust. They should apply a new “no” to future versions of models already trained on that content. And the standard should be finished, through the IETF working group, with a way for licensing terms to be stated as easily as a refusal. A clear “no” was the easy part. The harder part is making “yes, on these terms” just as simple.

Disclosure: AI Pulse is written with AI tools, including Claude, which is made by Anthropic.

## Sources

- [IETF Datatracker: Machine-Readable Training Preference Signal (draft-kelley-aipref-training-preference-signal-00)](https://datatracker.ietf.org/doc/draft-kelley-aipref-training-preference-signal/)
- [Bing Webmaster Blog: Your content, your choice: clearer preferences for training and grounding](https://blogs.bing.com/search/2026/10/your-content-your-choice-clearer-preferences-for-training-and-grounding/)
- [The Verge: The biggest names in AI are backing a new way to opt out of scraping](https://www.theverge.com/tech/1023763/openai-google-microsoft-ai-scraping-standard-opt-out)