Can You Block AI Training and Still Rank on Google?

You can now ask Google to keep indexing your pages for search—while saying no to AI training on your content. Cloudflare rolled this out on September 15, 2026.

The old “Block AI Bots” switch? Gone. Now you get three separate controls: Search, Training, and Agent. Each one handles a different job.

Why One Switch Never Worked

Certain crawlers wear two hats. They scan your site to build search results. They also scoop up your content for AI training. Cloudflare labels these mixed-use crawlers. Googlebot, Bingbot, and Applebot all belong here.

These bots carry serious weight. They drive 36.6% of all verified crawler traffic across Cloudflare’s network. That’s the biggest single category they monitor.

Here’s where it gets tricky. Site owners treat search and training as two very different things. Under 1% block search bots. Yet 17% try to block training somehow.

The old toggle left you stuck. Block one, block both. For an ad-supported site, that meant gambling with Google traffic just to keep content away from AI training.

What “Accountable” Really Means

Cloudflare slapped an Accountable label on crawler operators. It’s not decorative. Operators must clear four hurdles:

1. Offer you a way to opt out of AI training (via robots.txt or something similar).

2. Let you opt out of AI-generated search summaries.

3. Reveal which pages got used for training—down to the URL.

4. Promise that opting out of training won’t tank your search rankings.

Apple, Google, and Microsoft now clear that bar. So do Amazon, Anthropic, Meta, and OpenAI. Their training crawlers sit apart from their search crawlers. Cloudflare can block one without disturbing the other.

The Caveats You Must Know

First, “Block” changed its meaning. In the past, Block bypassed mixed-use crawlers. That’s over. Hit Block on the Training setting, and Bingbot, Googlebot, and Applebot stop coming. Not even for search. Want search but no training? Disallow AI Training is your only path.

Second, Bing lags behind. Microsoft is still building robots.txt support for a no-training preference. Early 2027 is the target. Until then, Disallow AI Training doesn’t pass this preference to Bingbot on its own. You’d need Bing’s content-removal tool or the NOARCHIVE meta tag.

Third, defaults roll forward. Cloudflare says most people won’t need to touch a thing. Had Block before? It shifts to Search: Allow, Training: Disallow AI Training, Agent: Block on pages with ads.

Training Was the Hard Part. Summaries Are Next.

Cloudflare calls mixed-use crawlers “the hard part of the training question.” Then it pivots to the next battle: AI summaries.

Over half of consumers read AI summaries in search. They’re also 40% more likely to bail without clicking through. Rough deal for sites earning per page view. But visitors arriving from AI search convert at 3x-5x the rate of traditional search visitors. Fewer visits—but stronger intent.

Cloudflare wants each site to control how much content shows up in an AI summary. One dashboard. Not separate settings scattered across every operator. That’s slated for early next year.

Behind all this sits a promise from three of the world’s biggest crawler operators: They claim your rankings won’t shift if you add a no-training flag. For now, Cloudflare’s scoreboard is the only way to verify they keep their word.