Cloudflare's Disallow AI Training: What 'Block' Does Now

Cloudflare shipped a setting called Disallow AI Training on 15 September 2026, and the same release changed the meaning of a switch plenty of site owners already had turned on. In Cloudflare’s own words, Block and Block on pages with ads now apply to mixed-use crawlers — bots that crawl for search and for AI training under one identity. Applebot, Bingbot and Googlebot are named explicitly.

So Block no longer means “block the AI bots”. It stops those crawlers altogether, search included. If you turned it on to keep AI off your content and you still want to rank, you are on the wrong setting; the one you want is Disallow AI Training.

Two numbers explain why Cloudflare built a separate setting instead of one “Block AI” switch: fewer than 1% of Cloudflare sites choose to block Search bots, while 17% enable some mechanism to block training. Almost nobody wants out of search; a sizeable minority wants out of training.

The four settings

All four apply at the domain (zone) level and are available on every plan:

Block on pages with ads is the one that used to be narrow and safe. It is still narrow in scope, and it now reaches Applebot, Bingbot and Googlebot on those pages.

If you never touched the granular controls, you were migrated

Cloudflare had already split bot controls into Search, Training and Agent. Sites that never configured them were migrated automatically, with the practical effect of the old Block AI Bots setting preserved:

• Disabled (unselected) → Search Allow, Training Allow, Agent Allow
• Block → Search Allow, Training Disallow AI Training, Agent Block on pages with ads
• Block on pages with ads → Search Allow, Training Disallow AI Training, Agent Block on pages with ads

Sites that had configured the granular controls kept their practical effect too: a Training setting of Block or Block on pages with ads becomes Disallow AI Training, while Search and Agent selections carry over unchanged. Block AI Bots and Managed Robots.txt are both retired in the same release, the latter replaced by Bot Preference Sync.

So in almost every case there is nothing to do — which is exactly why this is easy to miss. The migration protects you; a deliberate click does not. If you want the mixed-use crawlers gone entirely, you now have to say so, and Cloudflare states plainly that doing so stops Applebot, Bingbot and Googlebot.

The two gaps Cloudflare documents itself

Neither of these is an outside criticism. Both are in the announcement, and both matter before you treat the setting as settled.

Bingbot does not honour the preference yet. Microsoft is “currently building the mechanism to also respect a ‘no training’ preference in robots.txt at the domain/site level, targeted for early 2027”. Until that ships, Cloudflare writes, selecting Disallow AI Training “will not automatically convey a no-training preference to Bing through robots.txt”. Bing’s own route in the meantime is its NOARCHIVE meta tag, plus its Block URLs or Content Removal tool. On Bing, the setting’s name currently outruns its effect.

URL-level transparency is incomplete. One of the Accountable requirements is URL-level visibility into which pages were made available for training. Google says it is building that for Google-Extended, expected “in the weeks to come”; Applebot “does not yet provide a tool for URL-level inspection”, though Cloudflare says Apple has shared details of an in-progress solution for next year. Apple and Google have each stated that disallowing training does not impact search ranking.

On the Accountable label itself, Cloudflare requires four things: a robots.txt-style way to opt out of training, a way to opt out of AI summaries (directly with the operator today, through Cloudflare next year), URL-level training visibility plus search metrics, and an assurance that opting out does not affect traditional search. Apple, Google and Microsoft qualify, as do the relevant crawlers from Amazon, Anthropic, Meta and OpenAI, which separate search from training crawlers. There is no Disallow control for Agent traffic yet, because no established standard for it exists. And robots.txt alone is explicitly not the answer: it cannot identify who is crawling, determine why, or stop a crawler that ignores it. Enforcement has to happen at the network.

The next fight is AI summaries

Training and summaries are different questions. Training is about whether your content builds a model; summaries decide whether someone visits your site or reads the answer inside search. Cloudflare calls the current site-wide yes-or-no too blunt, because how much of your content appears in a summary matters as much as whether it appears at all. Summary opt-outs are already one of the Accountable requirements, and the stated goal is that by early next year you will set how much of your content may be used once on Cloudflare, rather than operator by operator.

The impact data cuts both ways, and Cloudflare cites third-party research for all of it: more than half of consumers read summaries in Search and are over 40% more likely to end their search after reading one, which can reduce visits — but consumers referred by AI Search convert at between three and over five times the rate of traditional search referrals. Fewer visits, much greater intent. A publisher funded by advertising optimises for volume; a retailer may prefer fewer visitors who are likelier to buy.

What to do with this

1. If your goal is “no training, keep search”, check that Training is set to Disallow AI Training — not Block. This is the whole story in one line.
2. If you genuinely want no crawlers, Block is now the honest way to say so — and it will cost you Applebot, Bingbot and Googlebot.
3. Do not assume robots.txt is doing the work. Outside Cloudflare you can still publish the directive, but nothing enforces it for you.
4. Verify the promise instead of trusting it. “Opting out of AI training will not affect traditional search results” is an operator assurance, not a measurement you have seen. Watch your own search visibility for a few weeks after changing the setting — keyword and rank tracking such as Semrush is how you turn that assurance into evidence about your own site.
5. If Bing matters to you, plan for the early-2027 date. Until then the robots.txt preference does not reach Bing, and NOARCHIVE is the lever that exists today.

Every figure and every quoted requirement above comes from Cloudflare’s announcement of 15 September 2026; the third-party research it cites for the AI-summary numbers is named in that post rather than restated here as our own measurement. We have not run these settings on a live domain, and this article does not claim otherwise.