Should you block AI from training on your website? What the new opt-out does, and what it doesn't.
On 30 October 2026, Google and OpenAI each said they will respect a new shared line in robots.txt that tells AI companies not to use your site to train their AI models. Microsoft (Bing) has signed up too and starts on 1 February 2027. Google says it aims to support it in 2027. The line is about training only. If your robots.txt doesn't name Google-Extended, Google says the line counts the same as blocking Google-Extended, which is separate from Google Search. You don't need to change anything today.
One line in robots.txt, shared by several AI companies.
robots.txt is a small file on your website that tells automated visitors (bots) what they may and may not do. Until now, each AI company had its own bot name, so blocking AI training meant listing them one by one. The new line is meant to work for all the companies that sign up to it.
- Google, 30 October 2026: it will honour the shared AI training opt-out. When robots.txt doesn't name Google-Extended, opting out with the shared line counts as disallowing it. If your file names Google-Extended directly, that rule is what Google follows. Existing controls keep working. Google says it aims to support the line in 2027. Google also says sites that specifically allow Google-Extended can appear in Gemini Apps grounding (Google Search Central).
- OpenAI, 30 October 2026: it will honour the same signal "in the coming months" (OpenAI).
- Microsoft (Bing): it backs the same signal and starts honouring the new preferences on 1 February 2027 (Microsoft Bing blog). Microsoft says these preferences don't change your site's eligibility for Bing search results (Microsoft Bing blog; also reported by PPC Land and The Verge). Search Engine Land reports that Bing adds separate AI Training, AI Grounding and Limited Grounding preferences, plus a page-level tag for Bingbot (Search Engine Land).
- Anthropic: The Verge and PPC Land report it backs the signal too, with no start date yet. PPC Land quotes an Anthropic engineer saying it has started building support. The IETF draft is written by an Anthropic standards lead.
- The standard is still a draft. The exact line and an example are in an IETF draft from the AIPREF group. It may change. The draft says a rule under "User-agent: *" (the "every bot" group) does not count (IETF draft).
Who starts when
- Microsoft (Bing): 1 February 2027.
- Google: aims for 2027 (PPC Land reports early 2027).
- OpenAI: "in the coming months", no date.
- Anthropic: no date.
If you use NoArchive or NoCache to keep your content out of Bing's AI: Microsoft says these sites should set the new preferences by 1 March 2027. It also says temporary grounding controls arrive in Bing Webmaster Tools from 15 December 2026 and expire on 30 September 2027 (Microsoft Bing blog, Search Engine Land).
Training, Search and AI answers are three different things.
Most of the worry comes from mixing these up. The new line is about the first row.
| What you're asking about | Does the new opt-out change it? | What that means for you |
|---|---|---|
| AI companies training their models on your pages | Yes. That is what the line is for, once each company supports it. | Your text should stop going into future training by companies that respect it. It doesn't remove what was already collected. |
| Showing up in Google Search | Google says it counts as disallowing Google-Extended when your robots.txt doesn't name Google-Extended. If it does, that rule is what Google follows (Google Search Central). Google-Extended is separate from Googlebot, the crawler Google Search uses. Bing results aren't affected either. | Don't block Googlebot, and don't put rules under "User-agent: *" unless you mean every bot, Google Search included. |
| Being named in AI answers, such as Gemini or ChatGPT | Possibly. Google says sites that specifically allow Google-Extended can appear in Gemini Apps grounding. | Opting out may mean fewer mentions in some AI tools. How each company handles this is not fully clear yet. |
For most small businesses: no rush, and probably not yet.
This is a trade-off, not a right answer. Here is how we'd think about it.
1. Decide what matters more
If you want customers to find you through AI answers, leaving training allowed is the safer default. If you publish original work, such as recipes, guides or photos, and don't want it used, blocking makes more sense.
2. Wait for support to start
Microsoft starts on 1 February 2027, Google aims to support it in 2027, and OpenAI says it will in the coming months. The standard is still a draft. Adding a line today that nobody reads yet does nothing.
3. Use today's controls if you're sure
Google says existing controls keep working. If you've already decided, blocking Google-Extended covers Google's AI training now. Your developer can set it in minutes.
Mistakes to avoid
- Blocking everything. A rule under "User-agent: *" hits every bot, including Google Search. And the draft says the new training line doesn't count there anyway.
- Copying a line from a blog. The exact wording is in the IETF draft and may change. Ask your developer to use the current version when support starts.
- Forgetting what's there. Many sites already have AI rules someone added in 2023 or 2024. Check your current robots.txt before adding more. If that includes NoArchive or NoCache for Bing, plan to update before 1 March 2027.
What you can change depends on where your site is built.
Before 2027, ask your platform whether it will add a simple switch for the new line.
WordPress you host yourself
You or your developer can edit robots.txt, often through an SEO plugin such as Yoast or Rank Math. This gives you full control over the new line once support starts.
Shopify
Shopify lets you edit robots.txt through a theme file. Shopify warns that mistakes can affect your traffic, so it's a job for a developer.
Wix
Wix has a robots.txt editor in its SEO settings. You can add lines there, but check the result carefully: a typo can hide pages from search.
Squarespace
Squarespace doesn't let you edit robots.txt directly. It has its own setting for AI crawlers instead. Whether it will cover the new line is up to Squarespace.
If you'd rather someone else handled this, we look after small business websites in the US, UK, Canada and Australia, including robots.txt and what AI tools say about you. Your first fix is free. Then from $100 a month. See our AI answer check, start a chat or message us on WhatsApp +92 326 2326191.
What's confirmed, what's likely, what's weak.
This changed on 30 October and will keep changing. We label each claim by how well it's backed.
| Claim | How sure | Source |
|---|---|---|
| Google will honour the shared opt-out, counting it as disallowing Google-Extended when robots.txt doesn't name Google-Extended. Google aims to support it in 2027. Existing controls keep working. | Confirmed | Google Search Central, 30 Oct 2026 |
| OpenAI will honour it "in the coming months". | Confirmed | OpenAI, 30 Oct 2026 |
| The signal is an IETF draft; a rule under "User-agent: *" doesn't count. | Confirmed (draft, may change) | IETF draft |
| Microsoft (Bing) has joined and starts honouring the new preferences on 1 February 2027. | Confirmed (Microsoft's own post) | Microsoft Bing blog, Search Engine Land, PPC Land, The Verge |
| Other Microsoft dates: temporary grounding controls in Bing Webmaster Tools from 15 December 2026, a 1 March 2027 deadline for sites on NoArchive/NoCache to set the new preferences, and temporary preferences ending 30 September 2027. | Confirmed (Microsoft's own post) | Microsoft Bing blog, Search Engine Land |
| Anthropic has started building support. | Likely (one quote) | PPC Land |
| Opting out of Google's AI features also cut normal clicks. | Weak (one small test, one site) | Search Engine Roundtable |
Should I block AI from training on my website?
It's your choice, and there is no rush. Blocking keeps your words out of future AI training by the companies that respect the signal. The cost is that some AI tools may mention or link to you less. If being found and recommended matters more to you than how your text is used, most small businesses are better off leaving training allowed for now.
Will blocking AI training hide my site from Google Search?
Google says the new shared opt-out counts the same as disallowing Google-Extended, its existing AI training control, when your robots.txt doesn't name Google-Extended. If it does, that rule is what Google follows (Google Search Central). Google-Extended is a separate control from the Googlebot crawler that Google Search uses. Don't block Googlebot or use a rule for every bot unless you mean to leave Google Search.
Will it affect my Bing results?
No. Microsoft says these preferences don't change your site's eligibility for Bing search results, subject to normal crawling and indexing (Microsoft Bing blog). Reports on the announcement add that Bing rankings are not affected (The Verge, PPC Land).
Will I still show up in ChatGPT, Gemini or Google's AI answers if I opt out?
Not fully known yet. Google says sites that specifically allow Google-Extended can appear in Gemini Apps grounding, so opting out may mean fewer mentions there. One small test found that opting out of Google's AI features also cut normal clicks, but that was one site and a different setting. Check each company's notes when support starts.
When does the new opt-out start working?
Microsoft says it starts honouring it on 1 February 2027. Google says it aims to support it in 2027 (reported as early 2027). OpenAI says it will honour it in the coming months. Anthropic has given no date. The standard itself is still an IETF draft, so the exact wording may change before then.
I use NoArchive or NoCache to keep my content out of Bing's AI. Do I need to do anything?
Yes, but not today. Microsoft says sites on these older directives should set the new preferences by 1 March 2027. It also says temporary grounding controls arrive in Bing Webmaster Tools from 15 December 2026 and expire on 30 September 2027 (Microsoft Bing blog).
Do I need to change anything today?
No. Google says existing controls, such as blocking Google-Extended, keep working. If you have already decided to block AI training, those controls do the job now. If you haven't decided, you can wait until support starts and the standard is final. If you use NoArchive or NoCache for Bing, plan to update before 1 March 2027.
Can I change robots.txt on Wix, Squarespace or Shopify?
It depends on the platform. WordPress you host yourself, Shopify and Wix let you or a developer edit robots.txt. Squarespace does not let you edit the file directly, but has its own setting for AI crawlers. Ask your platform whether it will add a switch for the new line before 2027.
Sources for this page
- Google Search Central blog: AI training control, 30 October 2026
- OpenAI: Helping websites express AI training preferences, 30 October 2026
- Microsoft Bing blog: Your content, your choice: clearer preferences for training and grounding, October 2026
- IETF: draft-kelley-aipref-training-preference-signal-00 (draft)
- Search Engine Land: Microsoft Bing gives publishers more control over how AI uses their content
- PPC Land: Google, Microsoft, OpenAI and Anthropic back one AI training opt-out
- The Verge: OpenAI, Google and Microsoft back an AI scraping opt-out standard
- Search Engine Land, 30 October 2026 (Bing and Anthropic report)
- Search Engine Roundtable: AI features opt-out and clicks (one site test)