Bing AI Training and Grounding Controls: Australian Impact

AI Search News | Australia | 8 November 2026

Last updated: 8 November 2026

Microsoft announced on 30 October 2026 that Bing is replacing the NoArchive and NoCache tags with separate AI Training and AI Grounding preferences, and Australian site owners who use those tags must act before 1 March 2027 to keep their choices. If you have never set either tag, nothing changes: both settings default to Allow.

What happened

On 30 October 2026, Krishna Madhavan, a Principal Product Manager at Microsoft Bing, published “Your content, your choice: clearer preferences for training and grounding” on the Bing Search Blog. The post says Microsoft was the first company to provide publisher controls for AI training and grounding in 2023, using the NoArchive and NoCache tags. It now wants to offer “newer, simpler, and clearer preferences” that let a publisher decide about training, grounding, or both, while remaining eligible for Bing search results.

The announcement covers web content collected through Microsoft’s crawlers and used for training or to inform generated responses, including web-grounded responses in Bing, Copilot and Microsoft 365. It splits the choice into two independent settings.

• AI Training is a shared preference expressed in robots.txt for a whole site or a path. Microsoft says it is extending the same preference to the page level, so a single page can override the robots.txt setting.
• AI Grounding is a separate, page-level preference from Microsoft. It controls whether your content can help inform generated answers. The options are Allow, Disallow, and a new middle option called DisallowExtended.

DisallowExtended limits grounding use to Microsoft-operated experiences such as AI Mode, AI Overview, Copilot and Microsoft 365. It excludes grounding services Microsoft provides to third parties, which the post names as Web IQ and grounding through API access in Foundry. Disallow stops grounding use altogether.

The training half of the change links to the work of the Internet Engineering Task Force (IETF) AI Preferences working group. Microsoft says it is working with other industry participants to bring a shared training preference to site owners, and describes the syntax in an IETF Internet-Draft. Microsoft is clear that the standard is not final: it says “additional work remains” and that it will keep taking part in the effort.

The syntax, as Microsoft published it

For training across a site or path, the post uses a robots.txt pattern with the user agent __usage_ai-training and a Disallow line. For a single page, you add a meta tag such as <meta name="Bingbot" content="ai-training:disallow"> or an HTTP header such as X-Robots-Tag: Bingbot: ai-training:disallow. Grounding works the same way at page level, using ai-grounding:disallow, or ai-grounding:disallowextended for the middle option. There is no site-wide robots.txt setting for grounding.

Two cautions from the post matter for anyone planning changes. First, these settings govern how content is used, not whether it can be crawled, so an Allow setting does not override a robots.txt rule that blocks Bingbot. Second, publishing a setting is not the same as activating it, because Microsoft has to crawl the page to discover it. The post notes that IndexNow can signal a change and improve the chance of a recrawl, but does not guarantee a recrawl date.

The timeline that matters

1. 15 December 2026: temporary grounding transition preferences open in Bing Webmaster Tools. They are only available to sites using NoArchive or NoCache as of that date.
2. 1 February 2027: Microsoft starts honouring the new AI Training and AI Grounding preferences as it crawls.
3. 1 March 2027: the deadline for sites using NoArchive or NoCache to publish new preferences and record a temporary grounding preference. After this date, NoArchive and NoCache are no longer read as a training or grounding preference.
4. 30 September 2027: any temporary grounding preference expires, or earlier if Microsoft recrawls the page and finds the new on-page setting.

The post also warns that choosing Disallow or DisallowExtended for grounding will “significantly reduce” content visibility in AI-generated experiences, without affecting eligibility for traditional Bing search results. It states that neither choice guarantees inclusion, traffic or payment, and that the preferences do not establish licensing arrangements. Microsoft’s own Bing Webmaster Tools help article carries a companion FAQ.

Is it live in Australia?

The announcement does not give a country-by-country rollout, and it does not mention Australia. It describes the change as applying to how Microsoft’s crawlers treat web content, which is not tied to a visitor’s location. The dates in the post are global: the new preferences begin taking effect on 1 February 2027 as pages are crawled. If your Australian website is crawled by Bingbot, which any indexed site is, this applies to you.

What is already live is the old system. NoArchive and NoCache tags still work today. The post says that if Microsoft detects the new preferences on a site, it will disregard any NoArchive and NoCache preferences remaining on that site. So the first practical question for any Australian site owner is simple: does my site use either tag at all?

What we saw in Australian AI search this week

This story is about the supply side of AI answers: which content can be used. To see the demand side, we ran our national spot check on 7 November 2026. The instrument is Google AI Mode, one query per city, using the buying-intent prompt “best electrician in [city]”. We chose a licensed trade because a local service buyer is a classic AI Mode user, and because many tradie websites run on older themes and plugins, which is where legacy tags are most likely to linger (our reasoning, not a measured finding). We did not test Bing, Copilot, ChatGPT or Perplexity, so nothing below says how Microsoft’s products answered the same question.

Every answer quoted a rating alongside the business it named. Every answer grouped businesses by speciality, such as emergency work, lighting, solar, or older-home rewiring, rather than giving one winner. The answers drew on listing data, and in Adelaide and Sydney, a small number of outside sites were cited alongside Google’s own listing data. In Perth, only Google was cited as a source. We report only what came back, and we are not rating any business. Sydney’s count is a floor because our parser reads a truncated excerpt of a long answer.

The pattern is the point. Even for a local trade, the AI answer leans on structured, third-party signals: reviews, listings, licence details, and clearly described services. Your own website is one input among several, and Microsoft’s announcement is about whether that input can be used at all in its products.

What it means for Australian businesses

This is our view, not Microsoft’s. The change sounds technical, but it creates a real decision for every business with a website, and most will make it by accident if they do nothing.

Most sites should leave grounding on

Grounding is the mechanism that lets an AI answer cite and link to your page. Microsoft says allowing it makes your content eligible to inform responses, “with links and citations providing a path to the original content”. For a Sydney accountancy, a Melbourne cafe group, a Brisbane builder, a Perth mining services firm, an Adelaide clinic or a regional retailer, being quoted is the goal. If the main commercial risk is invisibility, switching grounding off is the wrong lever. Microsoft itself warns of reduced brand visibility in AI experiences if you do.

Some content genuinely warrants a restriction

There are good reasons to restrict. A publisher with paid research, a membership site, a training provider with gated course material, or a law firm with proprietary precedent libraries may want to stop training use while still allowing grounding. The independent settings make that possible: you can set AI Training to Disallow and AI Grounding to Allow, which is the example Microsoft gives. A national news publisher or a regional newspaper might choose differently again. The point is that the two decisions no longer have to travel together.

The hidden risk is legacy tags on old sites

Many Australian small business sites carry tags nobody remembers adding. An SEO plugin default, a developer’s snippet from 2023, or a “hardening” checklist from a hosting provider can all have inserted NoArchive. After 1 March 2027, the post says those tags will no longer be read as a preference. A business that quietly relied on them to keep content out of AI use will find the restriction gone. A business that has the tag and wanted to be cited has nothing to fix, but should still know what its pages say. Either way, you want a deliberate setting, not an inherited one.

Agencies and multi-site owners have the biggest workload

Page-level grounding settings must be published page by page. For a franchise group with hundreds of location pages, or an agency managing dozens of client sites, that means a template or plugin level change and a plan for recrawl. Bing Webmaster Tools also carries the temporary transition preference from 15 December, which is another thing to track per site. Regional Australian businesses with a single WordPress site have it easier, but they are also the least likely to have heard about it.

This is one platform, and the controls are fragmented

This announcement only covers Microsoft’s crawlers and products. Google, OpenAI and others each have their own controls and names, and the shared training signal described in the IETF draft is still a draft. Microsoft says the goal is to “reduce the complexity of navigating different platform preferences”. Until a standard lands, Australian businesses need to treat each platform’s rules as separate checklist items. We cover the Google side of this in our earlier briefing on Google’s “control what you share” documentation.

What to do now

1. Audit for NoArchive and NoCache. Check your page source, HTTP headers and any SEO or caching plugin settings. Also check robots.txt for any existing rules aimed at AI crawlers. Write down what you find, because the old tags stop counting after 1 March 2027.
2. Decide training and grounding separately. For most businesses that want AI visibility, the answer is Allow for grounding. Decide training on its own merits, based on whether your content is a paid or proprietary asset.
3. Use the right level. Training can sit in robots.txt for a site or path, or on a page. Grounding is page-level only. Do not expect a site-wide grounding switch, because Microsoft says none exists.
4. Diarise the Bing Webmaster Tools window. If you currently use NoArchive or NoCache and want to withhold content from grounding, the temporary preference opens on 15 December 2026 and must be recorded before 1 March 2027.
5. Ask Bingbot to come back. Publishing a setting does nothing until the page is recrawled. Use IndexNow after you change key pages, knowing it improves the chance of a recrawl but does not guarantee a date.
6. Keep your content worth quoting. Permission is only the first gate. Clear answers, accurate business details, reviews and structured data decide whether you are used once you are allowed. Our AEO services and GEO services cover that work, and you can follow every briefing on our AI Search News hub.

If you want a deeper view of how Bing sees your content today, our earlier piece on the Bing Topics Portfolio and AI Performance reports shows what Bing Webmaster Tools already reveals. The LocalBusiness structured data vocabulary is also worth reviewing, as it supplies many of the business details AI answers tend to quote. For more on being named in AI answers, see our AI readiness guide for Australian websites.

Frequently Asked Questions

What is the difference between AI Training and AI Grounding in Bing?

AI Training covers using content to modify the learned parameters of a generative AI model. AI Grounding covers using content as input to help inform an AI-generated response when an answer is created. Microsoft says they are independent preferences, so your training choice does not change your grounding choice, and both default to Allow.

Do I need to do anything if I have never used NoArchive or NoCache?

Not to keep things as they are. Microsoft says both preferences default to Allow when nothing is published, and the transition steps apply only to sites using NoArchive or NoCache as of 15 December 2026. You can still publish explicit preferences if you want a clear record of your choice.

Will blocking AI grounding remove my site from Bing search?

No. Microsoft says these usage preferences do not change eligibility for traditional Bing search results, subject to normal crawl and indexing requirements. It does warn that choosing Disallow or DisallowExtended for grounding will significantly reduce visibility in AI-generated experiences.

What is the deadline for Australian sites?

Microsoft’s dates are global. The new preferences start taking effect on 1 February 2027 as pages are crawled, and sites using NoArchive or NoCache must act before 1 March 2027 to maintain their chosen preferences. Microsoft has not published separate Australian dates.

Does this change how Google AI Mode treats my website?

No. This announcement covers Microsoft’s crawlers and products only. Google has its own controls, and our spot check of Google AI Mode in Perth, Adelaide and Sydney shows those answers lean heavily on listing and review signals as well as web content.