Copilot and Gemini can use a law firm’s website in two ways. A page can help train future AI models, or it can be pulled in when someone asks a question.
On October 30, 2026, Google said it intends to support a shared robots.txt line for AI training in early 2027, and Microsoft announced separate training and grounding settings for Bing.
For a firm weighing an AI training opt out on its law firm website, training and grounding are two separate choices with very different results.
Copilot and Gemini can use a law firm’s pages for AI training and for grounding answers.
From February 2027, Bing will read a shared robots.txt line for training, Google intends to support it in early 2027, and Bing adds a separate grounding setting. Blocking training does not remove a firm from search results, but blocking grounding reduces how often the firm appears in AI answers.
Training and Grounding Are Two Different Choices
Training and grounding are the two ways an AI tool can use a law firm’s page.
- Training means the page may be used to build future versions of an AI model.
- Grounding means the page is fetched when someone asks a question, so the answer can rest on it. Google’s crawler documentation describes grounding as providing content from the Google Search index “to the model at prompt time.”
A firm can say yes to one and no to the other. OpenAI runs its own separate crawlers behind ChatGPT search.
What Google Offers for Gemini Now and in Early 2027
Google’s current control for Gemini is a robots.txt token called Google-Extended. According to Google’s common crawlers documentation, it covers both training of future Gemini models and grounding in Gemini Apps, so blocking it blocks both together. Google says Google-Extended “does not impact a site’s inclusion in Google Search” and is not used as a ranking signal.
An October 30 note adds that Google intends to support the shared training line, User-Agent: __usage_ai-training, in early 2027. AI Overviews and AI Mode use a separate switch, the Search generative AI control in Search Console.
What Bing Is Changing for Copilot
Bing’s new settings split training and grounding for Copilot, according to Microsoft’s October 30 announcement.
- AI Training: set in robots.txt for a site or folder, or with a page-level tag.
- AI Grounding: set page by page only, with no site-wide robots.txt option.
- DisallowExtended: keeps grounding within Microsoft’s own experiences, such as Copilot, and out of third-party services.
A site-wide training block in robots.txt looks like this:
User-Agent: __usage_ai-training
Disallow: /Bing says the settings “do not change eligibility for traditional Bing search results.” Blocking grounding, though, means “reduced brand visibility in AI experiences.”
The Dates Law Firms Should Put on the Calendar
Bing’s dates matter most for firms that used NoArchive or NoCache to stay out of Copilot.
- December 15, 2026: a temporary grounding setting opens in Bing Webmaster Tools, only for sites using NoArchive or NoCache.
- February 1, 2027: Microsoft starts honoring the new settings as it crawls.
- March 1, 2027: after this date, NoArchive and NoCache are no longer read as training or grounding preferences.
- September 30, 2027: any temporary grounding preference expires.
Pages relying on NoArchive or NoCache with no temporary preference by March 1 default to Allow for grounding until recrawled.
Which Setup Fits the Firm
The firm wants the widest AI visibility
A firm that wants to show up in AI answers should leave grounding open and decide on training separately.
The firm does not want its pages used to train models
A firm that objects to training can add the shared training line to robots.txt, which Bing honors from February 2027 and Google intends to support in early 2027. Search results are not affected.
The firm used NoArchive or NoCache to stay out of Copilot
A firm that relied on NoArchive or NoCache should add the page-level grounding tag and record the temporary Bing Webmaster Tools setting before March 1, 2027.
Frequently Asked Questions
Blocking AI training does not hurt a law firm’s regular search presence, according to both companies. Google says Google-Extended does not affect inclusion in Google Search and is not a ranking signal, and Bing says its new settings do not change eligibility for traditional Bing results.
AI training uses a page to help build future AI models, while AI grounding pulls a page in when someone asks a question so the answer can rely on it. Grounding is what lets a firm’s page appear in a Copilot or Gemini answer.
NoArchive and NoCache stop working as AI controls for Bing after March 1, 2027, when Microsoft stops reading them as training or grounding preferences. Firms using them should add Bing’s page-level grounding tag and record a temporary preference in Bing Webmaster Tools.
A law firm can block AI training on part of its website by naming a folder in robots.txt instead of the whole site, or by adding Bing’s page-level training tag to individual pages. Bing starts honoring both from February 1, 2027.
- Training and grounding are separate choices.
- Google-Extended covers Gemini training and Gemini Apps grounding, and Google says it does not affect inclusion in Google Search.
- Bing’s new settings do not change eligibility for regular Bing results, but blocking grounding reduces visibility in AI answers.
- Firms using NoArchive or NoCache as an AI control need to act before March 1, 2027.
Setting an AI Training Opt Out on a Law Firm Website
An AI training opt out on a law firm website is a fair choice that costs nothing in regular search. Grounding is different. It decides whether Copilot and Gemini can show the firm’s pages when a potential client asks a question, so treat it as a visibility decision.
For most firms, the sensible path is to set training to match the firm’s comfort level and leave grounding open. Firms that used NoArchive or NoCache should fix those pages before March 1.
Lexicon’s legal content experts can run a law firm website content audit to sort the pages that earn trust from the ones that need work. Call (877) 486-8123 to get started.
