How to get Copilot to recommend your brand

Copilot recommends brands it can retrieve from Bing. Consumer Copilot is grounded on the Bing web index, which is crawled by bingbot. Here is what surprises almost everyone: there is no Copilot crawler. A 2026 audit of documented AI robots.txt tokens found that none of the 15 belongs to Microsoft, so there is no Copilot token to allowlist. The whole job runs through bingbot, plus a pair of meta robots directives that decide whether Copilot may use your content. If bingbot is blocked, if your Bing presence is thin, or if your meta directives tell Copilot to stay away, your brand is invisible in Copilot's answers.

Why Copilot skips your brand

Copilot's failure modes are its own:

1. Your Google-only SEO does not cover it. Copilot retrieves from the Bing index. A brand that lives only in Google's index barely exists for Copilot, and brands with no Bing Webmaster Tools presence at all are common.
2. You chase a Copilot token that does not exist. Brands add a Copilot or Microsoft AI user-agent to robots.txt thinking it helps. It does nothing. The only crawler token that matters is bingbot's.
3. A meta tag quietly opts you out of Copilot usage. NOARCHIVE tells Copilot your content may not be used in its responses, grounding, or training; NOCACHE limits it to your URL, title, and snippet. This is reported from third-party summaries of Bing's guidelines, not confirmed by Microsoft directly, but it is the documented lever on how Copilot uses your pages. Many sites inherited a blanket NOARCHIVE years ago for SEO reasons and never reconsidered it.
4. Your enterprise buyers cannot see you no matter what you do. Microsoft 365 Copilot is primarily grounded on tenant data, and web grounding is an admin setting that can be switched off for the whole org. If your buyers ask Copilot inside M365 with web grounding disabled, no public-web work on your part makes you appear.

Step 1: Allow bingbot, not a Copilot token

Check your robots.txt at [yourdomain.com]/robots.txt, replacing yourdomain.com with your actual domain. If nothing loads, create the file at the root of your site. Then add near the top, before any broad disallow rules:

User-agent: bingbot
Allow: /

Do not add an invented Copilot user-agent. None exists, and made-up tokens change nothing. Then open Bing Webmaster Tools, register the site, and run its robots.txt tester and Fetch as Bingbot to confirm Microsoft's side sees what you intend.

Step 2: Prove the visits are real

Spoofed user-agents are cheap. Check your own logs for actual bingbot visits, then validate the IP behind each one:

grep -i "bingbot/2.0" /var/log/nginx/access.log

Microsoft publishes the official bingbot IP list and a Verify Bingbot tool. Never judge authenticity from the user-agent alone: a bingbot token from a non-Microsoft IP is a bot in a costume.

Step 3: Read your meta robots directives

Open a key page's HTML head and read the robots meta tag. If it says NOARCHIVE, Copilot is told not to use that content in responses, grounding, or training. If it says NOCACHE, Copilot is limited to your URL, title, and snippet. Decide what you actually want on the page before you remove anything; some brands set these deliberately on internal pages. The risk is the blanket sitewide NOARCHIVE nobody remembers setting.

Step 4: Keep Bing fresh with IndexNow

IndexNow feeds the Bing index, and Copilot retrieves from the Bing index. That is the honest chain: no Microsoft statement says "IndexNow feeds Copilot" directly, so treat freshness as flowing through Bing. To set it up, generate an API key, host the key file at your root, and submit changed URLs whenever product pages change. Responses: 200 received, 202 validation pending, 403 invalid key, 429 spam.

Step 5: Know your buyer's Copilot surface

Ask who asks about you. Consumer Copilot on the web, in Edge, or in Windows retrieves from Bing, and everything above applies. M365 Copilot inside a customer's tenant grounds on their own data first, with web grounding optional and admin-controllable, and identifiers stripped when it does send a query to Bing. If your buyers are teams inside M365 with web grounding off, your public web work is necessary but not sufficient. That conversation is about their tenant config, not your site.

What this guide does not cover

Getting retrievable is the start. Bing refreshes on its own schedule and your catalog drifts, which is monitoring work. The enterprise side is out of scope too: tenant data grounding and admin settings belong to your customer's IT department, not your website. And a recommendation is not a transaction: the checkout layer is separate work.

For the ChatGPT version of this playbook, see our guide to getting ChatGPT to recommend your brand.