Getting Cited by Copilot: What Is in Your Hands

As of October 2026, Microsoft Copilot, AI summaries in Bing, and Microsoft's developer grounding tools all draw on the Bing index. Microsoft documents that path. It does not publish how Copilot ranks or selects among the pages it retrieves.

It helps to split the journey into stages:

• Access: can bingbot fetch and store your page? You control this.
• Retrieval: does your page enter Copilot's context for a given question? Bing's systems decide, per query.
• Generation: does Copilot use and cite the page it retrieved? The model decides.

Everything below that you can act on sits in the first stage. For the wider Microsoft picture, see how Bing and Copilot assemble AI answers. If you also care about Google, read which AI answers use which index, because the two indexes behave independently.

Start in Bing Webmaster Tools. Verify the site, submit your sitemap, and confirm that the pages you want cited are actually indexed. Google Search Console can't tell you this. In Otterly's 2026 experiment, Google removed two test sites while Bing kept both indexed, and combined Bing impressions grew to 119,935 in July. One index being healthy told the owner nothing about the other.

If a page is missing from Bing, fix that first. The question of whether Copilot can cite a site that isn't in Bing's index has a short practical answer: don't plan on it.

A note on IndexNow. Bing consumes IndexNow pings, and Microsoft has said the protocol helps AI systems reference current page versions. That's an assertion with no published study behind it. The IndexNow documentation itself never mentions AI, and a 200 response confirms only that Bing received the ping. Treat it as Bing indexing hygiene, not a citation lever.

Two gates sit in front of the index, and both have to be open.

robots.txt

robots.txt rules are per bot. A blanket rule written to block AI training crawlers can catch bingbot by accident. A minimal explicit allow looks like this:

User-agent: bingbot
Allow: /

Your CDN or firewall

A firewall, CDN or bot-management rule can block a crawler that robots.txt allows. That failure is silent: robots.txt says yes, the edge returns 403, and no dashboard warns you. How often it happens hasn't been measured. Checking it means requesting your own pages as the crawler and reading the status code. The AI Crawler Checker does that request per crawler and shows which ones your firewall turns away.

Rendering

The strongest evidence that AI crawlers skip JavaScript (Vercel's crawler study) covers OpenAI and Anthropic bots. It excluded Copilot, so it says nothing about bingbot either way. Putting prices, product names and key claims in the initial HTML removes that uncertainty at little cost.

Bing added support for the data-nosnippet attribute in October 2025. Its announcement says content with data-nosnippet "is still indexed normally, but it will be excluded from snippets and AI summaries." The content still counts for ranking. It just can't be shown.

That makes it a useful tool and an easy accident. Templates often carry it from someone who wanted to hide a cookie banner or a footer. Check for:

• data-nosnippet on any block that holds facts you want quoted
• <meta name="robots"> values of nosnippet, max-snippet:0 or noindex
• the same directives sent as HTTP headers

There is no opposite control. No markup makes a section more likely to appear in a Copilot answer. These tags only take content out.

Once your page is indexed and reachable, selection is out of your hands. Retrieval runs per question, systems may rewrite the user's prompt into their own search queries, and the model chooses what to cite from what it retrieved.

The research is blunt about content tricks. A July 2026 critical survey of 45 GEO studies (arXiv 2607.14035) found no technique with a stable, cross-platform causal effect on discoverability. Schema, chunking, "AI-friendly" phrasing and llms.txt have no credible evidence as Copilot levers.

Two things stay worth doing:

• State specific facts plainly. A price, a number or a one-line definition on a fetchable page gives Copilot a sentence it can quote.
• Get named on sites Copilot already cites. There's a strong association between being mentioned on cited third-party pages and appearing in answers, with no promised effect.

Bing Webmaster Tools' AI Performance report, in public preview since February 2026 and expanded in June 2026, shows citations across Copilot, Bing AI summaries and some partner integrations. It also lists grounding queries: the phrases the AI actually searched when it retrieved cited content. Microsoft says the grounding query data is a sample. Its Citation Share figure is observational, not a ranking or a competitive scoreboard.

In the evidence we hold, the report had no API, though Microsoft signalled API access during 2026. Check the current state before you build on it. This is why setting up Bing Webmaster Tools for AI search is worth an hour, and why it's the only place you can see which queries Copilot used when it cited you.

Prompt-based trackers see less. In Otterly's test, Microsoft's report credited one site with 42,817 citations while Otterly's tracker logged 3,934. The windows may differ, but any prompt set is a sample.

Otterly published about 1,000 AI-generated posts on each of two new domains. Google deindexed one after about 14 days and the other after about 90, with no manual action reported in Search Console. Bing kept both. Of 3,934 tracked citations for one site, 2,718 (69%) came from Copilot, according to Otterly's write-up.

That looks like a loophole. It isn't one. Google's spam policies name "using generative AI tools or other similar tools to generate many pages without adding value for users" as scaled content abuse. The removal was silent, and Copilot citations don't replace Google visibility. The test also has hard limits: one site per arm, a low-stakes topic, and a vendor running it. Read the experiment's method and limits before drawing anything bigger from it.

The next move: confirm Bing indexing, test bingbot access at the firewall, and remove stray suppression tags. Then publish pages worth citing.

Does Copilot use the same index as ChatGPT? Not reliably. OpenAI says ChatGPT search sometimes partners with other search providers, but it doesn't quantify that link. In Otterly's experiment, sites that ranked in Bing and were cited thousands of times by Copilot drew effectively zero ChatGPT citations, even with ChatGPT's crawlers visiting. Treat Bing indexing as the documented route into Copilot, and manage ChatGPT access separately through OAI-SearchBot.

Does blocking GPTBot or ClaudeBot affect Copilot? No. robots.txt rules apply per bot, and those are OpenAI and Anthropic crawlers. Copilot's documented path runs through bingbot and the Bing index. The risk is a blanket rule, such as disallowing every crawler except Googlebot, that catches bingbot by accident. Read your robots.txt line by line and confirm bingbot isn't covered by a rule you meant for training crawlers.

Can I opt my content out of Copilot answers but stay in Bing search? Partly. Bing documents that content marked with data-nosnippet stays indexed and available for ranking but is excluded from snippets and AI summaries. Page-level nosnippet and max-snippet:0 work the same way for the whole page. These hide display in AI answers. They don't remove the page from Bing's index, which noindex would do.

Can AI visibility tools track what Copilot says about my brand? They can sample it. A tracker asks Copilot a set of prompts and records the answers, which shows what Copilot says for those prompts at that moment. It can't see the grounding queries Copilot actually ran, and its counts will differ from Microsoft's own report. For the trade-offs, see [whether tools can track Copilot answers](/questions/track-microsoft-copilot-answers).