Guide · Practitioner
How Microsoft Copilot Picks Sources and Citations
Copilot turns your prompt into a Bing search query, grounds its answer in the results and shows the query and sources it used ([Microsoft](https://support.microsoft.com/en-us/microsoft-365-copilot/how-web-search-works-in-microsoft-365-copilot-chat-and-agents)). Which pages get cited therefore depends on Bing's ranking, which Microsoft says weighs relevance, quality and credibility, user engagement, freshness, location and language, and page load time (Microsoft).
Facts are from Microsoft and Bing documentation, accessed 30 Oct 2026. Audit methods are ours.
What happens between prompt and citation?
| Step | Source |
|---|---|
| User prompt | — |
| Copilot generates a short search query | Microsoft |
| Query sent to Bing | Microsoft |
| Bing ranks results from its index | Microsoft |
| Response grounded in results, with source references | Microsoft, Microsoft |
For Microsoft 365 Copilot, admins can control whether web search is allowed (Microsoft). Work users at some organisations won't see web citations at all.
Why is the visible query useful?
Because Copilot shows the generated query (Microsoft), you can see what Bing was actually asked. Our view: this is the most useful diagnostic any AI engine offers. If you aren't cited, search that exact query in Bing and see where you rank.
How does Bing judge source authority?
Microsoft lists authority factors within quality and credibility (Microsoft):
| Factor | Question Bing asks |
|---|---|
| Reputation | What kinds of sites link to it? |
| Level of discourse | Is the content meant to harm? |
| Level of distortion | Does it separate fact from opinion? |
| Origination and transparency | Is it first-hand? Is ownership clear? Are sources attributed? |
Bing also says that without clear intent for low-quality content, it assumes users want high-authority results.
What does AI Performance show?
Bing Webmaster Tools' AI Performance report (public preview since February 2026) shows how your content is cited in Microsoft AI experiences (Bing; Bing). Use it to see which pages are cited and how often, then compare with pages that should be.
How do you audit Copilot citations?
- Pick 20 buyer prompts.
- Run each in Copilot; record the generated query and every source.
- Search the generated query in Bing; record your rank.
- Compare with AI Performance data.
- Repeat monthly.
| Prompt | Generated query | Sources cited | Your Bing rank |
|---|---|---|---|
| "what's the best CRM for a 10-person team" | "best CRM small team" | 4 sites | 14 |
Illustrative row.
How do you interpret results?
| Pattern | Likely cause | Action |
|---|---|---|
| You rank top 5 but aren't cited | Page doesn't answer clearly | Rewrite the opening |
| You rank 20+ | Bing ranking | Improve relevance, authority, freshness |
| Not in Bing at all | Indexing | Check Bingbot and Webmaster Tools |
| Competitor's review page cited | Authority gap | Earn coverage on that kind of site |
Does location matter?
Yes. Bing considers user location, hosting location and page language (Microsoft). Run audits from your buyers' market and in their language.
What about freshness?
Bing generally prefers fresh content but notes some content stays relevant for years (Microsoft). Our view: date pages and update them when facts change; don't change dates without real updates.
How does Copilot compare?
| Engine | Search index | Shows the query? |
|---|---|---|
| Copilot | Bing (Microsoft) | Yes |
| ChatGPT | Third-party providers and own (OpenAI) | No |
| Claude | Not published for text; Bing for images (Anthropic) | No |
Who should own the audit?
In most teams, SEO owns Bing ranking and the audit, while content owners fix the pages it flags.
Common audit mistakes
- Testing in Microsoft 365 Copilot where the admin has turned off web search (Microsoft)
- Ignoring the generated query and guessing what Bing was asked
- Mixing results from different markets or languages
- Counting citations without checking the claim matches the page
- Running a single test and treating it as stable
Checklist
- [ ] Generated queries recorded
- [ ] Bing rank checked for each
- [ ] AI Performance compared
- [ ] Audit in buyers' market and language
- [ ] Actions assigned
Related
Sources
- Microsoft, How web search works in Copilot; How Bing delivers search results; Manage public web access
- Bing, AI Performance; Introducing AI Performance
- OpenAI, Searching the web; Anthropic, Claude web search
All accessed 30 Oct 2026.
Questions
How does Copilot choose its sources?
Copilot generates a search query, sends it to Bing and grounds its answer in the results. Bing ranks results on relevance, quality and credibility, engagement, freshness, location and speed.
Can I see the query Copilot used?
Microsoft says Copilot shows the web search query it generated along with the sources used, so you can see what was searched.
Do ads affect Copilot's citations?
Microsoft says paid ads are labelled and payment can't improve organic Bing rankings.