Microsoft Copilot search optimisation is mostly Bing optimisation. Bing's guidelines say Copilot and Bing search share one crawling, indexing and ranking foundation, so fix Bing indexing, crawler access and snippet controls first. Then make key facts easy to quote and test your brand's mentions on a fixed prompt set before and after every change.
Key takeaways
- Treat Bing as the gate. A page Bing has not indexed is not available to Copilot as a source, so confirm indexing before you touch content.
- Check three technical things in order: Bingbot access in robots.txt and your CDN, robots meta directives such as NOARCHIVE and NOCACHE, and how fast Bing learns about changes (sitemaps and IndexNow).
- Verify your site in Bing Webmaster Tools and read the AI Performance report, which shows citations, cited URLs and grounding phrases.
- Test before you change anything: 20 buyer prompts, 3 runs each, logged by date. With 60 answers, a swing smaller than about 12 percentage points is within noise.
- Change one thing per test cycle, wait for Bing to recrawl, then rerun the same prompts. Do not rewrite 40 pages at once and then guess what worked.
How does Copilot pick sources through Bing?
Microsoft's own Bing Webmaster Guidelines say that Bing and Copilot search experiences rely on the same core crawling, indexing and ranking foundation as traditional search. They also list grounding results, the retrieval Copilot uses to build answers, among the outcomes that robots directives control. In practice, Bing's index is the pool Copilot draws from.
Third-party write-ups, such as Third Coast, describe the mechanism in more detail: Copilot turns a prompt into a short search query, sends it to Bing and composes the answer from the pages returned. Microsoft's guidelines page does not spell out that sequence in the text we reviewed, so treat it as a useful model, not a documented specification.
Microsoft has also put GEO into its webmaster rules. Search Engine Journal, February 2026, reported that the rewritten guidelines name Generative Engine Optimization (GEO) and cover how Bing handles content across traditional search and AI-generated experiences.
Two things follow, and both are inferences, not published ranking factors. The pages that rank in Bing for the query Copilot generates are the candidates. And the snippet Bing holds for your page matters, because a page limited to its title and snippet gives Copilot less to cite. Microsoft has not published a weighting for either, so treat any article that claims one with caution.
Copilot, Copilot Search and Microsoft 365 Copilot are different surfaces
Much of what ranks for this topic blurs them. Microsoft 365 Copilot searches a company's own files and mail, and its controls live in the Microsoft 365 admin centre. That is an enterprise tenant problem, not a public web visibility problem. This guide covers public web answers that cite outside pages, which is where a brand can earn a mention.
Is it worth the effort? Goodie, as checked on 2026-11-11, reports that Copilot accounted for 3.2% of 2.8 million web sessions that started from clicking an LLM citation, second only to ChatGPT. The period is not clear from the article and it is one vendor's dataset, so use it as a prompt to check your own Bing Webmaster Tools and analytics data, not as a forecast for your site.
Does Bing indexing affect Copilot answers?
Yes, in the way that matters most: Copilot retrieves through Bing's foundation, so a page missing from Bing's index is not available to be retrieved. Indexing is necessary, not sufficient. It gets a page into the candidate pool. Ranking for the query, a usable snippet and a clear, quotable answer decide whether it is cited.
Microsoft Copilot search optimisation starts with Bing indexing because, per the Bing Webmaster Guidelines, Copilot and Bing search share one crawling, indexing and ranking foundation; a page Bing has not indexed is not a candidate citation, and an indexed page still has to rank and offer a usable snippet.
| Claim | Status | Basis |
|---|---|---|
| Copilot and Bing search share a core crawl, index and ranking foundation | Stated by Microsoft | Bing Webmaster Guidelines |
| NOARCHIVE and NOCACHE change what Copilot may use | Stated by Microsoft | Bing Webmaster Guidelines |
| Citations are reported per URL and per grounding phrase | Documented in the AI Performance public preview | Bing Webmaster Blog, February 2026 |
| Copilot turns a prompt into a short Bing query | Reported by third parties, not confirmed in the guidelines text we read | Third Coast |
| A specific page type, schema type or word count raises citation odds | Not published by Microsoft | Treat as a hypothesis to test |
What should you fix first?
Work top to bottom. Each row is a check you can run in an afternoon, ordered by what blocks everything below it.
| Priority | Check | How to test | Fix |
|---|---|---|---|
| 1 | Site verified in Bing Webmaster Tools | Add the site and confirm ownership | Verify, or import from Google Search Console |
| 2 | Key pages indexed by Bing | Use URL Inspection in Bing Webmaster Tools, or a site: search in Bing | Submit sitemap, fix noindex, canonicals and redirects |
| 3 | Bingbot allowed | Read robots.txt, then check CDN or firewall bot rules | Remove blocks on Bingbot for pages you want cited |
| 4 | Meta directives permit citation | View source for NOARCHIVE, NOCACHE, NOSNIPPET | Remove them from pages meant for Copilot |
| 5 | Freshness signalled | Compare page update dates with Bing's last crawl date | Accurate lastmod in sitemap, IndexNow pings on change |
| 6 | Answer is extractable | Read the first 60 words under each heading | Put the direct answer first, with named entity and figure |
1. Verify and read Bing Webmaster Tools
Bing Webmaster Tools is the only first-party view of how Bing and Copilot treat your site. In its announcement, the Bing Webmaster Blog, February 2026, describes the AI Performance dashboard as showing when your site is cited across Microsoft Copilot, AI-generated summaries in Bing and select partner integrations. It reports total citations and the key phrases the AI used to retrieve cited content, and Search Engine Journal, February 2026, adds page-level activity and trends over time.
Microsoft frames the release modestly, calling it:
"an early step toward Generative Engine Optimization (GEO) tooling in Bing Webmaster Tools"
It is a preview, so expect the report to change. Microsoft's description covers citation counts, cited pages and grounding phrases, not ranking or click data, so do not report either from it. Use the grounding phrases as free prompt research: they are the words Copilot used to retrieve content it then cited.
2. Let Bingbot in, and check the layer above robots.txt
Most crawl blocks are not in robots.txt. They sit in a CDN bot-protection rule, a WAF country rule or a rate limit that returns 403 to unfamiliar user agents. Read robots.txt first, then fetch a key page with a Bingbot user agent string from outside your network and compare status codes with a normal browser. Our free AI crawler checker tests whether named AI crawlers can reach your site and generates a corrected robots.txt, and the AI Crawler Policy Kit explains how to decide training bots separately from search bots.
3. Remove directives that stop citation
Bing's guidelines spell out what each robots meta directive does to Copilot answers. This is the most actionable detail in the update, and easy to break by accident because SEO plugins often add these tags sitewide.
| Directive | Effect on Copilot, per Bing's guidelines | Action for pages you want cited |
|---|---|---|
| NOARCHIVE | Prevents content being used in Copilot responses and grounding results | Remove |
| NOCACHE | Limits Copilot to the URL, title and snippet, reducing citation depth | Remove; the guidelines advise avoiding it on content intended for Copilot |
| NOSNIPPET / DATA-NOSNIPPET | Reported by Search Engine Journal as possibly limiting citation quality | Remove, or apply data-nosnippet only to sections you do not want shown |
| data-snippet attribute | Reported as letting you specify which text Bing can display or cite | Use on the passage that states your core claim |
The NOARCHIVE and NOCACHE wording is from the Bing Webmaster Guidelines. The NOSNIPPET and data-snippet rows follow Search Engine Journal, February 2026, so re-read the guidelines before you change production tags, as the wording can be updated.
4. Tell Bing when pages change
Stale snippets produce stale answers. Submit an XML sitemap with accurate lastmod dates, and consider IndexNow, a protocol that notifies participating search engines, including Bing, when a URL is added, updated or deleted. Microsoft has not said IndexNow raises citation odds. The practical benefit is a shorter delay between a fix and Bing seeing it, which shortens your test cycles.
5. Make the answer extractable
With Bing as the retrieval layer, ordinary on-page clarity does the work. Put a 40 to 60 word direct answer under each question-style heading, name the entity in full, attach a dated figure and link its source. Keep one claim per sentence so a snippet can stand alone. Do not hide the answer in an image, an accordion that needs a click, or content that only appears after JavaScript runs; our guide to technical SEO for AI search covers the rendering test.
How can you check your page right now?
Run the free AI SEO checker on your three most important pages. It runs 30+ checks on whether AI assistants and Google AI Mode can read the page, covering crawl access, structured data and content, and returns the top fixes. Then confirm Bing's own view in Bing Webmaster Tools, because no third-party tool can see inside Bing's index for you.
How do you test Copilot mentions before and after changes?
Most guides stop at the fix list. The gap is proof. Copilot answers vary from run to run, so a single before-and-after check proves nothing. Use this plan.
- Write 20 prompts. Unbranded buyer questions in your category, plus a few comparison and problem prompts. Our guide to choosing prompts to track gives a scoring sheet, and the free AI prompt finder suggests starting questions. Add Bing Webmaster Tools grounding phrases once you have them.
- Freeze the set. Do not edit prompt wording between rounds. Record date, surface (Copilot on the web, signed out), country and language.
- Run each prompt 3 times. That is 60 answers per round. Log whether your brand is named, whether your domain is cited and which competitors appear.
- Record the baseline from Bing Webmaster Tools. Note total citations and the top cited URLs for the same date range.
- Change one thing. For example, remove NOCACHE from 10 product pages, or add direct-answer paragraphs to 5 guides.
- Wait for recrawl. Confirm in Bing Webmaster Tools that Bing has fetched the changed URLs, then rerun the identical 60-answer round.
- Compare against the margin of error. Only call a change real if it exceeds the noise.
Worked example: how big a change counts?
Suppose your brand is named in 18 of 60 answers at baseline, a 30% mention rate. The standard error of a proportion is the square root of p(1-p)/n, about 5.9 points at n = 60, so a 95% interval is roughly plus or minus 12 points. If the rerun shows 24 of 60 (40%), the gap is 10 points and sits inside that noise. You have not shown an effect. Run more repeats, or wait for the AI Performance report to show a sustained change in cited URLs. This is plain arithmetic on illustrative numbers, not a Microsoft figure.
Citations also churn. In a Noma by Nurdd study of 2,545 answers from ChatGPT, Gemini, Perplexity and Claude, two in three cited URLs appeared in one run only, and 36% of cited URLs were cited again within a week. That study did not cover Copilot, but it is a good reason to repeat runs rather than trust one result. For method detail, see how to measure AI share of voice across engines.
Which fix for which symptom?
| Symptom | Likely cause | First action |
|---|---|---|
| Brand never named, no citations in Bing Webmaster Tools | Pages not indexed, or Bingbot blocked | URL Inspection, then CDN and robots.txt check |
| Pages indexed, cited rarely | Weak rank for the grounding query, or no clear answer to quote | Rewrite opening answers; check grounding phrases |
| Cited, but answer misstates a fact | Old snippet or third-party page is the source | Update the page, ping IndexNow, correct the third-party source |
| Cited with thin context | NOCACHE or snippet limits | Remove the directive on those pages |
| Mentioned by Copilot but not by other assistants | Strong Bing presence, weak elsewhere | Compare engines on the same prompt set |
Common mistakes
- Optimising for Google only. A site can rank well on Google and be thin or unindexed in Bing. Check Bing separately.
- Blocking Bingbot to stop AI training. Bing's crawler feeds both search results and Copilot. Decide training controls deliberately and test the effect.
- Leaving sitewide NOARCHIVE or NOCACHE on. Plugins and templates add them. Audit the page source, not the editor.
- Treating schema as a switch. Microsoft has not published schema as a citation factor, so test it like any other change.
- Reading one answer as a result. One run is an anecdote. Sixty is a measurement.
Where does Noma fit?
Noma by Nurdd is an answer engine optimisation (AEO) and AI visibility platform that asks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, your buyers' questions every day. It reports where your brand is named, which competitors appear instead and which pages the answers cite. Noma does not track Microsoft Copilot or Meta AI, so use Bing Webmaster Tools for Copilot and Noma to measure the same changes on the engines it covers.
See it on your own brand: book a demo or start the 3-day free trial. A card is taken at signup so billing can start on day 3; cancel before then and nothing is charged. Current plans are on the pricing page.
How we researched this
We reviewed the top-ranking guides on Microsoft Copilot search optimisation, including Goodie's, to see what they cover and what they omit. We then read Microsoft's Bing Webmaster Blog announcement of AI Performance, the Bing Webmaster Guidelines page, and trade-press and third-party reports of them. Facts, dates and figures were checked on 2026-11-11. Microsoft does not publish how Copilot weights sources, so every ranking-factor claim here is labelled as documented, reported or a hypothesis. We did not run Copilot tests for this article, and the worked example uses illustrative numbers.
What to do next
This week: verify Bing Webmaster Tools, confirm your ten most important pages are indexed, strip NOARCHIVE and NOCACHE where they do not belong, and log your 60-answer baseline. Then change one thing and rerun. To see how those changes carry across ChatGPT, Gemini, Perplexity and Claude, book a demo or start the 3-day free trial.
See what AI engines say about your brand
Free · no account · or start a 3-day trial
Questions people ask
How do I get my brand cited in Microsoft Copilot?
Make sure Bing has indexed the pages that answer your buyers' questions, that Bingbot can fetch them and that no NOARCHIVE or NOCACHE directive limits them. Then put a direct, quotable answer at the top of each page and name your brand, category and a dated figure in plain sentences. Check the AI Performance report in Bing Webmaster Tools to see which of your URLs are cited.
Does Bing indexing affect Copilot answers?
Yes. Bing's guidelines say Copilot and Bing search share the same core crawling, indexing and ranking foundation, so a page absent from Bing's index is not a candidate source. Indexing only makes a page eligible: it still has to rank for the query and offer a usable snippet.
Is there a way to see Copilot citations for my site?
Yes. Bing Webmaster Tools has an AI Performance report, in public preview since February 2026, that shows when your site is cited across Copilot, Bing's AI summaries and select partners. It reports total citations and the grounding phrases used. Microsoft's description does not mention ranking or click data.
Do robots meta tags affect whether Copilot can use my content?
Yes. Bing's guidelines say NOARCHIVE prevents content being used in Copilot responses and grounding results, and NOCACHE limits Copilot to the URL, title and snippet. Audit the page source for these tags, because plugins often add them sitewide.
How do I test whether my changes improved Copilot mentions?
Freeze a set of about 20 unbranded buyer prompts, run each three times, and log whether your brand and domain appear. Change one thing, wait until Bing has recrawled the pages, then rerun the same set. With 60 answers, a change of roughly 12 percentage points or less is within normal noise.
Does Noma track Microsoft Copilot?
No. Noma by Nurdd tracks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, but does not track Microsoft Copilot or Meta AI. Use Bing Webmaster Tools for Copilot citations and Noma to measure the same changes on the engines it covers.