Bing and Microsoft Copilot

Four MCP tools read Bing Webmaster Tools for a connected Website: performance, a page-and-search drill-down, crawl issues and keyword ideas. All four require only the read scope, all four are free, and all four exist separately from the Google tools because Bing is the index Microsoft Copilot answers from, which no Google surface can see into.

The table above is generated from RankX AI itself rather than written by hand, as at 8 September 2026 (source 43d562be). When the product changes, this page changes with it.

Bing is the index Microsoft Copilot answers from

This is the reason these tools are worth calling at all, and it is the sentence to keep in mind when reading anything they return. Microsoft Copilot grounds its answers in Bing's index. A page Bing cannot fetch, or that Bing has been told not to fetch, is a page Copilot cannot cite, however good that page is and however well it ranks in Google.

Nothing on the Google side of the RankX AI MCP surface can see that. inspect_url and get_index_coverage answer the same question about Google's index and say nothing about Bing's, so a site with clean Google index coverage can still be invisible to Copilot.

Every Bing tool here is free, in two different senses

get_bing_performance and get_bing_crawl_issues read rows RankX AI has already synced, so they spend no credits and make no upstream call at all.

get_bing_drilldown and get_bing_keyword_ideas do call Bing live, and they are still free: Bing charges nothing for these reads. get_bing_keyword_ideas is worth singling out, because it is the free second opinion beside research_keywords, which is a spend tool buying Google-side volume. The two are not substitutes, for the reason in the next section but one.

Bing reports weekly, and that is a grain rather than a discrepancy

Bing's per-search and per-page figures are bucketed by week ending, where Search Console's are daily. So get_bing_performance takes a number of weeks and every figure it returns covers whole weeks, never a part week. A days parameter over a weekly grain would either round silently or slice a bucket in half, and both produce a number nobody can reconcile.

Bing's site-level totals are daily while these two grains are weekly, so the two do not add up to each other. Report that as the grain it is, not as a fault in the data.

One more property of Bing's own reporting: it publishes two different average position statistics, click-based and impression-based, which answer different questions. get_bing_performance names which basis a figure used rather than printing a bare number that invites a comparison neither supports.

The crawl report is an observation with a date on it

get_bing_crawl_issues lists pages Bing could not read or was told not to read: server errors, pages not found, DNS failures, timeouts, malware flags and pages blocked by robots.txt.

It is a point-in-time observation, not a live check. Microsoft's own notes on the underlying report say a fixed issue can stay listed for a few days, so the response carries the date Bing captured it and every answer built from it should quote that date. An assistant told "47 broken pages" with no date will report a live outage for something the site owner fixed on Tuesday.

The class worth acting on first is an important page blocked by robots.txt. Bing has judged the page important and been told not to read it, so the fix is one line in one file and the payoff is a page Microsoft Copilot can cite at all.

Bing impressions are not Google search volume

get_bing_keyword_ideas returns related search terms with the impressions Bing itself recorded for them. Bing is a minority of most sites' search traffic, so a figure here is a lower bound on interest and never a volume estimate. Every string the tool returns says "Bing" rather than "searches", so an assistant cannot quote it as one by accident.

For Google-side volume, difficulty and cost per click, use research_keywords instead, which is priced. Offering the Bing figure as a free second opinion is the point; offering it as a replacement would be a worse product sold as a saving.

get_bing_keyword_ideas also requires a market when the Website has not had one explicitly set. A Website's country defaults rather than being answered, so the tool asks instead of quietly reporting United States numbers to a British business.

What these tools need before they return anything

Every Bing tool needs Bing Webmaster Tools connected for the Website, with a site selected. When it is not connected, each one says so and names the remedy rather than returning zero. That is the same refusal contract the Google tools follow, and for the same reason: an absent connection and a quiet week look identical in a number, and only one of them is a fact about the site.

If Bing is not connected, say the Copilot question was not checked, rather than implying the site is fine on that axis.

Example prompts

"Which of my pages can Bing not read, and when did Bing last look?"

"Show me Bing clicks and impressions by search term for the last eight weeks, and tell me which grain those figures use."

"Which searches drive traffic to my pricing page in Bing?"

"Is anything important on my site blocked from Bing by robots.txt?"

Where to go next

• Traffic, for what each connected reporting source covers and the states in which a figure is refused.
• Rankings and Google data, for the Google half and why the two must never be read as interchangeable.
• Authentication, for what the read scope reaches.
• null is not zero, before reporting any absent figure these tools return.

Last updated 8 September 2026