Is there an API for Bing's AI Performance report?

As of August 2026, the latest date our evidence covers, there was no API for Bing's AI Performance report. You could see citations and grounding queries in Bing Webmaster Tools, but you could not pull them programmatically.

The reasoning is simple. The Bing Webmaster API reference is a .NET SDK document last stamped 2023-11-14. A search of it for "citation", "grounding", "copilot", "intent" or "topic" returns nothing. A February 2026 Microsoft Q&A thread also says there is no API. That thread alone is weak evidence, but the reference doc backs it up.

The AI Performance report itself is much newer. It entered public preview on 2026-02-10 and expanded on 2026-06-16. The API simply has not caught up. Fabrice Canel of Microsoft has indicated API access during 2026, so this gap has a stated expiry.

What it means: any tool that claims to sync your Bing AI citation data automatically needs to show you where that data comes from. The next move is to check Microsoft's current API docs, and if nothing has changed, plan a manual export.

Google has the same gap for its own AI report, which is covered in why tools can't pull Search Console's AI report.

The AI Performance report covers Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. It exposes:

• Total Citations and Average Cited Pages
• Page-level citation activity
• Grounding queries: in Microsoft's words, "the key phrases the AI used when retrieving content that was referenced in AI-generated answers"
• Intents and Topics
• Citation Share and a Compare view

Grounding queries are the valuable part. Most AI visibility tracking tests prompts that the tracker chose. But engines don't search for the prompt as typed. OpenAI says ChatGPT search typically rewrites a query into one or more targeted queries. Google's Gemini API documentation says grounding automatically generates one or more search queries. Bing's report is the only first-party place that shows the retrieval queries themselves.

Coverage is a separate question. The report does not claim to cover ChatGPT, so read whether it includes ChatGPT citations before you treat it as a cross-engine view.

Until an API ships, a short manual routine gets you a usable dataset.

1. Verify the site. You need a verified property in Bing Webmaster Tools. The setup playbook walks through it.
2. Open AI Performance and fix a date range. Use the same range every time, so exports compare cleanly.
3. Capture each view. Save grounding queries, intents, topics and page-level citations. Use the report's download where it offers one. Otherwise copy the table or take a screenshot.
4. Date-stamp every file. Record the export date and the date range shown. Preview features change, and you need to know what each file covered.
5. Repeat on a schedule. Monthly works for most sites. A gap in your exports is a gap you cannot backfill from outside.
6. Map queries to pages. For each grounding query, note which of your pages it retrieved. That shows which content Copilot actually used.

The detailed version, with analysis steps, is in exporting and analyzing Bing grounding queries.

Microsoft writes that "the data shown represents a sample of overall citation activity." That sentence sits inside the grounding queries subsection and refers to that metric.

Total Citations and page-level activity carry their own caveats, and those caveats do not mention sampling. So the accurate statement is narrow: grounding queries are a sample. Saying the whole report is sampled overstates it. Saying it is complete overstates it the other way. The sampling question has more detail.

Citation Share needs its own care. Microsoft calls it "an observational metric, not a ranking system or a competitive scoreboard." It does not expose competitor domains or traffic share. Because it is a share, it can fall while your absolute citations rise.

One public test shows why first-party counts matter. In Otterly's 2026 experiment, Microsoft's report credited one test site with 42,817 citations, while Otterly's prompt-based tracker logged 3,934. The time windows may not match, so it is not an exact ratio. It does show that a prompt sample sees less than the engine's own count.

A few situations change how you use the export:

• The API ships. Switch to it, but keep your manual files. They are your only history from before the API existed.
• The report is empty. No citations is a result, not an error. First check that Bing can reach and index your pages, since an engine cannot cite what its index doesn't hold.
• You want ChatGPT data. Bing ranking did not produce ChatGPT citations in Otterly's test, even with ChatGPT's crawlers visiting. Measure each engine separately.
• You blend sources. Don't add Bing's counts to tracker counts. They measure different things on different denominators.

The next move: export this month's report, date it, and read it with the AI Performance report playbook. Then set a calendar reminder for the next export.