How to Use Bing Webmaster Tools AEO Data to Get Ahead

Bing has released the first official AI performance tracking platform, focusing on citations and grounding queries. But beyond the excitement of something new, what can a report on Microsoft Copilot usage and Bing AI summaries actually offer us?

Here are several ways to put it to work and come out on top:

Hello, GPT-5

Despite the lukewarm interest Microsoft’s web ecosystem sometimes generates, often dismissed as a boomer hangout, there’s a nuance worth noting that should reassure everyone: Copilot is powered by GPT-5. That means some of the data shared through Bing Webmaster Tools actually originates from OpenAI. Do we know exactly which data? No.

Another interesting point: mentions are not part of the tracked metrics, only citations, for now, along with their corresponding grounding queries. Given the emphasis Fabrice Canel (Bing’s product manager behind all of this) has placed on monetizing organic search, this is an additional signal that citations are the most “reliable”, or least murky, metric for contributing to business objectives.

Technical Actions

Now for something concrete amid all the abstraction: what can you put in place, technically, to make the most of this new data?

First: cross-reference your Bingbot log data with citation data by page over a given time period.

Review the list of actions taken on your site during that period (if you don’t have a tracking document logging at minimum the action and its launch date, start one now without delay), study the fluctuations in Bingbot crawl activity, then look at the fluctuations in citation counts.

The goal is to find the winning formula: action + crawl rate increase + citation count increase.

Found it? Now it’s time to check your conversions from AI sources to confirm — or at least correlate — that it’s having an impact across multiple LLMs.

Got a lead? Replicate your action elsewhere to validate your hypothesis. Classic SEO, in other words.

Example workflow:

• You enabled server-side rendering on November 15th for a group of pages
• You observe citation counts from November 15–30 in Bing Webmaster Tools and compare to the previous 15 days
• You observe Bingbot crawl volume from November 15–30 and compare to the previous 15 days
• You look for a statistically significant increase in both metrics (≥5%)
• You draw a hypothesis (this action worked / didn’t work)
• If positive: you scale it up. If negative: you move on and find another action to test
• You build a Claude artifact or document your process so you can repeat it effortlessly. Next time, you’ll just need to load the new datasets for quick initial estimates and hypotheses

As with everything in SEO, this isn’t an exact science, but it gives your efforts a direction, maximizes your chances of impact, and makes use of the data available to you, without waiting for OpenAI’s Webmaster Tools.

Content Actions

It’s time to take a closer look at grounding queries. While they’re the closest thing to keywords, they are not the actual commands or prompts entered by users — they’re the labels assigned to those queries (see JC Chouinard’s write-up).

In other words, these are the key queries the AI uses to go fetch content.

With that in mind, it’s now possible to identify the most popular query types that prompt the LLM to pull up your pages, and more importantly, to spot the ones that aren’t performing as well.

Once you’ve mapped out your visibility opportunities, it’s a matter of reviewing the outputs generated by different prompts (not just on Copilot), seeing what surfaces, and adding passages to your content that fill in the gaps, then evaluating whether your visibility has improved.

NotebookLM is typically a great tool for this, even though it runs on Gemini. It lets you add a defined set of sources, text, links, videos (including competitors who are outranking you), then enter a prompt and get a response. Optimize your content, re-run the prompt, and instantly see whether you’re now mentioned or cited in places where you weren’t before, no waiting required. Here’s a video walking through the process.

Again, this isn’t OpenAI or any Microsoft tool, but it’s an imperfect way to evaluate what content changes to make with immediate feedback. Alternatively, you can always make the changes, wait a bit, and watch how things evolve.

In Summary

Bing Webmaster Tools is a first step, clearly improvable (a performance filter by URL would have been a welcome addition), but it points the way forward.

The currently recommended actions are:

• Cross-reference your Bingbot user-agent log data with Bing Webmaster Tools citations over specific time periods
• Dig into the lower-performing grounding queries surfaced by Bing Webmaster Tools, study the corresponding AI responses, and update your content accordingly

There’s a strong chance Google and OpenAI will release their own versions this year, possibly before summer, which is exactly why now is the time to get ahead before everyone else rushes toward these new toys.