I Found 1,200 AI Citations Hiding in Bing Webmaster Tools

For three months, Microsoft Copilot was citing my site hundreds of times and I had no idea. The count was sitting in a Bing Webmaster Tools dashboard I had never opened.

When I finally looked, the AI Performance tab showed 671 verified Copilot citations across the previous 90 days. As of late May 2026, that number has climbed to roughly 1,200. My site, chudi.dev, had a Domain Rating of 25 and almost no backlinks. Something was citing it anyway, and the only reason I can prove it is a free dashboard most people never open.

This is the case study behind my freeCodeCamp guide, How to Measure Your AI Citation Rate Across ChatGPT, Perplexity, and Claude. The guide covers the live-API method for ChatGPT, Perplexity, and Claude. This post covers the part that did not fit: the one Microsoft data source that shows your Copilot citations for free, and how a guest post fed mine.

You are probably being cited and flying blind

Most people tracking AI visibility watch ChatGPT, Perplexity, and Claude. Microsoft Copilot, which runs on the Bing index and reaches hundreds of millions of Windows and Microsoft 365 users, reports your citations in a dashboard almost nobody checks.

That blind spot costs you more than a vanity number. If you cannot see your Copilot citations, you cannot tell which pages earn them, whether a guest post moved the needle, or whether a content change helped or hurt. You are optimizing in the dark on the one AI surface that already hands you the data for free. I spent months building AI-visibility infrastructure before I realized the scoreboard had been on the whole time.

Where are your Microsoft Copilot citations hiding?

They are in Bing Webmaster Tools, under the AI Performance tab. It is free, it only requires that you verify your domain, and it reports how often Copilot cited your pages, which queries triggered the citations, and the trend over time.

Here is the setup, start to finish:

1. Go to Bing Webmaster Tools and add your domain as a property. The fastest path is the Google Search Console import, which verifies you in one click if you already use GSC. Otherwise verify by DNS record or meta tag.
2. Wait for data to populate. Bing backfills recent history, so you often see citations from the prior weeks within a day or two of verifying.
3. Open the AI Performance tab. You get a citation count, the queries that surfaced your pages, and a trend line.

The first time I opened it, the number was not zero. It was 671. That reframed everything I thought I knew about who was reading my site.

What the data showed for a small, new site

chudi.dev went from no assigned Domain Rating to DR 25 in roughly three months, and logged 671 Copilot citations in that first 90-day window. By late May 2026 the running count was about 1,200. This is a small, young site with no press coverage and a thin backlink profile by any traditional SEO measure.

The takeaway is not that the citations were huge in absolute terms. It is that they existed at all, were measurable, and were growing, on a site that every legacy metric would tell you to ignore. The AI Performance tab turned a guess into a number I could track week over week.

How a freeCodeCamp guest post fed the number

The clearest inflection came after I published on freeCodeCamp, a developer publication with a Domain Rating around 90. A guest post on a high-authority site does two things AI engines reward.

First, it corroborates your name and topic across an independent source. When “Chudi Nnorukam” and “AI visibility” appear on both chudi.dev and freecodecamp.org, a model has two independent sources binding the same entity to the same expertise. Corroboration across sources is one of the strongest signals an engine has for deciding whether a name is an authority on a topic.

Second, it gets crawled and indexed fast. High-authority publications are recrawled constantly, so a guest byline enters the retrieval pool in days, not months, and it carries a link back to your own domain.

I am hedging the causality on purpose: I cannot prove the freeCodeCamp post caused a specific number of citations, because Bing does not attribute citations to referring sources. What I can say is that the accumulation curve steepened after the guest post went live, and the mechanism, entity corroboration on a trusted domain, is exactly what the GEO research predicts should help.

This is a citation count, not a citation rate

Be careful with the 1,200 number, because it is easy to misread. It is a count of Microsoft Copilot citations from Bing. It is not a per-engine citation rate, and it is not AI visibility. Those are three different measurements, and conflating them is the most common mistake I see.

When I poll ChatGPT, Perplexity, and Claude directly through citability.dev, chudi.dev’s citation rate sits around 30% as of May 2026. For contrast, Ahrefs, at DR 88, shows 100% AI visibility but only a 5% citation rate: the model knows the brand everywhere, yet rarely links it as a source. Count, rate, and visibility each answer a different question. The Bing dashboard gives you the count for one engine. It is a starting point, not the whole picture.

What actually fed the citations

From the case study and the published research, a few things correlate with getting cited. Notably, domain authority is not one of them.

• Original statistics. The Princeton and IIT-Delhi generative engine optimization study (KDD 2024) found that including original data drives up to roughly 40% more AI visibility. chudi.dev publishes its own measured numbers, which is content a model cannot get anywhere else.
• Brand mentions on trusted domains. That same body of GEO research puts the correlation between brand mentions and AI Overview citations at a Spearman coefficient of 0.664. A guest byline on freeCodeCamp is a brand mention on a domain engines already trust.
• Topic specificity and structure. Narrow, well-structured content that answers a specific question gets pulled more reliably than broad editorial content. Tables, clear headings, and answer-first paragraphs make extraction easy.
• What did not matter: llms.txt. A SE Ranking study across 300,000 domains found no measurable correlation between llms.txt presence and citations. I keep one because it is free, but I would not call it a lever.

How to read your own AI Performance data without fooling yourself

Once the data populates, three columns matter. The citation count is your headline number, but the query list is where the value is: it shows the exact questions where Copilot reached for your page. Read that list like a content brief. The queries you already win tell you what to double down on. The adjacent queries you almost win, where a competitor got cited instead, tell you what to write next.

Then map the citations back to pages. If one or two URLs are pulling most of the citations, study what they have in common: a clear question-and-answer structure, original numbers, a tight topic. That pattern is your template. Apply it to the pages that earn nothing.

Two cautions. Bing’s data lags and samples, so treat short-window swings as noise and watch the multi-week trend instead. And remember the tab only covers Copilot and Bing-indexed surfaces; it tells you nothing about ChatGPT, Perplexity, Claude, or Gemini. For those you need live API polling, which is the method the freeCodeCamp guide walks through and the thing citability.dev automates. The honest workflow is to use Bing as your free, always-on Copilot scoreboard, and a live-polling tool when you need the per-engine breakdown the Bing tab cannot give you.

When the count is zero, that is the diagnosis

If you verify your domain and the AI Performance tab shows nothing, do not assume the tool is broken. A zero is information. It usually means one of two things: the engine is not crawling you because your infrastructure blocks or confuses it, or it is crawling you but your content is not structured in a way it can extract and attribute.

The infrastructure side is mechanical. Check that your robots.txt does not block the AI crawlers, that your pages render real HTML rather than client-only JavaScript, and that your structured data names you as the author. These are the signals that decide whether an engine can even read you, let alone cite you.

The content side is harder and matters more. Engines cite pages that answer a specific question in a self-contained block, lead with the answer, and carry a number or a fact they cannot find elsewhere. If your pages are broad, editorial, and built for a human skimming for vibe, a model has nothing clean to lift. Rewrite one page in the answer-first, original-data shape and watch whether it starts showing up. That single experiment teaches you more than any checklist.

Open the tab today

If you take one thing from this: open Bing Webmaster Tools and check your AI Performance tab. The data is probably already there. If the count is zero, that is its own useful signal, it means your structure is not getting picked up yet, and that is fixable.

When you are ready to see the full per-engine picture, the free scan at citability.dev runs 10 infrastructure and content checks in about two minutes and tells you which signals are and are not in place before you spend anything. The 1,200 citations did not arrive by luck. They arrived because a small site had the right structure in the right places, and because one guest post put the same expertise on a domain the engines already trusted.

Run the free scan at citability.dev