Microsoft Built the AI Visibility Tool Google Still Doesn't Have. We Found 3,200 Citations on One Page.

Microsoft Built the AI Visibility Tool Google Still Doesn't Have. We Found 3,200 Citations on One Page. Speaker 1: Now we're on a client call last week, sharing our screen, and looking at Bing Webmaster Tools. On a section none of us had really dug into properly before. AI Performance. One page, 3200 citations. The next highest page on that site, 334. We just stared at the screen for a few seconds. They got 10x more AI citations than anything else. This has been building for 3 weeks and we didn't even know. Because until now, there was no tool that could show you this information. AI is reading your content, using it to answer questions, shaping buyers' decisions, and you have no way of knowing which pages are doing that work and which ones aren't, or you didn't until a few weeks ago. I work with ecommerce brands on Amazon and AI search visibility. We've been tracking AI bot traffic, grounding queries and citation data for some time, and this particular tool is the most useful thing I've seen for understanding what AI is actually doing. With your content. Now in this video, I'm going to walk you through what this dashboard is, what it actually measures, and this is important, what it doesn't measure. Because there's a trap in this data that will send you into the wrong direction if you miss it. And at the end of This, I'll show you why the most valuable section in the entire dashboard is the one most people scroll straight past. Now until 6 weeks ago, if you wanted to know whether AI was citing your brand, whether ChatGPT, Copilot or Gemini were pulling your content into answers, you had two options. Now you could either type Hype questions into those platforms yourself and hope you showed up or you could track bot traffic in your analytics and make educated guesses. That was it. That was the state of VR. Google Search Console? Nothing on AI overviews specifically. They blend it into general web search data. You cannot isolate it. You can't tell what's AI-driven and what's traditional or gallic. Bing just changed that. On February 9, Microsoft launched AI Performance inside Bing Webmaster Tools. It's the first dedicated dashboard from any major search engine that shows publishers how their content performs in AI generation. So here's the quick win right now. Go to bing.com forward slash webmasters forward slash AI performance. If you have a site verified in Bing Webmaster Tools, you're already in. It's free and there's no application. If you're not verified in Bing Webmaster Tools yet, that takes two minutes. You should have done it already. So maybe pause this video and do it right now. Now if you're not subscribed Hit that now. Now I cover this kind of thing before most of the market world is paying attention. The dashboard has five things it shows you. Total citations, how often your content was pulled as a source in AI generated answers over a given period. Average cited pages, the average number of unique pages from the site cited per day. Site level citation activity, which specific URLs are getting cited and how How many times? Then you have visibility trends, a lifeline of how citation activity is changing over time, and the fifth one, grounding queries. That's the section that matters the most. But before I explain it, let me tell you what happens on the client call, because the context will help you understand everything else. We're working with a brand that sells beauty skincare products. In late February, they published a non-content piece with strong clinical backing. Still relatively early in mainstream consumer awareness, the article went live on the 23rd of February. On a call on March 18th, listed Then four weeks later, we pulled up Bing AI Performance Dashboard live for the first time. Click pages and there it is. 3,200 citations on that one article. The next page, 334. After that, 319. Then 268. The brand owner goes, you think we should do a couple more similar articles? That is the right instinct. But before you try to replicate a result, you need to understand what Actually, that's where grounding queries come in. When you click the grounding queries tab in this dashboard, your first instinct is to expect something like this. What is PDRN? Does PDRN help with anti-aging? Is PDRN safe for sensitive skin? Human questions. That's the kind of thing someone would type. But that's not what you see. The queries look more structured, more clinical. Sometimes they read like search strings rather than natural questions. And that's because of what they actually are. When someone asks ChatGPT, a copilot, A question the AI doesn't just answer from memory. Before it builds the response, it fires its own subqueries out against a web index, sometimes five to ten of them, to pull in verified accurate information. That process is called grounding. Those subqueries are what you're seeing in the tab. The human asked a question, the AI asked different questions to go find the answer. Grounding queries show you what the AI was looking for, not what the person typed. Hi, that one distinction to AI subqueries versus human prompts changes how you read every piece of data in this dashboard. So what does this tell you strategically? So, first layer, what the AI already associates with you. If the grounding queries firing against your PDRN article include things like clinical evidence, skin regeneration, or PDRN concentration, wound healing study, the AI is treating your content as a medical reference That's a different position than Best PDR & Serum to buy. So Layer 2, Language Drift. The AI might be querying your content using terminology you've never actually You didn't write those exact phrases, but the AI is connecting them to your content anyway. That's a content gap you can fill by writing about it. Layer 3. Programmatic Queries. Some of what you see in grounding queries look oddly automated. That's because it is. There are GEO monitoring tools, so Generative Engine Optimization platforms, That systematically query the web to benchmark which brand AI systems are treated as authoritative. Seeing those queries fire against your content is actually good. It means you're being evaluated and catalogued. Now here's the practical Download your grounding queries. Drop them into a spreadsheet. Look for phrases that appear repeatedly. Look for questions that suggest purchase intent. Look just at ingredient search. That data is telling you what AI thinks you're an authority on and what it's trying to confirm. When it reaches your pages. Now that's your next content brief. Not what ranks on Google, what AI is already looking to you to answer. So you see 3,200 citations and you think 3,200 people found this brand through AI? That's not what's happening here. There's a documented case where a page accumulated over a thousand Bing AI citations during a period when it had just three traditional Bing search impressions. Three, the AI reads the content, used it, Built answers with it, humans didn't necessarily click through. The citations in this dashboard are machines retrieval events. The AI fetched your page to build its answer. Whether a human then clicked through to the site, that data isn't here. There's no click-through rate in AI performance right now. Microsoft knows it. It's a gap they haven't filled yet. This is what I call the visibility gap. High citation volume to a potentially minimal actual traffic. And if you don't understand this going in, you misread what the data is telling you and optimized for the wrong thing. Think about citations in two separate buckets. So bucket one, authority. Every time your content is cited, you're shaping the AI's internal model of who you are, how expert you are, what topics you own. That then compounds. A brand that gets Constantly cited on a topic becomes a default reference for that topic and I believe that's going to become a mode. Now bracket to commerce, the question isn't just are we getting cited, it's what does the person do after they read the AI answer? Are we contributing If there's no path from the citation to the product, you're building brand equity with no way to capture it commercially. This is exactly the conversation that happened live on the call. Client sees 3,200 citations. He doesn't celebrate. Within 30 seconds, he says, we need to put banners on that page, banners that link directly to the PDRF. The article earns a citation. The citation pulls the article in front of the AI-informed buyer. The article has to convert that buyer into a customer. If there's no link from the informational content to the commercial page, you build a library with no gift If a citation gets you in the room, the conversation architecture closes the deal. So here's where I want to make sure you're not over-reading the data. When you see 3,200 citations in this dashboard, you might assume that covers all AI platforms, every chatbot, every AI search experience, the full picture. No, it doesn't. AI Performance tracks Microsoft Copilot and Bing AI summaries directly. What it describes as Select Partner Integration, those aren't named specifically, but the nuance is that it's ChatGPT. Because ChatGPT historically used the Bing Search API to browse the web, when ChatGPT fires a grounding query and retrieves a page from Bing Index, that can show up here. That's probably part of what the client is seeing. He confirmed on the call, he thought it was GPT all along. With Copilot but, ChatGPT has been moving away from Bing. One analyst tracking 240 million ChatGPT citations found that Bing's overlap with ChatGPT went from 26% in April last year down to 8% by July. ChatGPT is increasingly pulling from Google's index instead, so AI performance gives you practical ChatGPT insights, real data, just not the complete ChatGPT story. Perplexity? It's not here. Gemini? Not here. Claude? Not here. This is a Microsoft ecosystem tool. Now there's one more thing. The dashboard itself shows a disclaimer right on the top. The data shows below represent a sample of overall activity. The numbers are real. They're not capturing everything. If your page shows 3,200 citations, the actual number is probably higher. Here's the practical playbook. Step one, go to bing.com forward slash webmasters forward slash AI performance. Set it up if you have it. Step 2. Go to Page tab. Find the highest cited page. If one page is significantly outperforming everything else, understand why before you try and replicate it. Topic, format, external authority signals, like a press release maybe. Step 3. Go to grounding queries. Look at the phrases, look for language drifts, terms you didn't write that the AI is associating with your content. Those are your next big content pieces. Step 4. Whatever your highest cited page is, order it. Does it link to a product? Does it have a clear next step for someone ready to buy? Fix that before anything else. You've done the hard work by earning a citation. Don't waste it. Now for the last few years, the question has been, does AI know who you are? Now you can find out. 3,200 citations, 3 weeks, 1 article. Nobody knew until somebody looked. The brands that win in AI search aren't going to produce the most content. They're going to be the ones who understand what AI was already doing with their content and build around And while we're talking about platforms changing the rules without warning, the next video is already live. TikTok sent an email to every seller on February 17, reversing the entire shipping mandate. Sellers celebrated. Most of them missed 3 things that are still active right now. Yes, restriction, a performance metrics enforcement, and an operational gap that most multi-channel sellers have no idea exists. But the part I really want to see is Amazon's response. They're launching a pricing program 6 days before TikTok dropped the mandate. The timing This isn't a coincidence, and if you're doing any volume on TikTok shop, the numbers are worth 5 minutes of your time. It's the same pattern, different platform. Platforms take control, and the brands who see it coming build around it before it hits. That video is being promoted somewhere up above right now, so I'll see you there.