The Number Nobody Expected
I run a portfolio of WordPress sites. One of them — a media property publishing articles about AI tools, local business intelligence, and content strategy — started showing up in a place I didn’t expect: inside Microsoft Copilot’s answers.
Not as a search result. Not as a backlink. As a citation — the source that Copilot grounded its response on when enterprise users asked questions inside Word, Edge, Outlook, and the Copilot sidebar.
The Bing Webmaster Tools AI Performance tab — still in beta, still barely documented — told me exactly how much: 98,800 AI citations across 576 unique grounding queries in under 90 days.
That’s not a typo. Ninety-eight thousand, eight hundred times an AI engine pulled content from my site and embedded it in a response to a real user. And here’s the part that flipped my understanding of content economics: during that same period, the site received roughly 1,900 human clicks from Bing search.
The AI was reading my content 52 times more often than humans were clicking on it.
What the Bing AI Performance Tab Actually Shows
Most marketers don’t know this tab exists. It appeared in Bing Webmaster Tools sometime in late 2025, buried under the Performance section. Microsoft labeled it “AI Performance (beta)” and didn’t announce it with any fanfare. No blog post. No keynote mention. It just showed up.
Here’s what it tracks:
Citations: The number of times your content was used as a grounding source in a Copilot-generated response. This isn’t an impression — it’s a direct attribution. Copilot pulled from your page, used your information, and (in many cases) linked back to you as the source.
Grounding Queries: The actual questions users asked that triggered your content to be cited. These aren’t keywords — they’re natural language questions. Full sentences. “What is claude ai pricing in 2026.” “How do I connect Claude to Notion.” “What’s the difference between Claude Code and Cursor.”
Daily Trend Data: The day-by-day citation count. This is where the story gets interesting.
The Growth Curve That Changed My Strategy
When I first noticed the AI Performance tab, my daily citation count was sitting at around 672 per day. Modest. Interesting, but not transformative.
Ninety days later, it was 5,500 citations per day. That’s an 8x increase with no corresponding change in my publishing cadence, no new backlink campaigns, no paid distribution. The content was the same. What changed was Copilot’s appetite for it.
The growth wasn’t linear. It came in steps:
Days 1-30: Steady at 600-800 citations/day. Copilot was discovering the site.
Days 30-50: Jump to 1,500-2,200/day. A handful of articles got locked in as preferred sources.
Days 50-70: Acceleration to 3,000-4,000/day. The site was now a default grounding source for an expanding set of queries.
Days 70-90: Peak at 5,500/day. Citation velocity was compounding — the more Copilot cited the site, the more queries it became eligible for.
This looks like a flywheel, and I believe that’s exactly what it is. Copilot’s grounding algorithm appears to develop trust in sources over time. Once a domain proves reliable for a topic cluster, it gets promoted for adjacent queries in that cluster.
The 576 Queries: What Enterprise Users Actually Ask
The grounding queries are the most valuable dataset I’ve ever had access to. They reveal what Copilot users — overwhelmingly enterprise workers inside Microsoft 365 — are actually asking when they invoke the AI.
The top query by citation volume: “claude ai pricing” — generating 16,500 citations on its own. One query. One article. Sixteen thousand five hundred times Copilot used my page as the source for its answer.
The next tier includes queries like “claude code vs cursor,” “how to use claude code,” “anthropic console guide,” and “notion mcp setup.” These are highly specific, tool-comparison, how-do-I-use-this queries from people who are actively working. They’re not browsing. They’re not exploring. They’re in the middle of a task and they need an answer right now.
This tells me something fundamental about who Copilot serves: knowledge workers making decisions inside productivity software. They’re writing a memo and need a pricing comparison. They’re evaluating a developer tool and need a feature breakdown. They’re setting up an integration and need configuration steps.
The content that wins Copilot citations isn’t SEO content. It isn’t listicles. It isn’t keyword-stuffed landing pages. It’s reference-grade material that answers specific operational questions.
What Roofing Articles Got: Zero
I also publish content in trade verticals — restoration, construction, and local services. Those articles have solid traditional SEO performance. Google sends traffic. The content ranks.
Copilot citations for those articles: zero.
Not low. Not “a few.” Zero. Because the people using Copilot in their daily workflow aren’t asking about emergency water damage repair or roofing contractors in Houston. They’re asking about the tools they use to do their jobs — AI platforms, development environments, productivity software, and business strategy.
This is the first data point that made me realize: AI citation optimization is platform-specific. The topics that win on Copilot are not the topics that win on Google, and they’re not the same topics that win on ChatGPT or Perplexity. Each platform has a different user base with different intent patterns.
The Raw Numbers, Laid Out
Here’s the data from one domain over approximately 90 days, pulled directly from Bing Webmaster Tools AI Performance (beta):
Total AI Citations: 98,800
Total Grounding Queries: 576
Average Daily Citations (start): 672
Average Daily Citations (end): 5,500
Top Single Query Citations: 16,500 (“claude ai pricing”)
Human Clicks from Bing (same period): ~1,900
AI-to-Human Ratio: 52:1
Top Content Type Cited: Detailed comparison/pricing guides
Content Types with Zero Citations: Local service pages, trade industry content
What This Means for Content Strategy
The industry is currently arguing about SEO vs GEO vs AEO. That argument is already outdated. What the data shows is something more granular: different AI platforms are different audiences, and they require different content strategies, the same way that smart TV advertising requires different creative than mobile advertising.
I’m calling this Platform-Specific AI Optimization (PSAO) because nobody else has named it yet. Nobody else has named it because nobody else is measuring it. The tools are there — Bing Webmaster Tools shows Copilot citation data right now — but the marketing industry hasn’t caught up to the idea that AI engines are audiences, not just algorithms.
Here’s what I’m doing with this data:
I’m writing content specifically engineered for Copilot’s enterprise user base during business hours — detailed tool comparisons, pricing breakdowns, integration guides, and operational how-tos. I’m writing different content for Google’s organic audience — local business directories, event guides, and community resources. Same domain. Two completely different content strategies running simultaneously.
The Copilot content doesn’t need to rank on Google. The Google content doesn’t need Copilot citations. Each serves its platform’s audience where they actually are.
Why I’m Publishing This
I’m publishing this data because the industry needs a baseline. Right now, there is no public benchmark for AI citation volume. No one is talking about citation-per-query rates, daily citation growth curves, or topic-platform fit analysis. There’s no equivalent of “Domain Authority” or “organic traffic” for the AI citation economy.
Someone needs to be first. I have the data. So here it is.
If you run a content operation and you haven’t checked your Bing Webmaster Tools AI Performance tab, do it today. You might be sitting on citation data you didn’t know existed. And if you’re building content strategy without accounting for which AI platforms are actually consuming your content, you’re optimizing for one audience while ignoring the one that’s reading you 50 times more often.
The AI citation economy is already here. The question is whether you’re measuring it.
Frequently Asked Questions
What are AI citations in Bing Webmaster Tools?
AI citations are instances where Microsoft Copilot uses your website content as a grounding source in its responses to user queries. They appear in the AI Performance (beta) tab within Bing Webmaster Tools and represent direct attribution — Copilot pulled information from your page and used it to construct an answer for a real user.
How do I check my AI citation data?
Log into Bing Webmaster Tools, navigate to the Performance section, and look for the “AI Performance” tab. It’s currently in beta. You’ll see total citations, grounding queries (the actual questions users asked), and daily trend data showing how your citation volume changes over time.
Why does Copilot cite some content but not others?
Copilot’s user base is predominantly enterprise workers inside Microsoft 365 applications. They ask operational questions — tool comparisons, pricing details, integration guides, and how-to content related to their daily work. Content that answers specific, task-oriented questions earns citations. Generic listicles, local service pages, and broadly targeted SEO content typically receives zero citations because it doesn’t match what Copilot users are asking.
What is Platform-Specific AI Optimization (PSAO)?
PSAO is a content strategy framework that recognizes different AI platforms serve different audiences with different intent patterns. Copilot users are enterprise workers mid-task. ChatGPT users are explorers and researchers. Perplexity users want curated multi-source answers. PSAO means creating content tailored to each platform’s user behavior rather than treating all AI engines as interchangeable.
Is AI citation data more valuable than traditional search clicks?
The data suggests AI citations represent a fundamentally different type of content consumption. With a 52:1 ratio of AI citations to human clicks, AI engines are consuming content at dramatically higher volumes. Whether this translates to direct revenue depends on your monetization model, but from a reach and authority perspective, AI citations may represent the larger audience for many content categories.
Leave a Reply