By: Jason Dodge
Bing just gave search marketers something we’ve been asking for. Actual data on how content performs inside AI-generated answers inside their new AI Performance Report you can find inside Bing Webmaster Tools.
While the data is still very new, this is the first look inside AI data being shared by a major search engine. Currently in a public preview (beta), Bing is showing grounding queries, citation counts, and page-level citation activity pulled in across Microsoft’s Copilot, Bing AI Summaries and select partner integrations.
Wanting to understand this more and gauge what early signals could even be telling us, we pulled data from fifteen different brands that provide us with a diverse cross section of industries. While there certainly will be larger datasets to work with, the goal was to provide early signals.
Sticking with our belief that client/brand data is their data, we’ll keep this anonymous, but here’s what we found.
The Dataset
Our quick sweep across the brands we included resulted in a dataset of just over 365,000 total AI citations across more than 2,500 unique grounding queries. All of this in an effort to provide a very broad view into what we’re seeing.
- Travel & Tourism: Includes sites such as destination marketing organizations (DMO), resorts, and other hospitality type brands
- E-commerce: Consumer products that are more specialty online retail, outdoor gear, automotive aftermarket, etc
- Manufacturing: Heavily B2B focused within manufacturing, technical engineering services and support products
- Higher Education: Institutions such as university and college types
- EdTech: Virtual learning and educational platform based brands
- Financial Services: Personal and commercial banking, inclusive of lending
- Government or Adjacent: Government supported business services
- Healthcare: Clinic, specialty care or dental
Leveraging internal AI tools to assist in processing, the data available covers how often a site’s content is being cited as a source in AI-generated answers, and which grounding queries are triggering those citations. Note: This isn’t click data or traffic data as Bing hasn’t released that yet nor do we know if they will. You could likely align this with click data at the page level and we’ll plan to review that in deeper analysis. The focus here is how an AI system considers your content to be trustworthy enough to reference it when generating answers.
Numbers At a Glance
Citation volume varied greatly across brands and domains. On the smallest size of e-commerce property, this was 450+/- citations on up to the largest within the EdTech space of 160,000 citations. Volume alone isn’t telling the story and we would encourage you to look at pattern spotting instead as these trends reveal how AI systems are actually using different types of content.
There are few items that jumped out immediately across all properties we reviewed.
AI Is Being Used as a Help Desk
The single strongest pattern we see in the data is that across nearly every single vertical, login queries, portal access, and product support queries are dominating citation volume.
Within the higher education sector alone, shows us student portal login accounts for 27% of the institutions total AI citations by itself. In the EdTech platform space, we see a similar pattern, with 17% of all citations tied to login and portal access queries alone. Digging into one e-commerce brand we found single product app queries drove 33% of its entire citation profile. Similar to financial institutions, 25% of the queries are related to login type requests, noting these as the second-highest citation driver.
What does all this mean? At surface level it’s easy to glance past the obvious that is behavioral shifts much bigger than search. People are using AI (assistants or agents) as navigational layers to get things they already know exists. Re-read that and then ask yourself how you’re currently answering for navigational, “brand” queries? Individuals are asking AI to take them somewhere, and AI is citing your website as the answer (hopefully).
The implications for a brand and a site owner is pretty clear. If your site has login portals, support documentation or product setup guides, that content may already be one of your biggest AI citation drivers.
What Should Your Brand Take Away From This Data?
Make sure these informational components of your brand are updated and current. At a foundational level, check them over for structure and make sure they actually help people when they arrive. This also should be your prompt to spot check (or go deep) whether or not your brand is actually owning the support conversation. Whether you’re a product or service-based brand, the best thing you could look at is whether or not you’re providing your existing customers with the best after sales support possible. Your job is to reduce the friction, structure your content that is logical, helpful, and fulfills for both the audience and the machine.
One Piece of Great Content Can Dominate Your Entire Citation Profile
Several properties reviewed in our dataset had their AI citation profiles shaped almost entirely by a single content area.
An example would be a business services organization (B2B business services) saw 87% of their total citations come from a single topic. In this case it revolved around state-level employment law and HR-related queries. The organization had published thorough, authoritative information about the requirements for the law, and AI systems then latched onto it across dozens of search variants. Meaning, different phrasings, abbreviations, and year-specific searches, which all pointed back to the same content.
On the financial and fintech side, we picked up on a smaller grouping of fewer than 20 unique grounding queries that captured roughly 16% of its entire AI citation activity. All of this came in from a single blog post around how to prevent spam phone calls related to mortgages and loans. This continues to reaffirm that producing content at scale isn’t as important as making sure it aligns with the appropriate audience. This one piece of practical, helpful content, punched well above its weight for generative AI search inclusion.
Within higher education, a legacy tool built by the university’s library (years ago) drove 30% of all citations, that even included dozens of misspelling variants that AI still resolved back to the tool. Meanwhile, the same institution’s academic research archives accounted for another 7% of citations, drawing those in from users who likely have zero connection the school at all.
What’s the lesson? You don’t necessarily need a massive content library to earn meaningful AI visibility if you’re hitting on the right subject for those seeking what it is that you do. A single, well-executed piece of content that thoroughly addresses a specific need can generate outsized returns in AI citations (currently). The key unlock is depth, accuracy, and genuine helpfulness. Sounds an awful like E-E-A-T and is in alignment with Google’s own quality rater guidelines.
Travel and Tourism – Well Positioned, but Have Gaps
Collectively, we reviewed 714 unique grounding queries for brands that fall within the destination marketing (DMO) space and resort/hospitality verticals. This delivered a unique viewport across a diverse query set, with travelers asking AI a wide-range of questions.
Events and festival-related asks were the strongest across citation drivers. Seasonal festivals especially, including those that are annual and reoccurring events, such as holiday programming or signature cultural events are consistently generating the highest citation share. One destination we reviewed in particular, showed event-related queries as driving more than 27% of all citations.
Intent around lodging told a different story and was surprisingly underrepresented. Accounting for less than 1% of all citations, which could be an early indicator that if you’re not directly involved on the hotel/property side, you likely are not going to be referenced or showing up in the AI citations.
All properties we reviewed in the travel vertical skewed heavily toward branded destination queries. Between 82% and 93% of all citations included the destination (regional) name, indicating the audience had a general idea of “where” they wanted to travel, but were researching deeper on the typical approach of “see, play, do,” that we have witnessed over time in the destination marketing space. This is a healthy signal of authority that DMOs bring to both generative search and traditional search. The lower non-branded citations suggest room to grow on the category-level of travel queries that don’t mention the name.
What does this mean in non-geek language? If you’re a DMO, you likely are being cited and mentioned as it pertains to the destination, your region, etc. But if the audience is unclear on where exactly they want to travel, and perhaps they’re just interested in warm destinations that are ideal for family adventure, there’s likely an opportunity for you to make improvements.
EdTech and Higher Education are AI Citation Powerhouses
By a wide margin, the education sector was the highest-volume vertical in our dataset, accounting for more than 75% of all citations across the educational properties combined. Meaning, when you combine higher education with “EdTech,” colleges and universities will continue to win out.
It’s not that the EdTech space is invisible by AI citations, not at all. This sector still represented 44% of our dataset, but it’s primarily product and brand that accounted for about 35% of the platform’s citations. What stood out to us the most was the sheer volume of misspelled queries that AI handles, and the fact that they can resolve them back to the correct site. If you have spent any time in the SEO industry, this shouldn’t come as a shock to you, especially with auto-complete being injected into search for some time now. Roughly 6% of all citations came from misspelled brand name variations alone, with another 17% tied back to login and portal access queries.
Higher education fell within the top 20% of our total dataset, which gave us a lot to review and start to parse out preliminary signals. Beyond the login and portal queries that dominate the top, institutions are earning meaningful citations in AI across different state-level educational programs. Whether it’s programs related to free tuition initiatives or funding programs, these all drove a significant share of the non-branded citations.
The inverse of intent queries around “how do I pay for college,” or more enrollment-related asks, was a university brand that’s more rooted in its utility content and niche academic resources. Any grounded query that was grouped into “free online tools” accounted for 30% of the citations, whereas open course materials was less than 10% of citations. Research and archive content accounted for 15% of citations. Collectively, these types of citations are outperforming traditional admissions and academic program content. Meaning, AI tools are citing an institutional brand for information it’s giving away, not what it sells.
The takeaway for marketers in the higher education vertical is that your institutions AI visibility is likely driven by resources you’ve never thought to optimize. Library tools, open course materials, and research archives that are dated from years ago and serve and audience far beyond your prospective students.
Category Authority Drives Citations for B2B Services and Manufacturing
B2B marketers can all probably relate and share in the sentiment that this is the smallest volume of overall citations in the dataset. Setting that aside, manufacturing and technical services within the B2B sector, while having smaller volumes, have notable signals in the sophistication of queries.
Note: Much of this will come down to the size and scale of a brand, and we recognize that. This is also all very preliminary and your findings will likely vary.
One manufacturer reviewed, provided us with 107 unique queries. While it’s a smaller number, take stock again in the fact that 76% of those came from non-branded searches. Microsoft’s AI systems were citing this brand for broad category queries about specific industrial processes and applications. For a B2B company, that type of top-of-funnel AI visibility is directly in-line with what we should expect from Generative Engine Optimization (GEO) strategies.
Up to this point, the bulk of what we have been reviewing has been strictly English speaking language queries. The same brand also showed grounding queries in four separate languages beyond English. This indicates the content is being cited in AI answers that also serve an international audience. For companies with global operations, multilingual AI citations are a signal worth tracking, and potentially investing in. This also dovetails back into making sure your foundational and technical SEO elements are solid. If you’re a global brand and don’t have a strategic approach for how you’re going to handle Hreflang, that’s a missed opportunity.
Smaller and more regional B2B technical services brands we included in the dataset showed modest citation counts, but also highly specific grounding queries that tiered directly back to their areas of expertise. Both of these cases, even with preliminary data, validate the strategies around focused, authoritative content in a narrow vertical can earn and win AI citations regardless of the company or brand size. If you’re looking to mitigate the AI takeover of local SEO, take note here.
E-Commerce Brand Authority Is Not What You’d Expect
Possibly one of the most widest spread in citation volumes is e-commerce. But the interesting find wasn’t just a relative scale, it was where the citations actually came from.
For those of you wondering, this comes from sites built on both Shopify, WooComerce, and Magento, with Shopify dominating the dataset. The largest of e-commerce property’s citations were overwhelmingly product support or adjacent. Meaning, app setups, login troubleshooting, connectivity questions, firmware updates, and device configuration.
This level of insight tells us that AI has become your brand’s support agent. They have access to the information at scale, your documentation, and are becoming the primary answer source. The word of caution here is to review how this impacts your traditional support page visits, teams, etc. as it’s replacing those components with AI-generated answers that are citing more than just the content itself.
Again, e-commerce sites are vastly different from one another, and we can see that in our dataset. We noted a brand to non-brand split being roughly 35% to 65%, meaning the majority of its AI citations came from queries about other brands (third-party products) that they carried. AI engines were citing this site when users asked about specific manufacturers and product categories, which is a leading signal that the site had built topical authority beyond its own brand. Product manufacturers take note here as this could mean you are not owning your own brand story, but your retailers could be.
Financial Services and Healthcare Provide Early Signals
Financial institution data within our set was compact. We were only able to review 20 different grounding queries. While limited, it still signaled toward helpful needs for logins and support. All of this is to be expected. The standout however, was the practical support content around stopping spam phone calls, which accounted for over 16% of that institutions citations across multiple query variants. This also showed us that smaller institutions, either at a regional or local level, can very much compete with national brands. For financial institutions, this is your edge. Connect with your customer service representatives and find out what the customer pain points are. Put that in your roadmap as something to strive toward and make it a meaningful win.
Within the healthcare sector, the most interesting signals were on access-focused queries. Meaning the intent was with people who were searching for providers who accepted their form of insurance, payment, etc. In addition, there were standouts as the intent signals clearly defined individuals as new patients seeking a provider accepting new patients, and variants of “help me find care” type queries. AI was citing an organization specifically when people asked about finding affordable care, and affordable care near them (regionalized searches. Educational content about basic preventive health also performed well, with multiple query variations all pointing back to the same content.
Both verticals suggest that there is opportunity for local and regional service providers to gain exposure through practical, access-oriented content. AI is more likely to cite you directly if you’re owning the conversation around how people can find you, what forms of payment (and insurance) you accept, and how you solve common problems.
Key Takeaways from Early AI Citation Signals
AI systems are rewarding utility above all else right now.
Across every vertical, the content earning the most citations isn’t promotional, it was useful. This isn’t a case of brand storytelling and “vibes.” Useful content such as online tools, support, documentation, how-to guides, login portals, and practical advice is what consistently outperforms marketing-oriented content. If your site helps people do something, the probability that AI cites you will increase.
Support content might be your biggest AI Asset.
If one pattern presented itself the most, it’s that of the “AI as a help desk.” This was prevalent across e-commerce, EdTech, higher education, and financial services. If your site has support documentation, setup guides, FAQs, or portal access pages, you should be auditing them for accuracy and structure. These resources are likely already being cited, and you want to own as much of that narrative as possible.
Citation profiles can be carried by a single piece of content.
Contrary to what you might have been sold in the past, you don’t need a massive content operation to earn meaningful citations. You do however, need depth on the topics that matter most to your audiences. The facts don’t lie here either. We found that 87% of a site’s citations all referenced a state employment law explainer principles. A consumer advice blog post drove 16% of their citations. Free tools pulled in over 30% of citations for another brand.
Brand citations are the floor, not the ceiling.
Most sites have earned the majority of their branded queries. This ties back into the blunt fact that LLM knowledge cutoff dates are not discussed enough in marcom conversations. The real generative engine optimization (GEO) opportunity is non-branded citations. Earning the visibility when AI answers questions that don’t include your name. The sites and brands doing this well now, already have deep content that covers the broader topic space around their products or services.
Misspellings and query variants matter.
Google has handled misspellings since it was introduced to the search experience back in 2004. AI systems are handling the messy input of humans far better these days. While it’s not something you need to be optimizing for, you should take stock in how frequently it happens as it underscores the fact that AI is resolving the user intent, not matching keywords.
Citation data is NOT business outcome data.
Let’s be clear about what this review does and doesn’t tell us. It shows how often your content is cited in AI-generated answers. Period. It doesn’t show whether those citations drive clicks, traffic, or even conversions. That gap matters when you’re looking to align your digital marketing data and plan for larger conversations when C-suite sees you’re showing up in AI overviews, but you still can’t justify “why” that is important. For now, it’s a leading indicator of AI visibility, nothing more and nothing less.
What Comes Next?
What comes next and what marketers can do with this data are two great questions to be asking. Consider this your “day one” with data available from a tool that’s in public preview (beta). The sample sizes will continue to grow, the metrics will evolve and we might be able to expect Bing to eventually connect citation activity to traffic and business outcomes. Savvy marketers have been listening and are taking steps to model visitors from AI tools and how this is impacting their brand. Tip: create custom channel groups inside of GA4 now to see this.
In the meantime, make sure you are not sleeping on Bing Webmaster Tools. If you haven’t already, this is an easy sign-up and you can even import your site directly from Google Search Console. This will allow you to start to pull your AI performance data and building a baseline. As always, you can’t optimize what you don’t measure, and for the first time we have something directly from the engines that you can point back to.
If you’re not sure what to make of your data or you want help building a strategy around AI visibility and Generative Engine Optimization, that’s what we do. Reach out and let’s look at it together.
