Why changing AI citation data calls for a smarter B2B communications strategy

Every few months a study from a vendor is released and promises to tell us what AI is reading. The findings change, the recommendations are different, and B2B communicators are left wondering what is true.

We learn which sources AI tools cite when answering certain questions, but not everything those tools have read or used to generate an answer. The results depend on the questions being asked, the platforms being tested, and how researchers define their sources.

Rather than chasing every new citation trend, B2B communicators should focus on something more lasting: understanding what their buyers want to know and making sure their brands provide credible, consistent answers throughout the buying journey.

22%

journalist citations

17%

Owned + paid media

84%

earned media

Why do AI citation studies keep changing?

In October 2026, Generative Pulse, a product of Muck Rack, released a report called “What is AI reading?” If you are as in the weeds about AI visibility as we are, you know that is the same name as a report Muck Rack released in May. So what changed in five months?

At a glance, journalist citations are down from 27% to 22%. The report says most of those five points moved to other earned sources, not to paid content. Owned media, meaning a brand’s own content, made up 15.2% of citations. Owned and paid content together rose about one point, to roughly 17%, and earned media is still the largest category at about 84%. However, Muck Rack’s definition of “earned” is very broad. It includes journalism, academic research, government sources, Wikipedia, social platforms, and third-party corporate content. This is wider than the coverage we typically secure with media relations outreach, which makes it hard to measure.

The studies also don’t measure identical sets of prompts and platforms. For example, May’s report examined three major AI providers, while October’s included four. In other words, these numbers are useful for identifying patterns, but they shouldn’t be treated as a universal formula for what AI is actually reading.

What can B2B communicators take away from this report?

The October report found that the question shapes which pages get cited. Practical guides made up 68% of citations on how-to questions, while best-of prompts favored buyer’s guides and reviews at 45%. This gives us stronger evidence for connecting our Generative Engine Optimization (GEO) strategy to different stages of the marketing funnel.

At SVM, we are mapping the questions potential clients ask as they move from awareness to a decision.

Here are some examples based off of the stage in the buyers journey

The goal isn’t to guesstimate every single query or prompt and make a blog or a post about it. The goal is to have messaging targeted to each stage of the funnel to help inform your brand’s narrative intelligence.

Only 212 websites were cited against all four platforms, which reinforces the importance of a diversified strategy of earned, owned, and paid content.

What does this all mean for health IT communicators?

Figure out what your buyers are looking for and use that to build out your messaging and AI visibility platform. This is called narrative intelligence, and it focuses on making sure that the AI tools and your buyers are getting the same information and story about your brand. Google says there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode, and that SEO best practices still matter.

No one can guarantee that the LLM that recommended you today will do it again next week, but you can make sure the story it finds is accurate, consistent, and backed by evidence.

Don’t choose between earned media and owned content. Build one connected story that helps people discover, understand, and trust your brand. Our EARN framework is a good place to start.

Want to learn more about how to have AI start reading you? Reach out to us at [email protected]