# Press or Website? It Depends on What Your Future Student Is Asking > AI doesn't cite the same way for every question: press for industry trends, your website for how-to and program choice. What the data says, and Griddo's take. **URL:** https://griddo.io/en/journal/press-or-web-ai-answers/ **Language:** English **Published:** 2026-11-11 **Author:** Carmen Martin **Tags:** geo, digital-strategy --- ## Key Takeaways - **AI doesn't answer every question the same way:** for industry-trend queries it cites journalism in 46% of responses (Muck Rack, May 2026); for "how-to" questions it leans more on reference content and brand-owned material. - **More than 84% of AI citations are earned and third-party sources**, but that category mixes journalism, encyclopedias, and academic research — journalism alone was 22% of citations in September 2026 (Muck Rack, October 2026). - **Our position at Griddo:** for a university, its own website outweighs any other owned channel as the source of truth for "how-to" and program-selection queries. This is our inference, with no study measuring it in higher education. - **Earned media is still the foundation of reputation.** We propose splitting effort by query type, not abandoning the press. --- ## What a Future Student Asks Changes Where the Answer Comes From Think of three questions a family might ask an AI assistant. "Which universities are investing in artificial intelligence?" "How do you apply for a scholarship in a communications master's program?" "What's the difference between two degrees at the same university?" All three are about higher education, but they ask for different things. The first asks for industry context: who stands out, and what's being said about it. The second and third ask for detail: deadlines, requirements, curricula, career outcomes. In our view, that detail is rarely well covered in a press article. If your team splits effort between earning press coverage and tending to the website, that difference matters. At Griddo, we think press is the foundation of reputation, and that, at the same time, visibility is won or lost on the website itself: it's where the source of truth lives for what a future student needs to decide. This article asks when each one carries more weight, and what data can back that up. ## What the Muck Rack Report Measures, and What It Leaves Out Muck Rack is a PR software vendor, and in our view its [*"What Is AI Reading?"*](https://muckrack.com/blog/what-is-ai-reading-october-2026) report has an obvious commercial interest: it serves to argue that earning media matters. We use it with that caveat in mind. The headline figure we cite comes from its fourth edition (October 2026), which draws on 2026 Generative Pulse data covering more than 40 million AI responses and 207 million citations, plus a September analysis of 1,020 prompts across ChatGPT, Claude, Gemini, and Perplexity. By that edition, more than 84% of citations come from earned and third-party sources, a split that has held between 82% and 89% across all four editions. The [May 2026 edition](https://muckrack.com/blog/what-is-ai-reading-may-2026) used a different base: more than 25 million links, 17 industries, and three platforms (ChatGPT, Claude, and Gemini). That's why we don't compare figures across editions as if they were the same series; we only use the comparisons the October edition itself publishes. There are three things worth reading carefully before taking that figure to a committee. #### Three Caveats Before You Take That Figure to a Committee - **"Earned" is a very broad category.** It includes journalism, government and NGO sites, academic research, encyclopedic sources, and third-party corporate content. Journalism alone was 22% of citations in September 2026, down from 27% in May. An 84% "earned" share doesn't mean eight in ten citations are press. - **"Owned" has a narrow definition.** In the May edition, Muck Rack defined owned content as content created by the company or product the query refers to; a blog or corporate content that isn't from that company counts as earned. The October edition doesn't repeat that definition. By it, owned content was 15.2% of citations in September 2026, up from 13.7% in May, and paid content and advertorials were 0.3%. That's a global figure, not one broken down by education or query type. So the owned-content figure can't be read as "what a university's own website contributes." - **In May, the study didn't measure queries that name the brand.** Muck Rack used unbranded, top-of-funnel questions, and excluded branded ones because, in its words, they are "essentially a shortcut to visiting a company's website." That's the vendor's own reasoning, not a measured result. For October, the note says the 1,020 September prompts were unbranded consumer queries; for the Generative Pulse data, the main basis for the headline figure, it isn't stated whether branded queries are included. In both cases, what's left out is exactly the case of a future student who already knows your university and is asking about it. ## Industry Trends vs. "How-To" This is the part of the report we find most interesting. In the May 2026 edition, industry-trend queries cited journalism in 46% of responses, more than double the share for "how-to" and comparison queries, and press releases appeared almost exclusively in trend-related ones. That figure is from May and isn't compared with October's, which use a different base. When the query is about how to do something, Muck Rack says in that edition that AI leans on reference content and the brand's own material. It's worth being precise about the second part. The May report gives that line with no percentage breakdown of owned content by query type, and it doesn't distinguish owned from third-party content within the measured corpus. In October, Muck Rack says, again with no percentage, that product pages and useful guides are often cited in product and "how-to" queries. That's qualitative support. We read it as a clue in favor of our thesis, not as proof. Other studies point in the same direction, with their own limits. [Foundation Marketing and AirOps](https://foundationinc.co/lab/branded-unbranded-ai-citations/), two companies that sell AI-visibility services, analyzed 50 B2B brands. When the query names the brand, 77.6% of responses include at least one citation from its own domain. With unbranded queries, only 2.2% of cited links were from brand domains. [Yext](https://www.yext.com/research/articles/ai-citations-user-locations-query-context/), another vendor, measured 6.8 million citations in location-based business queries: 44% went to the brand's own websites and 42% to directory listings. In its unbranded, objective queries, owned websites and local pages together made up nearly 60% of citations, while branded or subjective queries leaned more on listings and reviews. Neither of those studies is about education. Foundation's and AirOps' figures use different units (responses in one case, links in the other), so they don't compare with each other. What they do show is that the weight of a brand's own website changes with the type of query. ## Griddo's Position for Universities, and Its Limits With those pieces, we form our thesis: earned media carries more weight in reputation and category queries; in specific queries that call for evidence, like "how-to" questions or the details needed to choose a program, agents turn to owned media, and among those, the website is the most important, especially for a university. This is our inference. We haven't found any study that measures citations to owned content by query type in higher education, and we say so precisely so no one cites it as a measured fact. There's also evidence that qualifies it. We don't have higher-education data from Spain or Latin America: what we've found isn't from those markets. Within that limit: - In the Education sector of the May 2026 Muck Rack report, journalism is 18.9% of citations, the lowest of any sector (tied with Industrials and Manufacturing). AI replaces it mostly with ranking platforms like US News, Niche, or Research.com, and Wikipedia is ChatGPT's leading domain in Education. - [Gradial](https://gradial.com/blog/higher-education-ai-search-visibility-study), a GEO services vendor, analyzed 51 higher-education institutions with 20 queries across seven AI providers. The average owned-domain citation rate was 10.5%, against a 35% brand-mention rate. Gradial doesn't define how it counts citations or state whether the queries named the institution. - When AI cites a source in a higher-education answer, it's rarely a .edu domain. In other words, in education AI frequently cites third parties. That fits with the importance of reputation and rankings, and our thesis allows for it. We're talking about which source AI consults when the question calls for a specific fact about the university, and that's where we believe the university's own website makes the difference — on one condition: that it holds complete, consistent, and current data. If it doesn't, AI will look for it elsewhere. ## Interest in GEO Is Real On our own website, we see demand for the topic. Between August 13 and November 8, 2026, the query "geo" had 46,204 impressions and 55 clicks in griddo.io's Search Console, with an average position of 6.6. In that same window, our article [What Is GEO and Why Is It Changing the Future of Digital Positioning?](/en/journal/geo-future-digital-positioning/) drew, counting only its main URL and excluding anchored URLs, 78,290 impressions and 186 clicks. That's data from our own website over that window: it speaks to interest in the concept, not to how much AI cites a university. ## What to Do With Your Website If you share our hypothesis, or want to test it, there's concrete work that doesn't compete with your press strategy: #### Concrete Work That Doesn't Compete With Your Press Strategy 1. **Decide which "how-to" and program-selection questions your website should answer.** Deadlines, requirements, scholarships, curricula, outcomes. These are the questions a future student asks once they're already comparing. 2. **Keep a single version of every fact.** If the admissions deadline differs across three pages, a model can't know which one is right. We go deeper on this point in our article on [information architecture and AI visibility](/en/journal/information-architecture-ai-visibility/). 3. **Make your website readable and citable.** At Griddo, we argue that fragmented, poorly described content in a generic CMS can't be reliably read or cited by AI, and that a unified, fully described knowledge base can. That's our position. 4. **Publish content in formats a model reads well.** Griddo publishes every page as markdown alongside the HTML, and its knowledge base as an [OKF bundle](/en/journal/open-knowledge-format-machine-experience/) at `/okf/`; griddo.io's is live, with an interactive graph visualizer. Since the Guiding Gingko release, in October 2026, llms.txt and each page's markdown version are managed from SEO Settings, in the editor. 5. **Keep working the press.** Earned media is still the foundation of reputation, and per Muck Rack's May edition, industry-trend queries are where journalism carries the most weight. If you want the full framework for bringing this to your university, it's in the [GEO Playbook for Universities](/en/journal/geo-playbook-universities/). ## In Summary "Press or website?" makes less sense than "what is my future student actually asking?" If they're after industry context, press carries the weight. If they're after the fact they need to decide, we believe your website does. The available data backs the first half with caveats, and the second only in part — which is why we present it for what it is: a Griddo position, waiting for someone to measure what actually happens at Spanish-speaking universities. ## Sources [Muck Rack: What Is AI Reading? (October 2026)](https://muckrack.com/blog/what-is-ai-reading-october-2026) | [Muck Rack: What Is AI Reading? (May 2026)](https://muckrack.com/blog/what-is-ai-reading-may-2026) | [Foundation Marketing × AirOps: Hidden Selection Phase](https://foundationinc.co/lab/branded-unbranded-ai-citations/) | [Yext: AI Citations, User Locations, & Query Context](https://www.yext.com/research/articles/ai-citations-user-locations-query-context/) | [Gradial: AI Search Results for Over 50 Colleges and Universities](https://gradial.com/blog/higher-education-ai-search-visibility-study) --- *Generated from [Press or Website? It Depends on What Your Future Student Is Asking](https://griddo.io/en/journal/press-or-web-ai-answers/)*