Digital PR for AI visibility: Why mentions matter more than backlinks

Why do mentions matter more than backlinks for AI visibility?

Mentions of your brand on other sites correlate about three times more strongly with visibility in AI answers than backlinks do. In the Ahrefs study of 75,000 brands (December 2025), web mentions reach a correlation of r = 0.664 with AI visibility, backlinks only r = 0.218. The strongest single factor is mentions in YouTube videos, at r ≈ 0.74.

The reason lies in how language models work. A backlink is a technical signal that Google's classic ranking algorithm treats as a recommendation. A language model reads text instead, and it gets to know a brand when that brand is named in many independent sources in the right topic context. Whether the mention is linked barely matters to the model. What counts is that brand, topic and statement appear together in the text and are confirmed across several sources.

What is digital PR, and what does it have to do with ChatGPT?

Digital PR is the systematic building of editorial mentions in online media, trade portals, industry directories and on platforms such as YouTube, so that third parties write and talk about your brand. Unlike classic link building, the goal is the mention in the right topic context, not the link.

For AI assistants, that is exactly what counts: 84% of all AI citations come from earned media, while paid and promotional content accounts for only 0.3% (Muck Rack, May 2026, more than 25 million source links analyzed from ChatGPT, Claude and Gemini). So AI assistants mainly cite what journalists, specialist authors and independent publishers have written about companies, not the companies' own websites.

If you're asking how your company gets mentioned in ChatGPT, the honest answer is: less through your own website, more through what others publish about you. A well-built website remains the foundation (our guide What is GEO and how does AI search work? explains what it needs), but the biggest visibility lever sits outside your own domain.

How much does digital PR affect visibility in AI answers?

Targeted news distribution increases how often a brand is cited in AI answers by a median of 239 percent. That was measured in a 30-day study by Stacker and Scrunch with 87 stories from 30 companies, across 2,600 prompts on 8 AI platforms (March 2026). It is the strongest documented effect of a single GEO measure so far: 97 percent of the distributed stories received at least one AI citation, compared with 82 percent for owned content alone.

Here are the main documented effect sizes side by side:

All of these figures come from US and English-language studies. For Switzerland and the German-speaking market they are guide values, not guarantees. What stands out is how consistent they are: measures that build on third-party sources and verifiable statements beat purely technical levers in every study.

What are ghost citations, and how do you prevent them?

Ghost citations are AI citations where your content serves as a source but your brand isn't named. According to an analysis by Kevin Indig of 3,981 domains, 61.7% of all AI citations don't mention the brand. The AI uses your knowledge anonymously.

The way out is the format. Brands get named mainly when the source explicitly compares, rates or recommends them: “Provider X vs. Provider Y”, “The best tools for Z”, product tests, rankings. Purely informational guides feed the machine with facts without making the sender visible.

For digital PR this means that not every mention has the same value. A placement in a comparison article or a “best of” list in a trade publication is far more likely to get your name into the AI answer than a topic article in which your company only appears in passing. If you plan digital PR for AI visibility, prioritize recommendation and comparison formats.

How do you build mentions that AI systems actually see?

Mentions only work if AI crawlers are allowed to read them, your core message is consistent and the source fits the target engine. In practice, four points decide the outcome:

1. Check the target publication's robots.txt before every placement. 75% of the news websites studied block at least one AI crawler such as GPTBot or PerplexityBot (Screaming Frog, 72 news sites, 2025). A mention in a publication that blocks the bots of your target engine is invisible to ChatGPT and the like, and the PR win does nothing for your AI visibility.
2. Keep your core message word-for-word consistent. Language models give more weight to statements they find in identical form across several sources. Describe your positioning (who you are, what you offer, where you are based) in the same words in press releases, on your website and in profiles, instead of rewriting it each time.
3. Choose the right sources for each engine. Each AI engine draws on a different pool of sources: 28.6% of Perplexity citations come from Google's top 10, for ChatGPT it is only about 8%, and 80% of all URLs cited by AI don't rank in Google's top 100 for the question at all (Ahrefs, 15,000 queries, 2025). A mention that Perplexity sees does little for ChatGPT. First check which publications the target engine actually cites for your relevant questions, and place your story there.
4. Keep adding fresh mentions. 65% of AI bot visits go to content less than one year old (Seer Interactive, 2025). A one-off PR push fades. A steady rhythm of new mentions keeps your brand within the models' relevance window.

Are backlinks now worthless for AI search?

No. Backlinks remain a central signal for classic Google rankings, and those rankings are still one of the strongest predictors of whether a page gets cited in AI answers, because the assistants check their answers against search indexes such as Google and Bing. If you drop link building entirely, you undermine your own foundation.

What has changed is the weighting: the mention gets you named in the AI answer, the link helps the linked page rank. The two goals don't exclude each other. The best digital PR placement delivers both: an editorial mention in the right context plus a clean link. As an SEO & GEO agency, we therefore plan both signals together instead of playing them off against each other.

One warning: mass-bought mentions, forum spam or fake Reddit posts don't work. Reddit deletes around 25,000 such spam posts every day (2026), and detected manipulation can even count against you in AI data. Only mentions with real editorial context pay off over time.

What does this mean for Swiss SMEs?

For Swiss SMEs, the gap is an opportunity. Most competitors still optimize only for Google rankings, while the shortlist for provider questions is increasingly drawn up in ChatGPT, Perplexity and Google AI Overviews. If you start building mentions systematically in Swiss trade media, industry portals and regional publications now, you take citation spots that will be expensive to win back later.

A practical start doesn't need a big budget. First, measure whether and where your brand appears in AI answers today. We track this continuously for our clients, separately for each engine and across many queries. Second, identify the most important customer questions where AI currently names competitors or no Swiss provider at all. Third, go after a small number of placements in publications that allow AI crawlers and are cited by your target engines. Quality beats quantity.

Our LLMO service shows how building mentions, structuring content and monitoring fit together into ongoing support. If you want to understand the basics first, start with our guide What is GEO?

As of July 2026 · Author: Baris Gündogdu, founder of SEOX GmbH

Sources: Muck Rack, Generative Pulse, May 2026, muckrack.com; Ahrefs, AI Brand Visibility Correlations, 75,000 brands, Dec. 2025, ahrefs.com; Stacker/Scrunch, citation lift study, March 2026, stacker.com; Kevin Indig/Semrush, ghost citations study, semrush.com; Screaming Frog, 72 news sites, 2025, screamingfrog.co.uk; Ahrefs, AI Search Overlap, 15,000 queries, 2025, ahrefs.com; Seer Interactive, Content Recency, 2025, seerinteractive.com; Princeton GEO study, KDD ’24, arxiv.org; The Next Web on Reddit spam, 2026, thenextweb.com.