Co-citations and unlinked mentions: the new links?

Short answer: Yes and no. Co-citations and unlinked mentions are not technically links, but they now do much of the work links used to do. In AI search they may matter more than backlinks: brand search volume correlates with LLM citations at 0.334 versus near-zero for backlinks, and Google has counted mentions as "implied links" since 2014.

For two decades, the link was the atomic unit of off-page SEO. You earned a hyperlink, PageRank flowed, rankings moved. That model is fracturing. Large language models like ChatGPT, Gemini and Perplexity do not crawl a link graph the way Googlebot does; they reason over text, and in that text a brand name sitting next to a topic carries weight whether or not it is wrapped in an tag. Meanwhile Google has spent a decade quietly building machinery to read mentions, not just links. This article explains what co-citations and unlinked mentions actually are, what the 2025-2026 data says about their impact, and how to earn them on purpose.

Co-citations vs unlinked mentions: not the same thing

These two terms get blurred together constantly, so let's separate them cleanly before we go further. They describe different signals, and the strategies to earn them differ too.

A co-citation happens when two pages or brands are referenced together by a third source. As Link-Assistant's SEO wiki defines it, a co-citation occurs every time two publications are cited together by any other document. If a roundup of "best link building tools" mentions both Ahrefs and your tool in the same paragraph, you have been co-cited with Ahrefs. Search engines read that proximity as semantic relatedness: you belong in the same neighborhood. Crucially, Vazoola notes co-citation does not require a hyperlink at all; it works through textual mentions, names and phrases that appear in similar contexts across the web.

An unlinked mention (sometimes called co-occurrence) is simpler: your brand name appears in content with no clickable link back to your site. "We switched to Notion last year" is an unlinked mention. No PageRank flows, but the entity association is created. As Ann Smarty puts it, because LLMs do not rely on links the way classic search does, co-occurrence (unlinked mentions) is as important as co-citation (link neighborhood) for AI visibility.

Google has counted mentions since 2014 (the implied links patent)

This is not a 2026 invention. The foundation was laid in US Patent 8,682,892, "Ranking search results," granted to Navneet Panda and Vladimir Ofitserov. The patent explicitly distinguishes two link types. An express link is a hyperlink a user can follow. An implied link is "a reference to a target resource, e.g., a citation to the target resource, which is included in a source resource but is not an express link." The patent states that links for a group "can include express links, implied links, or both."

In plain English, as DPOM summarized when the patent surfaced, Google's system can count mentions and citations of a brand, alongside regular links, when ranking a site. Bill Slawski's SEO by the Sea tempered the hype at the time, arguing entity mentions are good but are not a one-for-one replacement for links. Both things are true: a patent is not confirmation of live ranking use, but the conceptual machinery to read mentions has existed inside Google for over a decade.

What 2025-2026 data actually shows for AI search

Here is where mentions stop being theory. Multiple large studies in 2025 found that brand mentions and brand demand predict AI citations far better than backlinks do.

The 2025 AI Citation report from The Digital Bloom found that brand search volume is the single strongest predictor of LLM citations, at a 0.334 correlation coefficient, outperforming traditional signals like backlinks which showed weak or neutral correlation. The same report documented a striking gap: ChatGPT mentions brands 3.2x more often than it actually cites them with a link. The brand is being recognized as an entity long before it earns a clickable citation. The report also found sites appearing on 4 or more platforms are 2.8x more likely to surface in ChatGPT responses, and that only 11% of domains are cited by both ChatGPT and Perplexity, so spread matters.

Entity-SEO analysts push the gap even wider. Per Digital Applied's 2026 guide, brand mentions showed a 0.664 correlation with AI Overview visibility versus just 0.218 for backlinks, suggesting the classic link signal is meaningfully less predictive than brand-entity authority for AI citation.

The mechanism behind those numbers is community mentions. SE Ranking studied 129,000 unique domains and 216,524 pages across 20 niches using 100,000 prompts. Domains with minimal Reddit presence averaged about 1.8 ChatGPT citations; those with heavy Reddit engagement reached 7, roughly a 4x multiplier. Quora mirrored it almost exactly: 1.7 citations for low-mention sites versus 7.0 for sites with millions of mentions. These are unlinked mentions doing measurable work.

Why LLMs lean on mentions instead of links

The reason is architectural, not philosophical. A traditional crawler builds a graph of pages connected by hyperlinks and propagates authority across it. An LLM is trained on text and, at answer time, retrieves and reasons over passages. There is no PageRank step. What the model has internalized is co-occurrence statistics: which brands show up next to which topics, problems and competitors, and how often. A brand named in thousands of relevant contexts becomes a high-probability completion when a user asks for a recommendation.

That is why a Reddit thread naming your product with no link can outperform a dofollow backlink from a thin directory in AI search. Google's own systems are converging here too: per entity-SEO coverage, Gemini is trained on the Knowledge Graph, and your brand's representation in that graph, fed by co-occurring mentions, schema and a clear entity home, shapes whether you appear in AI Overviews and AI Mode. We cover the retrieval side in depth in how to get cited by ChatGPT.

How to earn co-citations and mentions on purpose

You cannot directly request "please mention us without a link." But you can engineer the conditions. The tactics overlap heavily with modern digital PR and entity building.

1. Get into the roundups your competitors are in. This is classic co-citation link building: target the "best X tools" and "alternatives to Y" articles where category leaders already appear. Being named in the same list inherits their topical neighborhood. Our digital PR guide walks through the outreach.
2. Become a source on Reddit and Quora. The SE Ranking data makes this non-negotiable for AI visibility. Answer real questions in your category, name your product where it genuinely fits, and accept that most mentions will be unlinked. That is the point.
3. Publish original data and quotable lines. The Digital Bloom found adding statistics improved visibility 22% and including quotations 37%. Data and quotes get re-stated by other writers, which is how unlinked mentions multiply. Our link building statistics hub is built for exactly this kind of citation.
4. Nail your entity home. A clear About page, consistent NAP, Wikidata presence and sameAs schema help Google and Gemini resolve which entity all those mentions point to. Mentions without entity clarity get wasted.
5. Earn DR55 editorial placements that double as both link and mention. A contextual placement in a relevant publication gives you the express link and the co-citation with whatever else that article discusses. You do not have to choose.

On the internal-strategy side, the same principle applies to your own site: dense, well-structured internal linking reinforces which entities your pages are about, helping AI and search systems map your topical authority.

So do backlinks still matter?

Absolutely. This is an "and," not an "or." Backlinks still drive classic Google rankings, still pass authority, and still bring referral traffic that builds the brand demand that, in turn, feeds AI citations. Wordtracker frames it correctly: co-citation and co-occurrence do not replace link building, they work alongside it to help engines understand context. The smart 2026 posture is to pursue links that also generate mentions and co-citations, and to stop treating an unlinked mention as a failed link-building attempt. It was never a failure; it was a different, increasingly valuable signal.

If you are weighing where to spend, our link pricing index helps benchmark placement costs, and for the editorial side specifically, see white-hat link building for SaaS and how to buy backlinks safely to keep the link half of the equation clean.

The bottom line

Co-citations and unlinked mentions are not literally the new links, but they are absorbing the role links played for AI. In generative search, where there is no PageRank to inherit, a brand named repeatedly in the right contexts is what gets recommended. The 2025-2026 data is consistent across independent studies: brand mentions and demand out-predict backlinks for AI citation by a wide margin. Build for both. Earn the link, but engineer the mention.