Backlinks are the backbone of traditional search rankings. They are also one of the most misunderstood signals in AI search. The short answer: backlinks still matter, but the mechanism through which they create value has fundamentally changed. Understanding the difference is essential for any AI visibility tracking strategy.
This is part of the broader comparison of AI search vs traditional search and our complete guide to AI visibility tracking.
Backlinks in Traditional Search
In traditional search, the role of backlinks is well-understood. Google’s algorithm uses links as votes of confidence — each link from an external site signals that your content is worth referencing. More links from higher-authority domains means higher rankings.
The optimization model follows directly: earn backlinks through outreach, content marketing, digital PR, and guest posts. Measure success by tracking referring domains, domain authority scores, and anchor text distribution. The relationship between backlinks and rankings is one of the most studied and validated signals in SEO.
Traditional search evaluates link quality through PageRank and its derivatives. The system explicitly rewards link volume from authoritative sources, making backlink acquisition a core investment for any SEO program.
Backlinks in AI Search
AI search engines do not use PageRank to decide what to cite. LLMs operate on two layers — training data and real-time retrieval — and backlinks influence each layer differently.
The Training Data Layer
Links matter here, but through a different mechanism. Common Crawl uses Harmonic Centrality — not PageRank — to decide which domains to crawl most frequently. Harmonic Centrality measures how close your domain is to all other domains in the web graph, weighted by inverse distance. It rewards structural connectivity, not raw link volume.
This means one link from Reddit, GitHub, or a major industry publication moves your web graph position more than hundreds of links from isolated sites. The goal is not accumulating links — it is shortening the average number of hops between your domain and the rest of the web.
The Retrieval Layer
When AI engines like ChatGPT Search or Perplexity retrieve sources in real time, they draw from search indexes (Bing for ChatGPT, Google for Gemini, Perplexity’s own index). Traditional backlink signals influence these indexes, so strong backlink profiles still help with retrieval. But the final citation decision depends on whether the content meaningfully answers the user’s query — not on how many links point to it.
Research confirms this disconnect: 90% of ChatGPT’s most-cited sources come from beyond position 20 in Google’s organic rankings. A DA 90 site can be invisible to AI engines if its content is not structured for extraction.
The Shift from Links to Mentions
One of the most significant shifts is the growing importance of unlinked brand mentions. In traditional search, an unlinked mention carries minimal SEO value. In AI search, it matters substantially.
A 2024 survey of SEO professionals found that 73% believe backlinks affect visibility in AI search results, 79% say even nofollow links move the needle, and 81% report that unlinked brand mentions help build authority in AI contexts. LLMs build brand-topic associations from co-occurrence patterns in training data — your brand name appearing alongside relevant keywords across multiple authoritative sources creates the semantic associations that drive AI citations, regardless of whether those mentions include hyperlinks.
As one industry analysis framed it: “By 2026, mentions are the new backlinks.” Traditional link building focused on optimizing for crawler signals is being replaced by citation engineering and strategic PR focused on getting referenced across external platforms.
What Still Works Across Both Systems
Not everything changes. Several link-building activities create value in both traditional and AI search simultaneously:
| Activity | Traditional Search Value | AI Search Value |
|---|---|---|
| Guest posts on industry publications | Backlink authority + referral traffic | Brand mention in training data + web graph edge |
| Reddit participation | Minimal direct SEO value | High — Reddit has top-tier Harmonic Centrality and is heavily weighted in training data |
| Data-driven original research | Earns natural backlinks | Generates citations, references, and brand-topic co-occurrences across the web |
| SaaS directory listings (G2, Product Hunt) | Referral traffic + backlink | Web graph connectivity from structurally embedded platforms |
| Wikipedia-adjacent sourcing | Indirect authority signals | Strong training data influence — Wikipedia is the most-cited source by ChatGPT |
Rethinking Link Building Metrics
Traditional link building success is measured by backlink count, referring domains, and DA changes. For AI visibility, the metrics that matter are different:
- Harmonic Centrality rank — Track via Common Crawl Web Graph Statistics to see your structural web graph position
- Citation frequency — How often your brand appears in AI-generated responses, trackable through platforms like PhantomRank
- Brand mention volume — How frequently your brand is discussed across authoritative platforms, including unlinked mentions
- Share of voice — Your citation rate versus competitors for the same queries
Backlinks are not dead for AI search. But the playbook needs to evolve from chasing link volume toward building structural connectivity, brand-topic associations, and cross-platform presence. For more on how link building must adapt, see our guide to why link building must evolve for GenAI. For the broader discipline, explore our complete guide to AI visibility tracking.