AI Citations Are Becoming the New Backlinks — But We Still Don’t Know How to Measure Them
For two decades, backlinks gave SEO teams a relatively understandable signal of online authority: another website linked to your page, analytics tools could discover the link, and marketers could compare domains over time. Generative search is creating a new version of that relationship. When ChatGPT, Perplexity, Gemini or another AI system cites a publisher, brand or research page inside an answer, that citation can place the source directly in front of a user at the moment a question is being answered. The problem is that the industry still lacks a universally accepted way to measure what that visibility is worth.
The shift is already visible in the product experience. OpenAI explains that ChatGPT search can include citations and a Sources panel, allowing users to open the webpages behind an answer. OpenAI also notes that search results and citations can be incomplete, outdated or incorrect, an important reminder that citation behavior is not equivalent to the comparatively stable link graph SEO professionals have spent years analyzing.
From link authority to answer visibility
A backlink exists on a webpage until somebody removes or changes it. An AI citation is more fluid. It can depend on the prompt, the engine, the model, the search index, freshness signals and the particular sources retrieved during that request. The same brand may be cited for one formulation of a question and disappear when the wording changes. That makes traditional rank tracking a poor fit for the new environment.
Emerging AI-visibility platforms are therefore experimenting with metrics such as citation frequency, brand mentions, source share, sentiment, competitor share of voice and AI-referred traffic. Recent measurement frameworks also recommend tracking the exact prompt, engine, timestamp and cited URL rather than treating a single generated answer as a permanent ranking. That approach reflects a basic reality of generative systems: visibility is probabilistic rather than a fixed position on a search results page.
The attribution problem
Even counting citations does not solve the larger question. A citation may build brand authority without producing a click. An AI system may use information from a page without visibly citing it. A user might encounter a company repeatedly in AI answers and later visit through search, direct traffic or another channel. Conventional analytics can capture some referral traffic, but they cannot reliably reconstruct that entire influence chain.
The volatility of citation sources makes the issue even clearer. Recent reporting has documented major swings in how frequently individual platforms are cited by ChatGPT, while OpenAI has said citation sources can evolve as its systems seek more relevant information. For marketers, this means optimizing around one publisher, platform or temporary citation pattern could be as fragile as building an SEO strategy around a single ranking factor.
A new metric without a standard
AI citations increasingly look like an important successor to backlinks because they represent something valuable: a machine choosing a source while constructing an answer. But citation count alone is unlikely to become the equivalent of Domain Authority, PageRank or organic position. A useful AI visibility model will probably need to combine citation frequency, prompt coverage, source quality, brand framing, consistency across engines and measurable downstream traffic.
Until analytics platforms and AI providers expose more standardized data, companies should treat AI visibility as a measurement discipline still under construction. The goal is no longer simply to rank first. It is to become a source that AI systems repeatedly find credible enough to retrieve, mention and cite — and then develop better ways to understand what those appearances actually contribute to the business.