Not all AI citations are equal. The Machine Relations Index, published on November 2, 2026, measured whether citations from the major AI engines actually lead somewhere — and the gap is enormous. ChatGPT's citations resolve 82.8% of the time. Claude's resolve just 6.1% of the time.
The Machine Relations Index measured citation resolvability by engine — whether a citation in an AI answer leads to a page that can actually be reached and checked. ChatGPT came out far ahead at 82.8% of citations resolvable. Perplexity was second at 47.3%, less than half of its citations resolving. Gemini managed 14.3%, and Claude was last at 6.1% — meaning roughly nineteen out of twenty Claude citations did not resolve.
The index also measured the supported share: the proportion of citations where the cited source actually supports the claim being made. ChatGPT led there too, at 54.5%, with Perplexity close behind at 52.0%. Even for the best engine, then, only just over half of citations were supported by their source. For the engines behind them, the supported share was lower still — the index reported supported figures for ChatGPT and Perplexity, and the resolvability gap suggests the others trail on support as well.
The two measures are different things, and the difference matters. A citation can resolve — the link works, the page loads — without supporting the claim, if the page says something else entirely or nothing at all on the point. And a citation can in principle support a claim without resolving, if the source exists but the link is broken. What the index shows is that ChatGPT is best on both counts, Perplexity is competitive on support but weak on resolution, and Claude and Gemini are far behind on getting the user to a real page at all.
The index does not explain the causes, and the reasons are likely a mix of product design and engineering choices. ChatGPT's search and browsing features have been built around returning clickable, verifiable sources, and its citation system appears to be the most mature of the four. Perplexity was built as an answer engine from the start, which may explain its relatively strong supported share, even though nearly half its citations still fail to resolve.
Claude's 6.1% is the striking number. Claude is widely regarded as a strong model for reasoning and writing, but its answers in this measurement rarely produced a citation that led anywhere checkable. That may reflect a product that cites less, or cites differently — naming sources in text without linking them, for example. Whatever the cause, the practical effect is the same: a reader who tries to verify a Claude citation will usually hit a dead end.
Gemini's 14.3% sits between the two extremes but much closer to the bottom. Given Gemini's position inside Google Search, that is a notable result: the engine with the largest search infrastructure behind it is producing citations that resolve only about one time in seven.
For SEOs, resolvability changes how to value a citation. A citation that does not resolve is a mention without a destination: it may put your brand name in front of a user, but it gives them no path to your site, and it gives you no referral traffic and no verifiable proof that you were cited. When you report AI visibility to a client, a count of citations across all engines flattens a real difference. Eighty citations from ChatGPT and eighty from Claude are not the same thing — the first group mostly leads somewhere, the second mostly does not.
That argues for weighting your tracking by engine. If ChatGPT is where your citations resolve, then ChatGPT is where the traffic and the verifiable brand exposure live, and it deserves the most attention. Perplexity is worth watching for its supported share, which suggests its citations are at least grounded in real sources even when the links fail. Claude and Gemini citations should be treated as brand mentions first and traffic sources second, until their resolvability improves.
There is also a lesson about the content itself. If an engine's citations frequently fail to resolve, part of the problem may be on the publisher side: pages that move, URLs that change, content behind logins or paywalls, and JavaScript-heavy pages that a crawler cannot render. Keeping URLs stable, keeping key content accessible without a login, and making sure the cited passage is actually on the page at the cited address are all things a publisher controls — and they are exactly the things that decide whether a citation resolves.
The question is whether the laggards close the gap. Citation quality is a fixable engineering problem, and all four engines have reason to fix it: users who cannot verify an answer lose trust in the product. If Claude and Gemini improve their resolvability, the value of being cited by them rises sharply — and the sites that are already citable there will benefit first.
Until then, the honest way to report AI visibility is per engine, with resolution in mind. Count citations, but also check whether they lead anywhere. The Machine Relations Index gives you the baseline to quote: 82.8% for ChatGPT, 47.3% for Perplexity, 14.3% for Gemini, and 6.1% for Claude. Any claim about AI visibility that ignores those numbers is overstating what a citation is worth.
Sources: Machine Relations Index, November 2, 2026.
Related reading on this site: Wikipedia Leads ChatGPT Citations at 10.1% in Ahrefs' November Ranking, AI-Generated Images Are Returning Inside AI Overviews, November Core Update Collides With Black Friday, and a Clean Read Will Not Come Until December, Core Update Volatility Started Days Before Google Announced It, Gemini 4 Hydrogen in Search: Four Changes Google Actually Announced, and November Search Console Click Spikes May Be Bots, SEJ Reports.
Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.
Start free trial