Claude Citation Change Log: 463 of the Top 500 Domains, a Panel the Six-Source Cap Barely Touched
A dated record of what Claude cites and what changes. Entry one: Claude reaches 463 of the 500 most-cited domains, third-widest head overlap of six engines, while AuthorityTech's own daily panel — which caught Perplexity and Gemini badly undercounted — was already recording Claude close to whole: 4.8 of its true 4.8 source URLs, truncated on only 41 of 263 runs.
This log is a dated record of what Claude cites and what changes, read from the Machine Relations Index and AuthorityTech's own daily panel. Entry one: Claude reaches 463 of the 500 most-cited domains in AI answers — third of six engines, behind Perplexity's 490 and Gemini's 452 — and cites the single most-cited source in 72 of the 93 published question segments. The panel that tracks it day to day is the interesting story here for the opposite reason than it was for Gemini and Perplexity: Claude returns a mean of only 4.8 source URLs per answer, so the panel's six-item cap barely touched it. That is disclosed here, with the number that shows exactly how little, before this log uses the panel for anything.
Every entry states the measurement, the release or window it came from, the primary source, and what a publisher can check. Entries are appended; nothing is rewritten. Entry one draws on the Machine Relations Index, which observes six engines — Perplexity, ChatGPT, Claude, Gemini, Google AI Mode and Google AI Overviews — against a fixed basket of buying and research questions and publishes a machine-readable release daily. Entry one also covers the same methodological finding this log's ChatGPT, Perplexity and Gemini entries carry about AuthorityTech's own daily panel — the panel's citation log keeps only the first six sources per answer — but for Claude the finding runs the other direction: Claude is one of the two engines the cap barely bites.
Entry 1 — September 22, 2026: the baseline
Release mri_score_v2.0, contract machine_relations_index_public_view_v2.0, generated September 22, 2026. Window May 10 to September 22, 129 days, 16,307 observed answer runs, 34,720 cited-domain observations, 22,828 cited domains. A subject category paired with a question type forms a stratum, and a stratum publishes a rate only after clearing an evidence floor of 10 observed runs across 7 distinct dates; 93 of the 151 measurable strata are published today.
The Index records, for each domain, which engines cited it at least once in the window. It does not publish a per-engine citation rate, so everything below is presence and counts, not per-engine rates — the same boundary this log's companion ChatGPT, Perplexity and Gemini entries run under.
Third-widest head overlap of six engines
| Engine | Top 25 | Top 100 | Top 250 | Top 500 |
|---|---|---|---|---|
| Perplexity | 25 | 96 | 244 | 490 |
| Claude | 23 | 94 | 237 | 463 |
| Gemini | 23 | 89 | 227 | 452 |
| Google AI Mode | 24 | 91 | 227 | 444 |
| Google AI Overviews | 24 | 89 | 203 | 378 |
| ChatGPT | 22 | 88 | 203 | 361 |
Claude misses 37 of the top 500 most-cited domains in the market — the fewest of any engine except Perplexity, and 11 fewer than Gemini's 48. At the top-100 mark Claude ties Perplexity for the widest reach of any engine (94 of 100), and it holds that lead through the top-250 band (237, second only to Perplexity's 244) before Gemini and Google AI Mode close in by the top-500 mark. Read against the domain-count section below, Claude's pattern is the inverse of Perplexity's: fewer distinct domains cited overall, but a higher share of what it does cite sits in the head the market already concentrates on.
Fifth on segment leaders
For each of the 93 published strata, take the domain with the highest citation rate in that stratum — the source the engines cite most when buyers ask that category's question — then ask which engines cite that domain anywhere in the window.
| Engine | Cites the leading source, of 93 segments |
|---|---|
| Perplexity | 93 |
| Gemini | 86 |
| Google AI Mode | 81 |
| Google AI Overviews | 79 |
| Claude | 72 |
| ChatGPT | 66 |
Claude misses the segment leader in 21 of 93 published categories — ahead of only ChatGPT. This is the one composition measure where Claude's strong head-overlap numbers do not carry through: reaching deep into the top 500 overall is not the same as reaching the single most-cited domain inside a specific category, and on that narrower test Claude sits closer to the bottom of the field than the top.
Domain count, stated without an exclusivity claim
Claude cites 5,326 of the 22,828 domains observed in the window — fewer than Perplexity, Gemini and Google AI Mode, more than ChatGPT's 5,054 and Google AI Overviews' 2,279. This log's ChatGPT and Perplexity entries each carry a floor-adjusted exclusive-domain count for their own engine, computed against the specific release each entry used; the Gemini entry withheld one for the same reason this entry does. A floor-adjusted exclusive-share figure has to be computed against this same September 22 release to be comparable to the numbers above, and that computation has not been done for Claude. This entry states domain count and stops there, rather than publish an unconditioned exclusive-share number that this log's ChatGPT entry already had to correct once — most of an unconditioned count's weight sits on domains cited a single time.
Community platforms: tied for fourth of six
Twenty-seven classified domains in the Index are community and social platforms — a small set where engines diverge most.
| Engine | Community platforms cited (of 27) |
|---|---|
| Google AI Mode | 20 |
| Perplexity | 18 |
| Gemini | 15 |
| Claude | 10 |
| Google AI Overviews | 10 |
| ChatGPT | 8 |
Claude ties Google AI Overviews on community reach, well behind Gemini, Perplexity and Google AI Mode. Combined with the segment-leader table above, Claude's composition profile is uneven rather than uniformly strong or weak: it reaches deep into the head of the index by raw domain rank, but underperforms on the two measures — segment leadership and community-platform reach — that track breadth into specific categories rather than overall citation frequency.
Three named crawlers where Gemini has none, and Perplexity has two
Unlike Gemini, which has no dedicated fetch-time citation crawler and governs training and grounding through a single log-invisible Google-Extended token, Claude is the most granular of the four engines this log has covered so far. Anthropic's own crawler documentation, last updated April 7, 2026, names three separate, log-visible user agents, each controllable independently in robots.txt: ClaudeBot, which "collect[s] web content that could potentially contribute to their training"; Claude-User, which "supports Claude AI users" by fetching a page when someone asks Claude a direct question about it; and Claude-SearchBot, which "navigates the web to improve search result quality for users" — the crawler that builds the index behind citations. Disabling Claude-SearchBot "prevents our system from indexing your content for search optimization, which may reduce your site's visibility and accuracy in user search results"; disabling Claude-User blocks retrieval of a specific page a user asked about. A publisher can, in principle, opt out of training while remaining fully citable — block ClaudeBot, allow the other two — a policy this log's Perplexity entry showed is also available there, where PerplexityBot and Perplexity-User split the same way, but which Gemini's single combined token does not permit.
This is a structural comparison, not a measurement of any specific domain's robots.txt policy — that check, run at this scale, is future work for this log, not this entry. What entry one can state is the shape of the choice available to a publisher: Claude and Perplexity each expose two or three named tokens with distinct purposes; Gemini exposes one that folds training and grounding together; and the measurement below, on how much of each engine's answer the daily panel actually saw, is unrelated to any of these robots.txt policies — it is a property of the panel's own storage, not of what a publisher blocks.
What the daily panel was actually recording
AuthorityTech runs its own tracked panel alongside the Index: 35 fixed queries, five retrieval surfaces, once a day, the same instrument behind this log's ChatGPT, Perplexity and Gemini entries and the site's 89-day per-engine churn measurement. That page's September 22 correction found the daily log stores only the first six sources returned by each engine, and the cap does not bite every engine equally. Claude is one of the two engines it barely caught.
Measured against the stored per-run answer artifacts, which preserve each provider's complete returned source list — available from September 1, 2026 onward, not for the full panel history — Claude returns a mean of 4.8 source URLs resolving to 4.1 distinct domains per answer. The daily log's six-item cap exceeded that most of the time: Claude's answers went past six sources on only 41 of 263 runs in that window — 15.6% truncated — against Perplexity's 350 of 350 (100%), Gemini's 175 of 193 (90.7%), ChatGPT's 12%, and Google AI Mode's 0%. Of the four engines whose complete-list domain counts this log has now reported, Claude's 4.1 is the lowest, and its truncation rate is the second-lowest of the five surfaces the panel tracks.
The practical consequence, recomputed rather than assumed small: day-over-day domain churn on Claude's complete source lists for the September 1–22 window (453 day pairs) is 34.2%, against 34.7% on the panel's six-source cap — an overstatement of 0.5 percentage points, the smallest correction of the four engines with a usable September sample, in the direction the truncation predicts, with the cross-engine churn ranking unchanged. That correction and its full recomputation table live on the churn page linked above; this log will not restate the panel's older 89-day retention, drop-out, and core-domain figures as if they described the complete list, because the complete-list artifacts do not reach back past September 1 and the churn page says so explicitly. Any figure in this log drawn from the panel is labelled by which list produced it.
Where other measurements land, and why they stay off this log
Third-party trackers publish their own Claude citation-depth numbers, on their own prompt baskets, and they do not agree with each other or with this log's 4.8. Orbit Media's study of 13,184 citations across four engines reports Claude averaging 3.6 sources per answer, the lowest of the four it tracked, with brands mentioned in 70% of Claude's answers but cited in only 40% — a 30-point gap. A separate tracker cited by QuickSEO puts Claude at 5.67 citations per answer, attributing Claude's retrieval to a Brave Search backend. Ahrefs data reported via GEOxylia describes Claude as citing "rarely" with "the longest citation tail of any major engine," noting only 36% of Claude's journalism citations come from the past 12 months, against a faster-turning ChatGPT citation set. None of these figures is stated here as a benchmark for this log's own number; each comes from a different query basket, a different capture method, and in most cases an unstated measurement window.
This log's 4.8 comes from a fixed panel of 35 buying and research queries measured daily since September 1, checked against the complete per-run artifact rather than a display-layer or third-party scrape. It is the lowest source-depth figure this log has reported for any of the four engines covered so far — Perplexity's 20.8, Gemini's 10.8, and Claude's own 4.8 span more than a fourfold range on the same instrument — and it is stated as what it is, not as an industry consensus figure, because the four third-party numbers above do not converge on one.
How to read this log
Each future entry compares a release or a panel window against the one before it and reports what moved. Four numbers are the ones to watch:
- Head overlap. 237 of the top 250, 463 of the top 500.
- Segment leaders cited. 72 of 93.
- Community platforms. 10 of 27.
- Panel-recorded source depth. 4.1 distinct domains per answer on the complete list, against a six-item cap that only bit 15.6% of runs; watch for that share moving as the answer-evidence store's history lengthens past September 1.
A change in any of them is an entry. A release with none is recorded as no change.
What a publisher should take from entry one
Claude is the clearest case in this series of an engine whose overall reach understates its concentration. It cites fewer distinct domains than Perplexity, Gemini or Google AI Mode, yet reaches deeper into the top 100 and top 250 of the index than every engine but Perplexity — a narrower, more head-weighted citation set rather than a wide one. Where it falls down is segment leadership: missing the top-cited source in 21 of 93 categories means a brand that is not already the leading answer in its specific category cannot assume Claude's strong aggregate numbers will carry it there.
The measurement finding cuts the other way, and this time it is good news for AuthorityTech's own panel rather than a correction against it. A dashboard, a competitor tracker, or an internal panel that caps a citation array at six entries per answer will report close to the true count for Claude and Google AI Mode, understate ChatGPT modestly, and understate Perplexity and Gemini severely. Claude's answers exceeded the six-item cap on only 41 of 263 runs — the panel's Claude numbers needed the smallest correction of any engine measured, 0.5 percentage points on churn, once checked against the complete artifact.
Watch list for entry 2
- Whether Claude's head-overlap or segment-leader numbers move at the next Index release — the release updates daily, so entry 2 will report on genuine movement, not a fixed weekly cadence.
- Whether a floor-adjusted exclusive-domain count for Claude gets computed against a specific release, making that comparison publishable without the caveat in entry one.
- Whether the complete per-run artifact history extends further back than September 1, letting the panel restate churn over a longer window than 22 days.
- Whether any additional top-500 domain moves into or out of Claude's cited set, or whether the 21-of-93 segment-leader gap narrows.
Frequently asked questions
Is Claude's citation set simply smaller than the other engines', or is it different in kind? Both. Claude cites fewer distinct domains overall than Perplexity, Gemini or Google AI Mode, but a higher share of what it does cite sits inside the head of the index — the pattern of a concentrated set, not just a small one. It falls to the bottom half of the field only on segment leadership, where reaching a specific category's top-cited source matters more than overall reach.
Why did the six-source panel cap barely affect Claude when it distorted Gemini and Perplexity so badly? Because the cap only bites when an engine's true answer exceeds six sources, and Claude's complete-list mean is 4.8 URLs — below the cap most of the time. Claude's answers passed six sources on only 41 of 263 runs in the September 1–22 window, against 100% of Perplexity's and 90.7% of Gemini's. The cap is a property of the panel, not of the engine, and for a shallow-answering engine like Claude it simply has little to bite into.
Does Claude having three named, log-visible crawlers instead of Gemini's one make it easier to measure Claude's crawl behavior? It makes the robots.txt policy easier to read and control — a publisher can distinguish training, search indexing and live user fetches in server logs the way they cannot with Gemini's single combined token. It does not by itself change what this log can say about citation counts or churn, which come from the Index and the daily panel rather than from crawler logs; a domain-by-domain robots.txt check for Claude, in the style of this log's Perplexity entry, is future work, not something entry one attempts.