AI Source Index · one engine · 2026-W41
The most-cited websites in Claude answers — updated weekly
In week 2026-W41, measured 5 October 2026, the websites Claude cited most were en.wikipedia.org (6 questions, 15%), forbes.com (6 questions, 15%) and techradar.com (5 questions, 13%). Each count is the number of the 40 questions it answered whose answer cited that site at least once.
Week 2026-W41, measured 5 October 2026. One sample per engine per question; a difference of one or two questions is noise, not a trend.
What stands out
We ask Claude through Anthropic’s Messages API with its web search tool, up to three searches per answer. Like every API answer, it carries no account memory or custom instructions.
This week its 40 answers cited 288 different websites, 8.1 per answer on average (across the 6 engines: 3.4 to 10.7).
7 of its 40 answers cited no website at all.
Among the source types we classify, media sites were its largest group, at 7% of its citations (a domain counts once per answer); 89% matched no type — retailers, vendors’ own sites and the rest.
The 25 most-cited websites
| Domain | Type | Questions | Share (95% interval) | vs week before |
|---|---|---|---|---|
| en.wikipedia.org | reference | 6 / 40 | 15% (7–29%) | +5 (from 1) |
| forbes.com | media | 6 / 40 | 15% (7–29%) | +1 (from 5) |
| techradar.com | media | 5 / 40 | 13% (5–26%) | ±0 |
| tomsguide.com | media | 4 / 40 | 10% (4–23%) | ±0 |
| nerdwallet.com | media | 3 / 40 | 8% (3–20%) | −2 (from 5) |
| aol.com | other | 2 / 40 | 5% (1–17%) | +1 (from 1) |
| clearvaluelending.com | other | 2 / 40 | 5% (1–17%) | new |
| cnbc.com | media | 2 / 40 | 5% (1–17%) | ±0 |
| emailtooltester.com | other | 2 / 40 | 5% (1–17%) | +1 (from 1) |
| emailvendorselection.com | other | 2 / 40 | 5% (1–17%) | −1 (from 3) |
| gulfnews.com | other | 2 / 40 | 5% (1–17%) | +1 (from 1) |
| health.yahoo.com | other | 2 / 40 | 5% (1–17%) | new |
| money.usnews.com | other | 2 / 40 | 5% (1–17%) | ±0 |
| northflank.com | other | 2 / 40 | 5% (1–17%) | −2 (from 4) |
| popsci.com | other | 2 / 40 | 5% (1–17%) | new |
| techsifted.com | other | 2 / 40 | 5% (1–17%) | +1 (from 1) |
| thehappylovedlife.com | other | 2 / 40 | 5% (1–17%) | new |
| ventureharbour.com | other | 2 / 40 | 5% (1–17%) | +1 (from 1) |
| wallethub.com | other | 2 / 40 | 5% (1–17%) | ±0 |
| wise.com | other | 2 / 40 | 5% (1–17%) | +1 (from 1) |
| 13labs.au | other | 1 / 40 | 3% (0–13%) | new |
| 1vps.com | other | 1 / 40 | 3% (0–13%) | ±0 |
| aeoeye.com | other | 1 / 40 | 3% (0–13%) | new |
| agiled.app | other | 1 / 40 | 3% (0–13%) | ±0 |
| amazon.com | other | 1 / 40 | 3% (0–13%) | new |
Share = questions whose answer cited the site at least once, out of the 40 this engine answered. The interval is a 95% Wilson score interval; “thin sample” marks an interval wider than a third of the scale, where the rate cannot be read off the sample.
The change column compares with 2026-W40, a week in which this engine answered the same number of questions under the same panel. A move of one or two questions is noise, not a trend.
Method
A fixed, published panel of 40 buyer questions in English is put once to each of 6 answer engines every Monday. Every URL an answer cites is stored, and a domain counts once per question per engine. One sample per engine per question.
How we measureEvery observation behind 2026-W41, as JSON (this engine’s rows carry "engine": "claude")
The same week, other engines
Most-cited sites in PerplexityMost-cited sites in ChatGPTMost-cited sites in GeminiMost-cited sites in Google AI OverviewsMost-cited sites in Google AI Mode