Most-cited websites by Claude
The publishers and platforms Claude actually reaches for when it answers, ranked by citation share, with weekly movement.
Rolling 4 weeks: 43,117 citations · 8,653 model responses · every named domain seen across ≥ 2 independent projects
Top-50 stability: 18% retained week to week: a fast-changing source list
This week's movers
Citation-share change, week commencing 2026-09-28 vs 2026-10-05. Minimum 25 citations in both weeks and a 0.2pp move to qualify; domains newly added to monitoring don't count as movers.
No qualifying decliners this week.
Claude's top sources
Top 50 domains by share of all citations in the rolling window. The trend column is the weekly share trajectory.
| # | Domain | Share | Citations | Trend |
|---|---|---|---|---|
| 1 | wikipedia.org | 0.7% | 316 | |
| 2 | trustpilot.com | 0.4% | 175 | |
| 3 | yahoo.com | 0.4% | 160 | |
| 4 | autohero.com | 0.4% | 152 | |
| 5 | ayvens.com | 0.3% | 123 | |
| 6 | lizy.be | 0.2% | 104 | |
| 7 | linkedin.com | 0.2% | 97 | |
| 8 | tesla.com | 0.2% | 96 | |
| 9 | bmwgroup.com | 0.2% | 88 | |
| 10 | finn.com | 0.2% | 78 | |
| 11 | forbes.com | 0.2% | 77 | |
| 12 | hyundai.com | 0.2% | 75 | |
| 13 | amazon.com | 0.2% | 75 | |
| 14 | facebook.com | 0.2% | 75 | |
| 15 | volvocars.com | 0.2% | 73 | |
| 16 | kia.com | 0.2% | 70 | |
| 17 | automobile-propre.com | 0.2% | 66 | |
| 18 | drive.com.au | 0.2% | 65 | |
| 19 | highmotor.com | 0.1% | 61 | |
| 20 | usnews.com | 0.1% | 61 | |
| 21 | vivacar.fr | 0.1% | 61 | |
| 22 | autoexpress.co.uk | 0.1% | 60 | |
| 23 | gulfnews.com | 0.1% | 59 | |
| 24 | medium.com | 0.1% | 59 | |
| 25 | carwow.de | 0.1% | 57 | |
| 26 | cars24.com.au | 0.1% | 57 | |
| 27 | carwow.co.uk | 0.1% | 56 | |
| 28 | nih.gov | 0.1% | 55 | |
| 29 | carexpert.com.au | 0.1% | 55 | |
| 30 | autoscout24.it | 0.1% | 54 | |
| 31 | play.google.com | 0.1% | 54 | |
| 32 | whichcar.com.au | 0.1% | 53 | |
| 33 | mercedes-benz.com | 0.1% | 52 | |
| 34 | guideflow.com | 0.1% | 52 | |
| 35 | autofact.cl | 0.1% | 51 | |
| 36 | autorai.nl | 0.1% | 51 | |
| 37 | racv.com.au | 0.1% | 51 | |
| 38 | walmart.com | 0.1% | 50 | |
| 39 | improvado.io | 0.1% | 49 | |
| 40 | carsguide.com.au | 0.1% | 48 | |
| 41 | edu.pe | 0.1% | 48 | |
| 42 | glassdoor.com | 0.1% | 48 | |
| 43 | infobae.com | 0.1% | 46 | |
| 44 | drivingeco.com | 0.1% | 45 | |
| 45 | autoscout24.nl | 0.1% | 45 | |
| 46 | thenationalnews.com | 0.1% | 45 | |
| 47 | uspto.gov | 0.1% | 42 | |
| 48 | indeed.com | 0.1% | 42 | |
| 49 | arxiv.org | 0.1% | 42 | |
| 50 | wecovr.com | 0.1% | 41 |
The race for the top
Where the leading sources sat against each other, week by week. Rank space makes every crossing an event.
Data table
| Series | 21 Sept | 28 Sept | 5 Oct | 12 Oct |
|---|---|---|---|---|
| wikipedia.org | 1 | 1 | 1 | 1 |
| trustpilot.com | 3 | 3 | 3 | 2 |
| yahoo.com | 2 | 2 | 4 | 5 |
| autohero.com | 4 | 5 | 2 | 3 |
| ayvens.com | 5 | 4 | 5 | 4 |
| lizy.be | 6 | 6 | 6 | 6 |
3 places lost: yahoo.com, #2 → #5 (12 Oct)
How sourced are the answers?
Citations per response over the last 12 complete weeks: whether the models are leaning on the web more or less.
Data table
| Series | 20 Jul | 27 Jul | 3 Aug | 10 Aug | 17 Aug | 24 Aug | 31 Aug | 7 Sept | 14 Sept | 21 Sept | 28 Sept | 5 Oct |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Claude | 9.6 | 12 | 9.9 | 8.7 | 4.1 | 7.4 | 4.6 | 4.2 | 4.6 | 5.1 | 4.9 | 4.9 |
Methodology
- What's counted: every citation/source URL Claude returned across the rolling 4 ISO weeks, aggregated from every analysis run on BrandInsightAI. Citations deduplicate to one registrable domain per model response.
- Quality gate: a domain must appear across ≥ 2 distinct projects to be named; shares are computed against ALL citations in scope, so the gate hides names, never distorts percentages.
- Movers: week-over-week comparisons use the two most recent COMPLETE ISO weeks (2026-09-28 → 2026-10-05); the in-progress week is excluded because it refreshes hourly, and a domain must clear the citation floor in BOTH weeks, so a newly monitored category entering the dataset never shows up as a gainer.
- Freshness: the snapshot behind this page recomputes every hour. All figures are measured, never estimated.
Where next
This board answers "which websites does AI cite?". These answer the questions that come next — which brands it recommends, which sources each model prefers, and how any of it is counted.