Skip to content
AI Search Index

What AI assistants actually cite

Updated daily. The same prompts run against five assistants every day, so a change on one is visible against the other four.

289,482 citations
28,507 source domains
30,745 answers
Rolling 90 days

reddit.com's share of citations, by assistant

Share of all cited sources, 7-day rolling average

0.0%2.8%5.5%8.3%11%May 27Jul 11Aug 244.7%3.8%3.5%0.4%0.0%

Claude stops where we no longer see enough daily activity to report a share. The line ends rather than continuing at zero, which would claim a measurement we do not have.

google.com rose from 9.5% to 10.8% of Google AI Mode citations over the past week.

What kinds of sources get cited

ChatGPT

Other source
71.2%
Institutional
13.3%
Reference
10.3%
Community
2.8%
Editorial
2.5%

Perplexity

Other source
83.7%
Community
7.4%
Institutional
4.9%
Editorial
2.4%
Reference
1.7%

Gemini

Other source
75.9%
Community
17.0%
Institutional
4.1%
Editorial
2.5%
Reference
0.4%

Google AI Mode

Other source
82.0%
Community
13.8%
Editorial
1.9%
Reference
1.5%
Institutional
0.7%

Measured over the past 7 days. Categories cover the sources we can positively identify; most of the long tail is ordinary business websites.

Biggest moves this week

  • Source
    google.com
    Assistant
    Google AI Mode
    Change
    +1.3%
    Share now
    10.8%
    Share before
    9.5%
  • Source
    techradar.com
    Assistant
    ChatGPT
    Change
    -1.2%
    Share now
    0.4%
    Share before
    1.6%
  • Source
    reddit.com
    Assistant
    ChatGPT
    Change
    -0.8%
    Share now
    0.4%
    Share before
    1.2%
  • Source
    reddit.com
    Assistant
    Google AI Mode
    Change
    +0.7%
    Share now
    3.5%
    Share before
    2.7%
  • Source
    reddit.com
    Assistant
    Gemini
    Change
    +0.7%
    Share now
    5.3%
    Share before
    4.6%
  • Source
    youtube.com
    Assistant
    Gemini
    Change
    -0.5%
    Share now
    10.5%
    Share before
    11.0%

Each share is averaged across the websites we track, so a change in what is being measured cannot move a line. Days with too little measurement to be meaningful are left out rather than shown as zero.