Citewake

Edition 2

Edition 2, collected 29–30 Oct 2026

What ChatGPT, Claude, Gemini and Perplexity recommended in seven categories, with every answer kept and the data open.

Published
Collected
29 October 2026, 13:38 – 30 October 2026, 04:18 UTC
Method
v0.9
Benchmark
v1

Reports

  1. Accounting software 1Xero69.6%2QuickBooks Online63.3%3Zoho Books32.2% 240 of 240 answers 300 brands named read the report
  2. Business bank accounts United States1Mercury65.5%2Relay39.5%3Chase39.3%Europe1Wise Business42.0%2Revolut Business41.3%3Qonto31.1% 240 of 240 answers 166 and 169 brands named, per panel read the report
  3. Cloud computing platforms 1AWS64.3%2Google Cloud62.3%3Azure36.0% 240 of 240 answers 464 brands named read the report
  4. Corporate cards United States1Ramp86.0%2Brex77.3%3Rho26.1%Europe1Pleo42.6%2Revolut Business37.7%3Spendesk27.4% 240 of 240 answers 132 and 200 brands named, per panel read the report
  5. Endpoint protection 1Microsoft Defender for Endpoint41.3%2CrowdStrike Falcon39.9%3SentinelOne Singularity25.3% 240 of 240 answers 281 brands named read the report
  6. International money transfer United States1Wise53.7%2Remitly44.4%3Western Union33.3%Europe1Wise69.4%2Remitly45.8%3Western Union36.5% 240 of 240 answers 87 and 97 brands named, per panel read the report
  7. Online payment processing 1Stripe75.2%2Adyen50.4%3PayPal24.3% 240 of 240 answers 246 brands named read the report

Recommendation rate: percent of decision answers in which the AI recommended the brand, all four AIs weighted equally. Categories asked per market are ranked per panel, never pooled.

What we saw

  • 50.0–90.0%of questions on which Gemini (gemini-3.1-flash-lite) ran a web search. ChatGPT, Claude and Perplexity searched on 100.0%.
  • 34–44.5pages ChatGPT found for a typical answer (median), of which it cited 2–3.
  • 93 of 306questions on which all four AIs recommended a brand first and named the same one (23, 5, 8, 12, 6, 2, 5, 5, 11, 16 by category and panel).

Ranges run across the seven categories. Each report has the full figures.

What this edition covers

  • Seven categories, 60 scored questions each, asked once of each of four AIs with their own web search on, in English and with no location sent.
  • Recommendation rate, top-pick rate, mention rate, own-site citation rate and negative mentions, overall and per AI, with 90% intervals.
  • Every brand named, in the data files.

What it doesn't

  • Change over time: this edition is not compared with the ones before it in these reports.
  • A language other than English; and, outside the categories asked per market, any particular country.

Data

Licence: CC BY 4.0. How the numbers are made: the method.

Version 1, published 30 October 2026. Published editions never change; corrections appear as new, dated versions.