Published October 14, 2026 | Version 1.0.0
Other Open

Mentioned Isn't Recommended: How ChatGPT and Gemini Recommend HR Software (CobaltKite Research 001)

Authors/Creators

  • 1. CobaltKite

Description

I asked ChatGPT (GPT-5.6 Sol) and Gemini (Gemini 3.1 Pro) which HR software a 50-person US startup should buy. Three prompts, ten runs each, on both assistants: 60 answers, collected in October 2026. For every vendor, I tracked whether it was named at all and whether it made the top three of the recommendation list.

Deel was named in 68% of answers (41 of 60) but made the top three in just 1 of 60. Rippling was ranked first in 52 of 60 answers (87%) and was in the top three of every one. Four vendors (Rippling, Gusto, BambooHR and Justworks) took 99% of all top-three slots. Workday was named in 13 of 60 answers and never recommended; every one of those mentions advised against it.

Being named by an AI assistant is not the same as being recommended. This dataset includes every answer word for word, the coding behind every number, and the script that reproduces the results. See README.md for the full method, definitions, and limitations.

Methods (English)

60 runs: 3 prompts x 2 assistants x 10 runs. Fresh temporary chat for every run, personalization off, no follow-up messages, prompts rotated. Web search left at each assistant's default. Runs made from outside the US without a VPN. Recommendation lists coded by hand; mentions detected automatically from the answer text. Rates reported with Wilson 95% intervals, differences with a stratified bootstrap (10,000 resamples), and headline findings checked by leaving out each prompt and each assistant.

Files

mentions.csv

Files (356.5 kB)

Name Size
md5:8e98e479560c56ed69a607e223f9c61d
17.8 kB Download
md5:b7c34fa51a487a21f41bdf3011dd2959
15.1 kB Preview Download
md5:073de4e8c7577d4b5037e2ca8f946d02
440 Bytes Preview Download
md5:341f651f12ea1140e033dd4e1dabc1f2
14.0 kB Preview Download
md5:9f18b9eeaa28953cebb0d3bc69e9e9dd
273.3 kB Preview Download
md5:5ca029fc73a6b39a7eabc308e3dd7956
35.8 kB Preview Download

Additional details

Dates

Collected
2026-10
Data collection