Four products · sixteen agent decisions each

See how AI agents evaluate products people already know.

Each report follows the same buying situations across ChatGPT, Claude, Gemini, and Grok. Explore the headline result here, then open the complete research, evidence, and recommendations.

Independent public analysis · Education

Duolingo

September 3, 2026

01 / 04

Observed behavior · 16 completed paths

Every agent found Duolingo. They still disagreed on whether it was enough for serious learners.

Its brand was unmistakable. Evidence for advanced speaking and writing outcomes shaped the harder decisions.

Selected3/814 considered

ChatGPT3/4selected · 3/4 trial

2/2 found it unprompted

Claude2/4selected · 2/4 trial

2/2 found it unprompted

Gemini1/4selected · 1/4 trial

2/2 found it unprompted

Grok3/4selected · 4/4 trial

2/2 found it unprompted

Independent public analysis · Consumer subscription

Spotify

September 3, 2026

02 / 04

Observed behavior · 16 completed paths

Spotify entered every discovery path and was selected whenever it reached the final decision.

Agents still used different evidence to weigh discovery, audio quality, device support, bundles, and price.

Selected8/815 considered

ChatGPT4/4selected · 3/4 trial

2/2 found it unprompted

Claude4/4selected · 3/4 trial

2/2 found it unprompted

Gemini3/4selected · 1/4 trial

2/2 found it unprompted

Grok4/4selected · 3/4 trial

2/2 found it unprompted

Independent public analysis · Travel marketplace

Airbnb

September 3, 2026

03 / 04

Observed behavior · 16 completed paths

Airbnb was easy to discover. Total-price clarity weakened the recommendation when alternatives entered.

Agents repeatedly compared fees, cancellation terms, location, and hotel alternatives before choosing.

Selected1/813 considered

ChatGPT2/4selected · 0/4 trial

2/2 found it unprompted

Claude1/4selected · 0/4 trial

1/2 found it unprompted

Gemini3/4selected · 0/4 trial

1/2 found it unprompted

Grok2/4selected · 0/4 trial

2/2 found it unprompted

Independent public analysis · Consumer product

Dyson

September 3, 2026

04 / 04

Observed behavior · 16 completed paths

Agents knew Dyson. Its broad homepage did not help them choose the right product for a specific need.

The missing bridge was category and model guidance, especially when price and specialist competitors mattered.

Selected0/811 considered

ChatGPT0/4selected · 0/4 trial

2/2 found it unprompted

Claude0/4selected · 0/4 trial

1/2 found it unprompted

Gemini0/4selected · 0/4 trial

2/2 found it unprompted

Grok0/4selected · 0/4 trial

2/2 found it unprompted

Your product belongs in the next report

See what the agents would find, understand, and recommend.

Test your page free