How Claude Decides Which B2B Brand to Recommend: A 1,000-Query Study

How Claude Decides Which B2B Brand to Recommend: A 1,000-Query Study

SEO & AEO
Slate
Research study
8,989 brand mentions across 253 brands, showing that winning an awareness query predicts almost nothing about winning a comparison query.
Organisation
Slate
Written By
Akilan Arumugam
PUBLISHED
August 5, 2026
Read
3
min

What it is.

Slate ran 10 prompts 100 times each in independent Claude sessions with no memory carried between them, across CRM, payroll, email, eSignature and project management.

That produced 8,989 brand mentions across 253 brands. The correlation between a brand's awareness rank and its comparison rank, for the 61 brands appearing in both, is 0.25, which is close to random.

For anyone working on AI visibility the blunt finding is the citation column: across all 1,000 runs, Claude returned zero citations.

What you get.

  • The method: 10 prompts run 100 times each in independent sessions, across five verticals
  • 8,989 brand mentions across 253 unique brands, parsed by hand
  • Awareness-to-comparison rank correlation of 0.25 across the 61 brands in both
  • Stability by category, where one payroll prompt returned the same ten brands in the same order in 63 of its 100 responses
  • A confidence spread of 0.11 to 0.63 between categories inside the same model on the same day
  • Sentiment movement, with 6.1% of awareness mentions cautious and 0% in comparison mode

Read it at the source.

Opens on Slate. Free, and the figures on this page all come from it.
Read the guide

Frequently Asked Questions

Do repeat runs mean anything?
Who ran it?
Does Claude cite its sources when recommending brands?
Is it worth working on anything below the top two positions?
Does winning best-in-category queries help with alternatives queries?
How stable are AI brand recommendations?
Strengths
  • Independent sessions with no memory, so repeat runs measure the model and not the conversation.
  • Reports the correlation itself, so you can judge how weak the link is.
  • Names the zero-citation result plainly, even though it complicates the AEO story.
Limitations
  • Slate's own research, and Slate sells AI visibility software.
  • One model on one set of days, so treat the ranks as a reading and not a leaderboard.
  • Five verticals and 10 prompts, so category-level conclusions rest on 100 runs each.
Skip this if
  • Skip it if you already track whether assistants name you on awareness questions.
Ideal for
  • AEO and GEO leads
  • Product marketers on competitive positioning

The plays behind the reading.

We build inbound GTM engines for B2B software teams, and these are the frameworks we build from. Bring us a pipeline target and we'll map the route.