How often each brand appears in AI’s top-3 recommendations — plus who it merely mentions, and whose own site it actually cites. Measured across the real buyer questions people ask ChatGPT, Claude and Perplexity. Observed counts.
Asana
41%Top 3 recommendation rate
51%Mentioned by AI
18%Own site cited
359 / 882 eligible answers
02
monday.com
36%Top 3 recommendation rate
45%Mentioned by AI
18%Own site cited
322 / 891 eligible answers
03
ClickUp
31%Top 3 recommendation rate
42%Mentioned by AI
18%Own site cited
280 / 891 eligible answers
Based on 100 buyer questions · 3 AI engines · 3 repeats · 900 observed answers · September 2026
Read the full report →AI can recommend a brand without using its own website as a source. The distance between the two layers — mentioned by AI vs own site cited — is the owned-source opportunity.
How we measure it →Asana
monday.com
ClickUp
Trello
AI already knows these brands — the opportunity is making their own content part of the answer. See your brand's AI visibility →
September 2026 is Project management, measured on the 100-question methodology. Published categories stay in the archive and are re-measured on their next edition — August 2026 was the CRM pilot, kept as its rank-only record.
In the pipeline — categories queued for a future edition:
Each category is measured with a fixed set of 100 buyer questions — the same questions re-run every edition, so month-over-month movement reflects AI behavior, not question drift. Every question is asked three times to ChatGPT, Claude and Perplexity with live web search, producing 900 observed AI answers per edition.
Real decision-stage questions across 6 intent types.
ChatGPT, Claude, Perplexity — live web search enabled.
Each question runs 3 times per platform for stability.
100 × 3 platforms × 3 runs — every answer analyzed.
Mentions, top-3 recommendations and own-site citations, with raw counts.
Which project management tool is best for freelancers?
Category intentAsked 3× per engineBest project management tool for a nonprofit with limited funds?
Recommendation intentAsked 3× per engineBest software for an agency juggling multiple clients?
Recommendation intentAsked 3× per engineThree of the 100 frozen questions, verbatim — every question goes to all three platforms, 3 times each, with live web search on. Answer-level evidence — every raw AI answer behind the scores — publishes with the October edition.
From each answer we extract the ordered list of products it actually recommends and fold name variants to one canonical brand. A brand counts as Recommended in an answer when it appears among the top three extracted recommendations (top-3 inclusion); Mentioned when it is named anywhere in the answer; Cited when the engine's cited sources include the brand's own domain.
Anchored-query exclusion. Questions that explicitly name a brand (e.g. “What are the best alternatives to Asana?”) are excluded from that brand's denominator — the question creates an artificial opportunity for Asana to be named, while every other brand still has to earn the recommendation, so those answers stay eligible for everyone else. Brand-specific questions outside the comparison/alternative types are not allowed into the set at all.
Worked example. Asana appeared in the top three recommendations in 359 of its 882 eligible answers (900 observed − 18 answers to two Asana-anchored questions): 359 ÷ 882 = 40.7% → 41% Recommended.
Why 100 questions? The set is a designed benchmark, not an arbitrary list: each question is grounded in live search demand, phrased the way real buyers ask chat assistants (a mix of quick short asks and detailed context), and empirically verified to trigger product recommendations before being locked in. The September project-management set covers six buyer intents:
Per-brand raw counts (e.g. 359 / 882) are shown in each row above. Answer-level evidence with per-engine breakdowns is being added from the October edition onwards.
Free to cite and share under CC BY 4.0 — please link back to this page.
See how AI platforms recommend your brand across real buyer questions — per engine, with the sources they cited and what to fix.
Check my AI visibility →