The AIs Don't Agree

Buyers now ask ChatGPT, Gemini, Claude and five more AI models what to buy and who to hire. Each answers with confidence. We asked all 8 the same buying questions across 185 categories and froze the record. They are not giving your buyers the same answer.

• 40% of categories: even the #1 brand is not named by all 8 AI models. 74 of 185 winners have at least one AI model that leaves them out entirely.
• 46/47 Australian category rankings have a different #1 than the global ranking. Ask from Sydney, get a different winner.
• 0.0 the score gap in the closest races: Amazon vs DHL Supply Chain, PwC vs KPMG, FreshBooks vs Xero, GitHub Copilot vs Cursor. Dead ties at the top.

Every number in this briefing comes from the July 2026 edition of the CiteHawk AI Index: an immutable public record of 185 categories and 1,791 ranked brands, with every verbatim AI answer stored and published. Recompute anything at citehawk.com/leaderboards/editions/2026-07.

How unanimous are the winners, really?

For each of the 185 categories, the #1 brand earned its position across all 8 AI models. But look at how many AI models actually name each winner:

• Named by 8 of 8: 111 (60%)
• Named by 7 of 8: 56 (30%)
• Named by 6 of 8: 16 (9%)
• Named by 5 of 8: 2 (1%)

Read that from the winner's side: 4 winners in 10 are invisible to at least one major AI model. And these are the category leaders. Below rank 1, agreement falls away quickly; the same brand routinely leads one AI model's answer and misses another's entirely.

The most contested crowns, July 2026

"A brand can lead the category and still be missing from a quarter of the answers your buyers actually hear."

The 16 categories where the #1 is named by only 6 of 8 AI models include digital marketing agencies and branding agencies. Service businesses are where the AI models disagree most.

Cross a border, get a different winner.

For 47 categories where local answers genuinely differ (agencies, banks, insurers, professional services), we also asked as a local buyer would. The winners barely survive the trip:

Outside the US, the global ranking almost never predicts the local answer. Visibility earned in one market silently fails to transfer, and a brand tracking only "what does ChatGPT say" from a US vantage point is reading someone else's scoreboard.

What this means if you own a brand

One AI model's opinion of you says almost nothing about the other seven, and one country's answer says little about the next. The only honest measurement is per-model and per-market, refreshed on a fixed rhythm, with the receipts kept. That is what the AI Index is: 8 AI models, monthly frozen editions, every verbatim answer published.

This briefing precedes The AI Recommendation Report, our full August study with month-over-month movement across all 8 AI models. Data: the CiteHawk AI Index July 2026 edition (immutable). Method: citehawk.com/leaderboards/methodology. Rankings are computed from AI responses only and are not for sale.