US Coffee & Tea: AI Shelf Visibility Insights for Brands

Interactive Dashboard

US Coffee & Tea: AI Shelf Visibility Insights for Brands

A study of AI Recommendation Coverage (ARC) across ChatGPT, Claude and Gemini, measuring 494 coffee and tea brands against 100 shopper prompts

Request Demo

Filter the report

Every chart below follows this selection.

AI Recommendation Coverage Leaderboard

AI coffee and tea brand leaderboard

  • Nobody owns this shelf. Starbucks leads on 96 recommendations and 4.1% share of voice, and the fifteen brands on this chart account for 751 of 2,333 recommendations between them — under a third.
  • Craft roasters beat the big names to first place. Stumptown wins 21 first answers from 74 recommendations and Onyx Coffee Lab 16 from just 39, against 10 for Starbucks from 96.
  • Syrups punch above their weight. Torani and Monin take 58 and 43 recommendations from a segment that is only 13% of the dataset.
  • A long tail, not a shelf. 494 distinct brands and 1,771 distinct products appear across 2,333 recommendations, and no brand clears 5% share of voice.

*All interactive charts are best viewed on desktop or tablet.

Which brands hold up across all AI surfaces?

AI coffee and tea brand consistency by assistant

  • Only Starbucks and Peet’s hold across all three. Starbucks sits at 4.0%, 4.6% and 3.7%, Peet’s at 3.6%, 3.3% and 3.3% — the only two brands above 3% on every assistant, and Peet’s is the flattest row on the chart (1.1x spread).
  • Counter Culture is a ChatGPT brand. 4.1% there against 1.4% on Claude, a 2.9x gap and the widest on the list.
  • Lavazza is a Claude brand. 4.0% on Claude against 2.0% on Gemini, exactly double.
  • Stumptown, Onyx and Blue Bottle lean Gemini. All three peak there — 4.1%, 2.2% and 2.2% — with Onyx and Blue Bottle at roughly half that on Claude (1.1% and 1.0%).

*Each cell is the brand's share of that assistant's total recommendations. Only the top 15 brands are shown, so a column sums to roughly a third rather than 100%.

First place is its own contest

Being mentioned is not the same as being recommended first. Across 300 first-place answers the order changes completely: Stumptown takes 21, Lavazza 20 and Torani 18, while the most-recommended brand on the page manages 10.

*All interactive charts are best viewed on Desktop or Tablets

Recommendations, redrawn by parent company

AI coffee and tea recommendations by parent company

  • Four in ten recommendations have no named owner. Only 1,408 of 2,333 recommendations carry a manufacturer; the other 925 go to brands with no company recorded behind them.
  • The largest owner holds under 6%. JDE Peet’s at 130 recommendations, then Nestlé at 95 and Keurig Dr Pepper at 74 — no single company sets the agenda here.
  • The next tier is tightly bunched. Lavazza (71), J.M. Smucker (51) and R. Torre, the maker of Torani (46), sit within 30 recommendations of Keurig Dr Pepper.
  • Ownership is a long tail. 312 different manufacturers appear, and the twelve largest account for 639 of the 1,408 attributed recommendations — 45%.

*Covers the 1,408 of 2,333 recommendations that name a manufacturer. Recommendations with no manufacturer recorded are excluded from this view.

Sources the assistants get their evidence from

  • Each assistant shops somewhere different. ChatGPT cites target.com 165 times and never amazon.com, Claude cites amazon.com 115 and costco.com 25, and Gemini cites almost no retailer pages (2 amazon.com, 1 walmart.com).
  • Gemini’s evidence is unreachable. 515 of its citations route through grounding redirects, so its sourcing cannot be audited or influenced through the usual channels.
  • Roasters’ own sites are evidence here. counterculturecoffee.com (40), stumptowncoffee.com (30), onyxcoffeelab.com (27), torani.com (25) and monin.us (23) are all cited directly — in a category this fragmented, a brand’s own domain carries weight.
  • The evidence base is wide. 2,227 of 2,333 recommendations carry a citation, drawn from 440 different domains.

*Counts are citations rather than product recommendations. All interactive charts are best viewed on desktop or tablet.

What this means for your brand

This report is built from 100 real shopping prompts run across ChatGPT, Claude and Gemini. Every brand each assistant recommended was captured using DataWeave’s platform, along with the position it held in the answer and the sources cited. Recommendations are rank-weighted so first place counts for more than eighth, and brands are normalized and grouped under the manufacturer recorded for each product.

Run against your own portfolio, the same method answers three questions most brands cannot currently answer: where you rank in AI answers and against whom, which assistants you are losing on, and whether the fix sits in your retailer product pages or in third-party editorial.

This report analyzes the Coffee and Tea category. DataWeave measures the same thing across any category and market, continuously rather than as a snapshot. Talk to us to find out where your brands stand on the AI shelf.

Get in touch