AI assistants cited General Mills’ websites in 0 of 94 answers. Publishers supplied the rest.
When a shopper asks an AI assistant which cereal, cake mix or dog food to buy, whose website does the answer lean on? We asked four assistants the buying questions for eight General Mills categories, counted the sites they cited, and set that beside search share and the company’s own filings. This is one half of machine availability, the step between being thought of and being bought: whose pages the assistant reads before it names anyone. On the questions buyers ask when choosing, General Mills supplied none of the sources.
0 of 94 answers. Not one cited cheerios.com, bettycrocker.com, oldelpaso.com, naturevalley.com, bluebuffalo.com, totinos.com, progresso.com or annies.com.
No brand’s own website, General Mills’ or a rival’s, appeared among the ten most-cited sources in any category. In three categories one publisher owned the answer: tastingtable.com in 8 of 12 taco-kit answers, health.yahoo.com in 8 of 12 soup answers, dogster.com in 6 of 11 dog-food answers. Meanwhile General Mills spent $873.6 million on marketing and advertising in its latest fiscal year.
The decision this is for
Brand plans fund two availabilities: mental (does the buyer think of you) and physical (can the buyer get you). Machine availability sits between them: does the assistant the buyer asks name you, and whose pages does it read to decide? General Mills is strong on the first two. This study measures the source side of the third.
Where a brand’s content and media money should go when buyers ask assistants instead of searching. The reader is a brand or portfolio lead at a packaged-food company.
If the brand’s own site carries the answer, the job is to keep it current. If third parties carry it, the brand’s spend reaches the buyer only through what those publishers write. Then machine availability is built through the publishers the assistants already trust, and it needs an owner and a line of its own.
Who the AI answers cited, category by category
| Category | Questions asked | Brand site cited | Category | Most-cited sources |
|---|---|---|---|---|
| Cereal Cheerios | What is the healthiest breakfast cereal to buy? / Which breakfast cereal is best for toddlers? | 0 of 12 | contested | health.yahoo.com (4), healthline.com (4) |
| Baking mixes Betty Crocker | What is the best boxed cake mix? / Which brownie mix tastes most like homemade? | 0 of 12 | contested | foodnetwork.com (4), allrecipes.com (4), americastestkitchen.com (4) |
| Mexican meal kits Old El Paso | What are the best taco kits at the grocery store? / Which taco shells are best for taco night? | 0 of 12 | owned | tastingtable.com (8), sporked.com (6) |
| Snack bars Nature Valley | What are the healthiest granola bars? / Which snack bars are best for hiking? | 0 of 11 | contested | consumerreports.org (3) |
| Pet food Blue Buffalo | What is the best dry dog food? / Which dog food brands do vets recommend? | 0 of 11 | owned | dogster.com (6), petmd.com (5) |
| Frozen snacks Totino's | What are the best frozen pizza snacks? / Which frozen snacks are best for a party? | 0 of 11 | contested | eatthis.com (5), chowhound.com (4) |
| Soup Progresso | What is the best canned soup? / Which canned soup is the healthiest? | 0 of 12 | owned | health.yahoo.com (8), delishknowledge.com (6) |
| Mac and cheese Annie's | What is the best boxed mac and cheese? / Which mac and cheese is best for kids? | 0 of 13 | contested | tasteofhome.com (4), allrecipes.com (4), nytimes.com (4) |
“Owned” means one site was cited in half the answers or more; “contested” means no site was. The top five sources held across the two askings in 4 or 5 of 5 places in every category, so these are not one-off answers. Retailers appeared in four top tens (amazon.com, target.com, chewy.com) and led none.
Search tells a different story
| Category | Brand | Rank | Share | Leader and notes |
|---|---|---|---|---|
| Cereal | Cheerios | 1 of 4 | 33.4% (-2.6 pts) | Cheerios leads |
| Baking mixes | Betty Crocker | 2 of 4 | 31.0% (-0.3 pts) | Ghirardelli 56.7% |
| Mexican meal kits | Old El Paso | 2 of 3 | 22.6% (+5.8 pts) | Ortega 60.9% · Ortega is also a surname; flagged before the run |
| Snack bars | Nature Valley | 3 of 3 | 21.4% (+2.6 pts) | Clif Bar 56.8% · KIND not measured (no clean 24 months) |
| Pet food | Blue Buffalo | 2 of 3 | 31.8% (+2.5 pts) | Purina 45.6% · Royal Canin not measured |
| Frozen snacks | Totino's | 3 of 3 | 22.4% (+2.7 pts) | Bagel Bites 48.7% · Hot Pockets not measured |
| Soup | Progresso | — | — | Refused: only one rival had 24 clean months |
| Mac and cheese | Annie's | 1 of 3 | 62.7% (-6.5 pts) | Annie's leads · Goodles not measured |
Search share is a measure of mental availability; it is not machine availability. Cheerios and Annie’s lead their sets on search, at 33.4% and 62.7%. Their sites were cited in 0 of 12 and 0 of 13 AI answers. Leading the search does not make the brand the source. General Mills ranked first in 2 of the 7 sets measured.
The company behind the brands
For the fiscal year ending May 31, 2026, General Mills reported net sales of $18.42 billion, down 5.4%, against a peer median of -3.5% (Conagra, Campbell's, Kraft Heinz (3 of 5 asked; Kellanova and Post not measured)), rank 4 of 4. Its gross margin was the highest of the group: 33.6% against a median of 28.1%. It spent $873.6 million on marketing and advertising, 45.2% more than in fiscal 2019. The US was 79.8% of sales.
Part of the sales decline is the sale of the US yogurt business, Yoplait among it, to Lactalis on June 30, 2025. General Mills does not report marketing spend by brand, so no category’s spend can be set against its answers.
What this cannot tell you
Sources, not mentions. The instrument counts the websites an answer cites, not the brands it names. An answer can recommend Cheerios without citing cheerios.com. This study measures the source side of machine availability; whether the assistant names the brand is the other side, and it was not measured here.
Buying questions, not recipe questions. Every question asked was a choice: best, healthiest, for toddlers, vets recommend. A brand’s own site is not an independent source on whether it is best, so assistants answer those from reviewers. “How do I make…” questions were not asked and may go to a brand’s recipe site.
Gemini’s sources are hidden. Gemini's citations return as vertexaisearch.cloud.google.com, a Google redirect host, in 3 or 4 answers per category; its real sources are not visible. Every other platform cited General Mills zero times. For a recipe site to reach a quarter of its category’s answers, Gemini would have had to cite it in every one of its own.
One day, two questions per category. Answers move. Search sets lost KIND, Royal Canin, Hot Pockets and Goodles to a data-quality control, and “Ortega” is also a surname. Soup’s search set could not be measured.
How it was built
Seven predictions were registered before any data was pulled, including the one we expected to win: that General Mills’ recipe sites would carry the recipe categories. It lost. The explanation we rated unlikely, that third parties carry every category, held in all eight. AI answers come from CULT+MATH’s answer-share run: two buyer questions per category, each asked twice of ChatGPT, Claude, Gemini and Perplexity with web search on. Search shares come from the share-of-search run. Company figures come from General Mills, Form 10-K for the fiscal year ending 2026-05-31, filed 2026-07-01 (SEC EDGAR, accession 0001628280-26-046466).
| Prediction | Registered | Result |
|---|---|---|
| D: General Mills' recipe sites (bettycrocker.com, oldelpaso.com) carry the recipe categories at 25%+ of answers | Roughly even (45-55%) | Missed: 0 of 12 in each |
| A: third parties carry 7 or more of 8 categories; no brand domain at 25%+ | Unlikely (20-45%) | Held: 8 of 8 |
| B: General Mills' sites at 25%+ in 4 or more categories | Very unlikely (5-20%) | Held as unlikely: 0 |
| C: a rival brand's site cited more than General Mills' where General Mills leads search | Very unlikely (5-20%) | Held as unlikely: no rival brand site in any top 10 |
| P-S1: General Mills first on search in 3 to 5 of 8 | Likely (55-80%) | Missed: 2 of 7 measured |
| P-S2: Nature Valley not first among Nature Valley, Clif Bar, KIND, RXBAR | Likely (55-80%) | Held: Clif Bar first |
| P-M1: General Mills revenue growth below the peer median | Likely (55-80%) | Held: -5.4% against -3.5%, 4 of 4 |
The same pattern held for Kraft Heinz in our meals study, and machine availability sets out why it matters. Any of these runs can be repeated by an agent at cultmath.com/ai.
Want this for your portfolio of brands?
We measured machine availability for one portfolio with public data. The same runs work for yours: your brands, your categories, the questions your buyers actually ask, and whose pages the assistants read instead of yours.
Email Ipalibo to request thisOr write to ipalibo@cultmath.com. $995, commissioned for your brands and categories. The questions are registered before any data is pulled; delivered as a dated PDF with every source. 14-day guarantee.
Where the money goes is one of the brand manager's decisions
This study reads where the answers come from for one portfolio. The CPG Brand Manager Decision Brief takes the decisions a brand lead owns, including where content and media money should go, and answers each from sourced evidence with the limits stated.
See the brand manager's decisions$495 standard, $995 custom. Pre-made and delivered on payment. If the decision turns out not to be decidable from evidence, it is refunded.