Chapter 3. AI Product Visibility by Category: Electronics, Beauty, Health, Home, Sports, Toys, Auto, and Women's Dresses

AEOsome Research. Vijaya Kumar Channalli. Chapter 3 of the AI Shopping Visibility Study 2026, version 2026-10. The hub carries the abstract, the ten findings, the methods summary, and the other chapters. This chapter reports the category dimension of Study 2 and sets women's dresses (Study 1) beside it.

Abstract

Study 2 asked 5,584 shopping questions across seven categories, five subcategories each, on four AI engines. This chapter reports how the product-card rate, the shelf's brand concentration, the merchant network, the citation mix, and the price, rating, and review profile of surfaced products vary by category and subcategory, and which brands win each funnel stage inside each category. The category effect on the card rate is modest next to the engine and the stage: 22 points between Electronics (50.3%) and Auto (28.7%), against 59 points between engines and 40 between stages. But categories differ sharply in structure. Electronics is the only concentrated category (five brands hold 29% of its slots) and the only one in which Perplexity shows cards. Beauty and Health are the flattest shelves, with 300 to 470 brands per subcategory, and the most review-backed. Auto is the most community-cited category and Home the most editorial. Copilot beats Google AI Mode only in Health. Subcategory rates run from 54.4% (Headphones) to 21.0% (First Aid), a spread as wide as the category spread itself. Women's dresses, measured on ChatGPT alone three months earlier, are the most direct-to-consumer shelf in the study, with 42% of card slots on brand-owned stores against 2% to 7% in Study 2's categories, and the least review-backed.

1. Scope and Reading Guide

All Study 2 rates use the 22,335 de-duplicated answers; category cells hold 2,120 to 3,600 answers (530 to 900 per engine), subcategory cells 256 to 720. Study 1 (women's dresses, 8,520 ChatGPT answers, February 2026) appears in the category tables as a comparison row, never as an eighth Study 2 category: its engine, date, and collection run differ (hub, section 3.6). Five category × stage cells are absent by design: Home and Toys have no Budget Framing or Post Purchase questions, and Toys has no Validation questions. Those cells are marked absent, not zero, and the pooled category rates for Home and Toys are computed on seven stages rather than nine.

Definitions of product card, brand, source type, HHI, and confidence labels are in the hub and chapter 7.

2. Product Cards by Category and Engine

Figure 1. Product-card rate by category and engine, Study 2.

Table 1. Category card rates by engine, with ads and products per card-bearing answer.

Figure 12. For the same question, how many of the four engines show a card, by category.

The category effect is real and modest. Electronics leads at 50.3% pooled and Auto trails at 28.7%; every category's bootstrap interval is ±1.5 to ±2.1 points, so the order is Confirmed. Within that range, three structural facts matter more than the order.

Perplexity's shelf exists only in Electronics. It shows a card on 30.3% of electronics questions and on 0.3% or fewer in every other category (0.0% in Home). Its 1,842 card slots are 98% electronics. For six of the seven categories, Perplexity is a text engine.

Copilot is the Health engine. Health is the only category in which an engine other than Google AI Mode leads: Copilot cards 53.0% of health questions against Google AI Mode's 44.9%. Vitamins, cleaning products, paper goods, and first-aid items map onto Copilot's Amazon and Walmart index. Copilot is weakest in Auto (23.0%, below ChatGPT's 24.0%).

Google AI Mode is above 56% in every category and reaches 79.9% in Electronics. Its category spread (44.9% to 79.9%) is wider than ChatGPT's (22.4% to 40.8%) in absolute terms, but ChatGPT's figures are mixtures of its served models (chapter 2). Electronics questions get a card on three or more engines 39.2% of the time; Auto questions 5.7%.

3. Category by Stage

Figure 4 (parts A and B). Product-card rate by category, stages 1 to 5, four engines pooled. Electronics, Sports, Beauty, and Toys in part A; Home, Health, and Auto in part B.

Figure 5 (parts A and B). Product-card rate by category, stages 6 to 9.

Table 6. Product-card rate by category, stage, and engine, Study 2 (% of answers). n is per engine; absent cells are design gaps.

The easiest cells in the study are the four electronics shopping stages: Budget Framing 73.5%, Attribute Constrained 70.0%, Scenario Confirmation 69.2%, Purchase Execution 60.0% (400 answers each, four engines pooled). The hardest are Problem Recognition in Toys (7.8%), Post Purchase in Sports, Auto, and Beauty (8.0% to 8.3%), and Health's Scenario Confirmation and Validation (13.5%, 14.3%). The share of questions that get no card on any engine runs from 79% (Toys, Problem Recognition) to 3% (Electronics, Budget Framing).

Two category-specific patterns stand out. Health's scenario questions ("what protein treatment for 4C hair", "what to take for …") are answered as advice on every engine, 13.5% against 39.6% pooled across categories. Toys' problem questions ("can't keep kids entertained indoors") are the single hardest cell in the study, answered with activities rather than products. Chapter 1 reports every stage in full.

4. Who Fills the Cards

Table 2. The shelf by category: brands, concentration, merchants, and hosts.

Figure 6. Brand concentration by category (HHI × 100), with women's dresses for comparison.

Figure 11. Distinct brands per card-bearing answer, by category and engine.

Figure 13. Where the card slots sit: brand-owned store against marketplace, by category.

Electronics is the only concentrated category. Samsung, Apple, Sony, LG, and Google hold 29% of its 1,812 card-bearing answers' slots; its HHI is 0.044, twice the next category (Home and Toys, 0.021). The Smartphones subcategory is the most concentrated shelf in the study (HHI 0.185: Samsung 534 slots, Apple 275, Google 241), followed by TVs (0.164) and Laptops (0.131). Beauty (0.008) and Health (0.009) are the flattest: 300 to 470 distinct brands per subcategory, no brand above 20% of any subcategory, and a top brand (Maybelline in Makeup, CeraVe in Skincare, Lysol in Cleaning) that holds under 10% of slots.

Hosting follows the category's retail structure. Copilot's marketplace share (Amazon, eBay, Etsy, Newegg) is highest in Toys and Beauty; Walmart hosts a fifth or more of ChatGPT's and Google AI Mode's slots in Health and Home. Slots on brand-owned stores are rare in every Study 2 category (2.2% in Beauty to 6.6% in Sports) and common in women's dresses (42.3%): the dress shelf is Lulus, Petal & Pup, Quince, and Fashion Nova selling from their own stores, while electronics, health, and home are sold through Walmart, Target, Best Buy, and Amazon.

Table 8. Brands most often in the cards, by category and engine.

Table 9. Which brands win each funnel stage, by category (share of card-bearing answers at that stage that include the brand; four engines pooled).

The leading brand changes along the funnel inside a category, and the pattern repeats across categories: an incumbent leads exploration and attribute questions, a premium or trusted name leads comparison and validation, and store brands or consumables lead budget and post-purchase questions.

• Electronics: Samsung leads Category Exploration (38%), Attribute (33%), Budget (30%), and Scenario (29%); Apple leads Comparison (37%), Validation (22%), Purchase (20%, tied with Samsung), and Post Purchase (21%).
• Beauty: CeraVe and La Roche-Posay lead Problem, Category, Budget, and Scenario; Cetaphil leads Comparison (12%); Native leads Validation (9%); The Ordinary leads Attribute (11%); Olaplex leads Post Purchase (15%).
• Health: Kleenex and Puffs lead Problem; Levoit leads Category; Great Value, Equate, and up&up, three store brands, lead Budget (17%, 12%, 11%); Seventh Generation leads Validation (14%) and Scenario (20%); Lysol leads Purchase.
• Home: Breville leads Category Exploration (24%); Cuisinart leads Attribute (18%), Validation (21%), and Purchase (13%); Ninja leads Comparison (23%) and Scenario (16%); Shark leads Problem (18%).
• Sports: HOKA leads Problem (17%) and Scenario (13%); Nike leads Category (15%); adidas and Ozark Trail lead Budget; Rogue leads Comparison; Manduka leads Validation and Post Purchase; Specialized leads Purchase.
• Toys: Magna-Tiles leads Problem (29%); Fisher-Price leads Category (13%); LEGO leads Comparison (17%), Purchase (12%, tied with Hasbro), and Scenario (10%).
• Auto: Chemical Guys, Meguiar's, and WeatherTech rotate across stages; Michelin appears only at Category Exploration (11%); Meguiar's leads Purchase (16%).

Confidence: Confirmed for the category-level shelves (1,000 to 1,800 card-bearing answers per category); Suggestive for the stage-level winners, where most cells hold 50 to 250 card-bearing answers.

5. Who Is Cited, by Category

Table 3. Citation and answer-shape profile by category.

Figure 7. Cited source types by category, with women's dresses for comparison.

Figure 8. Share of answers citing reddit.com, by category.

Figure 14. Google AI Mode's ad rate by category.

Table 5. Category rankings on six axes (1 = highest).

Auto is the community category: 25.6% of its sources are Reddit, YouTube, forums, and social pages, the highest share, and it is the least editorial (31.2%). Home is the most editorial (49.3%; recipe, kitchen, and home-improvement publishers). Toys is the most store-cited (34.0% retail plus marketplace; Amazon and Walmart listings) yet among the least review-heavy. Beauty's cards are the most review-backed (84.4% with 100 or more reviews) and women's dresses the least (33.0%). Reddit appears in 27% to 41% of answers in every category, most in Electronics (40.8%) and least in Home (27.1%); in women's dresses, 16.4%.

Google AI Mode's ads cluster by category: Home 10.1% of its answers, Sports 5.9%, Health 4.2%, Electronics 3.1%, Auto 2.9%, Toys 2.1%, Beauty 0.3%. Chapter 5 reports where ads sit relative to organic cards.

6. Price, Rating, and Review Profile

Table 4. What gets surfaced: price, rating, and review profile by category.

Figure 9. Median price of surfaced products by category.

Figure 10. Review profile of surfaced products by category: share with a rating, share with 100 or more reviews, share rated below 4.0.

Median card prices follow the category: Electronics $272 (TVs $839, Laptops $699, Smartphones $381, Headphones $105, Smart Home $80), Home $120, Sports $85, Auto $65, Toys $34, Beauty $23, Health $19, women's dresses $70. Within every category, only 2.6% (Beauty) to 7.6% (Home) of surfaced products are rated below 4.0, and mean ratings sit between 4.43 (Home) and 4.62 (Toys). The shelf is pre-filtered to four stars everywhere; what differs by category is review volume, which follows the engine mix and the merchant network: Beauty and Health are Copilot-heavy and Amazon-hosted (median 3,238 and 1,329 reviews), Electronics and Auto are Google-heavy (147 and 180), and dresses are long-tail SKUs (31).

7. The 35 Subcategories

Figure 2. Subcategories with the highest product-card rates.

Figure 3. Subcategories with the lowest product-card rates.

Table 7. All 35 subcategories: card rate, ads, shelf size, concentration, price, rating, review median, and top three brands.

The subcategory spread (54.4% to 21.0%) is as wide as the category spread. The top of the list is all electronics (Headphones 54.4%, Smartphones 52.1%, TVs 51.2%, Laptops 49.9%, Smart Home 44.2%), then Baby & Toddler Toys (40.0%), Makeup (39.4%), Running Shoes (39.3%), Cycling (37.9%), Yoga (37.8%). The bottom is First Aid & Medicine (21.0%), Board Games & Puzzles (21.2%), Vitamins & Supplements (23.1%), Tires & Wheels (24.7%), and Auto Lighting (25.1%). Within a category the range can be 15 points (Health: Cleaning Supplies 36.9% to First Aid 21.0%; Toys: Baby 40.0% to Board Games 21.2%).

Concentration is a subcategory property as much as a category one: Smartphones 0.185, TVs 0.164, Laptops 0.131, and LEGO & Building Sets 0.239 (LEGO holds 151 of its slots) against 0.02 to 0.05 for Beauty, Health, and most Sports and Auto subcategories.

8. Engine Profiles Inside Each Category

The seven tables below give each category's measures by engine: card rate, shelf, hosts, citations, answer shape, and price and rating profile.

Table 1.1. Electronics: measures by engine. Answers: All engines 3,599, Google AI Mode 900, Copilot 899, ChatGPT 900, Perplexity 900.

Table 2.1. Sports: measures by engine. Answers: All engines 3,600, Google AI Mode 900, Copilot 900, ChatGPT 900, Perplexity 900.

Table 3.1. Beauty: measures by engine. Answers: All engines 3,600, Google AI Mode 900, Copilot 900, ChatGPT 900, Perplexity 900.

Table 4.1. Toys: measures by engine. Answers: All engines 2,120, Google AI Mode 530, Copilot 530, ChatGPT 530, Perplexity 530.

Table 5.1. Home: measures by engine. Answers: All engines 2,216, Google AI Mode 554, Copilot 554, ChatGPT 554, Perplexity 554.

Table 6.1. Health: measures by engine. Answers: All engines 3,600, Google AI Mode 900, Copilot 900, ChatGPT 900, Perplexity 900.

Table 7.1. Auto: measures by engine. Answers: All engines 3,600, Google AI Mode 900, Copilot 900, ChatGPT 900, Perplexity 900.

9. Women's Dresses Against the Seven

Study 1 measured one category, women's dresses, on ChatGPT in February 2026. Compared on the same definitions, it differs from Study 2's seven categories in four ways that are category properties rather than engine or time properties.

It is the most direct-to-consumer shelf in the study: 42.3% of card slots sit on brand-owned stores (Lulus, Petal & Pup, Quince, Fashion Nova), against 2.2% to 6.6% in Study 2's categories. It is the most brand-cited: 10.3% of its sources are brand-owned domains, against 3.4% to 8.4%. It is the least review-backed: median 31 reviews per card and 33.0% with 100 or more, against medians of 147 to 3,238. And its brand concentration (HHI 0.017) sits mid-table, between Toys and Auto, with Lulus in 30% of carousels.

What cannot be read from the comparison is a trend. Study 1's 59.5% card rate against Electronics' 50.3% or Beauty's 35.9% mixes a different engine mix, a different month, and different served models; chapter 2 shows that ChatGPT's full models in May 2026 card 66.9% of non-fashion questions. The category facts above hold regardless; the rate comparison does not.

10. Discussion

Category sets the floor and the structure, not the verdict. The 22-point category spread is within the range a brand can move by choosing the engine it measures on and the stages it writes for. Three structural facts follow a brand into its category.

First, who hosts the shelf. In Study 2's categories the cards link to Walmart, Amazon, Target, and Best Buy; a brand reaches the shelf through those retailers' catalogs, and its own store is a marginal host. In women's dresses, the brand's own store is the main host. A direct-to-consumer brand in electronics or health should expect the retailer route; a direct-to-consumer brand in apparel can expect its own store to be indexed.

Second, how concentrated the competition is. In Electronics and in LEGO, a handful of incumbents hold the slots and the winner changes by stage; in Beauty and Health, hundreds of brands share a flat shelf and no brand holds 10%. Flat shelves reward presence at the right stage; concentrated shelves reward displacing a named incumbent at a specific stage.

Third, what the engines read. Auto is answered from forums and YouTube, Home from publishers, Toys from store listings, Beauty from heavily reviewed Amazon listings. The citation layer a brand needs differs by category even when the card layer does not.

11. Limitations Specific to This Chapter

1. Five category × stage cells are absent by design (Home and Toys: Budget Framing and Post Purchase; Toys: Validation). Home's and Toys' pooled rates are computed on seven stages and are not strictly comparable with the other five categories' nine-stage rates; on a common seven-stage basis they would be slightly lower relative to Electronics.
2. Subcategory cells hold 256 to 720 answers across four engines, 64 to 180 per engine; subcategory-by-engine figures are Suggestive and are not reported.
3. Stage winners inside a category are computed on 50 to 250 card-bearing answers per cell and are Suggestive.
4. Perplexity's price and rating figures are excluded from category comparisons except in Electronics, the only category where it shows cards, and even there its prices include non-US stores.
5. Women's dresses is one category on one engine at one date; category conclusions drawn from it are about dresses on ChatGPT in February 2026, not about apparel in general.

Navigation. Previous: Chapter 2. Four engines, one question · Next: Chapter 4. What predicts a product card · AI Shopping Visibility Study 2026 (hub, abstract, and all chapters).

About the author. Vijaya Kumar Channalli is a serial SaaS founder and the founder of AEOsome (Tartu, Estonia), which helps eCommerce brands get their products recommended when shoppers ask AI what to buy. Read how we write, source and correct posts in our editorial policy.

Sources: AEOsome Research, AI Shopping Visibility Study 2026, chapter 3. Study 2: 22,335 de-duplicated answers across ChatGPT, Google AI Mode, Microsoft Copilot, and Perplexity, seven categories and 35 subcategories, United States, May 5 to May 14, 2026. Study 1: 8,520 ChatGPT answers, women's dresses, United States, February 28 to March 1, 2026. Hub: https://www.aeosome.com/blog/research/ai-shopping-visibility-study-2026/