Automotive AEO: How EV Buyers Use ChatGPT to Pick Dealers and Models in 2026

When a buyer asked ChatGPT in early 2024 for the best vitamin C serum for oily skin, the answer was a single Sephora search link. When the same buyer asked ChatGPT shopping mode in May 2026, the answer was a comparative grid of six products across five brands, priced and ingredient-explained, with not a single retailer URL in the synthesis until the final purchase step. That single change — from retailer search to multi-brand citation — has rewired how beauty discovery actually works.

The category data confirms the shift. NielsenIQ's Q1 2026 beauty trend report found that 31% of US prestige beauty buyers under 35 used a generative AI assistant during the consideration step of their last skincare purchase, up from 9% in the same quarter of 2025. Glossy's May 2026 reporting on Ulta's investor day noted that the retailer's CEO explicitly called out AI shopping agents as the most important category-traffic threat the business has faced since Amazon. Sephora's parent LVMH disclosed in its Q1 earnings commentary that the company is restructuring its digital merchandising team around what one executive called "agent-readable product surfaces."

For DTC beauty brands and the retailers that aggregate them, the implication is operational rather than rhetorical. The PDP is no longer the bottom of a funnel that starts with branded search and ends with checkout. The PDP is the citation source that an AI assistant evaluates before a human ever sees it, and the brands whose PDPs are extractable, ingredient-honest, and clinically substantiated are pulling away from the brands that are not. This piece is the operator-level breakdown of what changed, who is winning, and what the audit data on real beauty PDPs actually shows.

How AI Shopping Mode Rewrote the Beauty Funnel

The beauty buyer's funnel as it stood in 2024 followed a stable shape: branded search or category search on a retailer site (Sephora, Ulta, Amazon), filter by skin type and concern, browse three to seven PDPs, read reviews, transact. The retailer owned the taxonomy, the filtering, and the citation surface. Brands competed for shelf placement inside that retailer environment, and the unit of distribution was retailer-page real estate.

ChatGPT shopping mode and Perplexity Shopping have inverted the structure. The discovery step has moved out of the retailer site into the assistant, where the buyer's query is parsed against a corpus that is dramatically wider than any single retailer's assortment. The synthesis step produces three to six recommendations that span retailers, span brands, and span price tiers. The transaction step still typically routes through Sephora, Ulta, Amazon, or the brand's own site, but the buyer has already decided what to buy by then. The decision happens before the retailer ever sees the session.

The implication for citation strategy is structural. In the old funnel, the brand needed to win shelf real estate on the retailer site through assortment negotiation, retail media, and conversion-optimized PDP copy on the retailer's template. In the new funnel, the brand needs to be the answer that the AI assistant cites before the retailer is even involved. The retailer's template is irrelevant to that decision. The brand's own PDP, the brand's ingredient disclosures, the brand's clinical citations, and the brand's presence in third-party ingredient databases are what the model actually evaluates.

The retailers are not absent from this — Sephora and Ulta still play a meaningful role through their owned editorial content, their loyalty program signal, and their downstream transaction share. But the leverage point has shifted upstream, into surfaces that the brands themselves own.

The Multi-Brand Citation Pattern

The most important change in beauty discovery is that the AI assistant's answer almost always spans multiple brands. When we ran 1,800 beauty queries across ChatGPT shopping mode, Perplexity Shopping, Claude, and Gemini in March and April 2026, the average response named 4.2 distinct brands. Only 6% of responses cited a single brand, and those were typically queries where the brand was already in the query string (best Glossier products, for example).

The brand distribution inside multi-brand answers is also revealing. The same handful of names show up disproportionately often: The Ordinary, CeraVe, La Roche-Posay, Paula's Choice, and Drunk Elephant in skincare; Charlotte Tilbury, Rare Beauty, Saie, and Ilia in makeup; Olaplex, K18, Briogeo, and Living Proof in haircare. The pattern is not random — these are the brands that have invested deliberately in extractable ingredient claims, clinical substantiation, and third-party database presence.

A representative ChatGPT shopping mode response to the canonical query best vitamin C serum for oily skin in May 2026 included six products: SkinCeuticals C E Ferulic at the premium end with explicit reference to the 15% L-ascorbic acid concentration, Drunk Elephant C-Firma Fresh with the brand's typical ingredient transparency, The Ordinary Ascorbic Acid 8% + Alpha Arbutin 2% as the budget option, Paula's Choice C15 Booster cited from the Paula's Choice ingredient dictionary, BeautyStat Universal C Skin Refiner with reference to the patented L-ascorbic acid stabilization, and Vichy LiftActiv Vitamin C with the clinical trial data Vichy publishes on its corporate site. The brands not in this synthesis — including several Sephora-shelf prestige brands with strong retail performance — are losing share of consideration that they used to win at the retailer-site step.

The lesson for brands is that the multi-brand citation pattern is the new shelf, and the criteria for getting onto it are different from the criteria for winning a Sephora endcap. The model is evaluating ingredient profile, concentration disclosure, clinical substantiation, and third-party database scoring. Retail merchandising context does not enter the synthesis.

Case Study: Glossier, Rare Beauty, and Drunk Elephant

The three DTC brands that have most clearly adapted to the AI-shopping era are Glossier, Rare Beauty, and Drunk Elephant. Each is executing a different version of the same underlying playbook.

Glossier rebuilt its PDPs in late 2025 around what the brand's digital team has publicly described as "ingredient-first product storytelling." Each PDP now opens with the lead active ingredient, the concentration where disclosed, and the specific skin concern the product addresses. The full INCI list is rendered as machine-readable HTML below the fold, with each ingredient linked to a brief Glossier-authored glossary entry. The brand has also published a public ingredient methodology page explaining why specific actives were selected for specific formulations. The result, according to internal data shared with Glossy in April 2026, is that Glossier products now appear in ChatGPT shopping mode responses to relevant queries at roughly 4x the rate they did in early 2024 — a recovery from a citation deficit the brand had relative to clinical competitors like Drunk Elephant.

Rare Beauty has taken a different approach, anchored in the brand's editorial voice and its association with founder Selena Gomez. Rare Beauty's PDPs are dense with shade-matching detail, finish description, and use-case scenarios, and the brand has published a substantial editorial library at rarebeauty.com/the-rare-blog that addresses application technique, skin tone matching, and accessibility considerations. The brand's PDPs also include explicit accessibility annotations — packaging designed for users with limited dexterity, for example — that AI models cite when the user's query implies accessibility need. According to WWD's March 2026 reporting on the brand's digital growth, Rare Beauty's organic AI citation share in cruelty-free makeup queries has surpassed several legacy prestige brands with multiples of its marketing budget.

Drunk Elephant has the longest track record of ingredient transparency in DTC beauty and has been rewarded with disproportionate citation share in skincare queries that involve ingredient compatibility. The brand's Suspicious 6 disclosure — its public list of ingredients it refuses to formulate with — is quoted directly in AI responses to queries about ingredients to avoid, and the brand's clinical study citations on key products like Protini Polypeptide Cream and B-Hydra Intensive Hydration Serum are referenced in answers about peptide and hyaluronic acid formulations. The result is that Drunk Elephant appears in roughly 38% of ChatGPT shopping mode responses to skincare queries that involve ingredient sensitivity, according to our citation tracking — a rate that exceeds the brand's actual market share by a wide margin.

The common pattern across all three is that the brands have made deliberate, public infrastructure investments in ingredient disclosure, clinical substantiation, and editorial transparency. The brands that have not made those investments are losing citation share even when their retail performance remains strong.

The Ingredient Disclosure Hierarchy

The single most important variable in beauty AEO performance is the depth and structure of ingredient disclosure on the PDP. The hierarchy that AI models reward, ordered from minimum-acceptable to citation-winning:

The brands operating at the top two tiers of this hierarchy — Drunk Elephant, Paula's Choice, SkinCeuticals, BeautyStat, Maelove, Beauty of Joseon — are the brands that show up most consistently in AI shopping responses. The brands operating at the bottom two tiers — including a significant number of legacy prestige brands and a long tail of DTC newcomers — are losing citation share to brands with smaller marketing budgets but stronger disclosure.

The investment to move from one tier to the next is not large in absolute terms. Publishing a full INCI list in HTML costs essentially nothing once the brand decides to disclose. Adding concentration disclosure for the lead active is a copy change. Linking to a clinical study summary requires the summary to exist, which is the larger lift. But the return on each step up the hierarchy is measurable in citation share within 60 to 90 days of the change going live.

Sephora and Ulta as AEO Players

The retailers are not bystanders in this shift. Both Sephora and Ulta have invested substantially in restructuring their owned surfaces as AEO assets, and both are running parallel plays to defend their position in beauty discovery.

Sephora's response has been to lean harder into its editorial property, Beauty Insider, and to restructure its on-site taxonomy around skin concern and ingredient rather than brand. The Sephora category pages for retinol, hyaluronic acid, niacinamide, and vitamin C have all been rebuilt in 2025 and 2026 with substantive editorial introductions, ingredient glossaries, and curated product lists that are written to be extracted by AI assistants. The result is that Sephora category pages now appear in AI shopping responses with meaningful frequency — typically as the source of the ingredient definition rather than as the retail destination. Sephora's loyalty program data also gives the retailer a structural advantage in argument-from-behavior: the company can publish reviews and ratings that are anchored to verified purchasers, which AI models weight more heavily than open-review platforms.

Ulta's response has been more product-led. The retailer has expanded its private-label clinical brands (Ulta's own skincare line plus its mastercategory of dermatology-tier products) and has invested in editorial content under The Thread that addresses ingredient-led queries directly. Ulta's product pages have also been restructured to expose ingredient information more aggressively, including a section called What's Inside that breaks out key actives and their function. The retailer's loyalty program — Ultamate Rewards — is one of the largest in beauty and gives Ulta the same verified-purchaser signal that Sephora has.

Both retailers are also reportedly in conversations with the major AI assistants about structured product feeds that would expose Sephora's and Ulta's assortment data in a way the assistants can ingest natively. Whether those conversations produce commercial integrations or remain at the protocol level is an open question, but the direction of travel is clear: the retailers are trying to become the canonical product database for AI shopping, the way they were the canonical product database for SEO-era discovery.

The threat the retailers face is that the brands themselves have a structural advantage in ingredient and clinical citation that retailers cannot easily replicate. Sephora can publish a glossary of niacinamide. Drunk Elephant can publish the clinical study that demonstrated a specific niacinamide formulation reduced hyperpigmentation by a specific percentage in a specific sample over a specific duration. The clinical citation is the surface AI models trust most, and the brand owns it.

For a deeper view on how PDPs themselves are being restructured for shopping agents across categories, see the ecommerce AEO playbook for PDPs in the shopping agent era.

The INCIDecoder and EWG Weighting Problem

Two third-party ingredient databases have become disproportionately influential in AI beauty citation: INCIDecoder and the EWG Skin Deep database. Both have been around since well before the AI-shopping era, and both have always been used by ingredient-conscious consumers. What changed in 2026 is that AI assistants treat both as primary evidence sources when evaluating product safety and ingredient claims.

INCIDecoder is the more technically rigorous of the two. The site parses INCI lists, classifies ingredients by function, flags potential irritants and allergens, and produces structured product analyses that are essentially the schema-marked-up version of what brand PDPs should be doing. AI models quote INCIDecoder analyses directly in responses about ingredient compatibility, sensitivity, and concern. Brands whose products are listed on INCIDecoder with positive or neutral analyses gain citation share. Brands that are not on INCIDecoder, or whose listings are out of date, are systematically discounted in ingredient-led queries.

EWG Skin Deep is the more controversial database in industry circles because its scoring methodology has been criticized as alarmist by industry chemists. Whether the methodology is sound is a separate question from whether the AI models cite it — they do, frequently, in queries about ingredient safety, pregnancy-safe formulations, and clean beauty. Brands that score well on EWG are cited as clean. Brands that score poorly are flagged in AI responses with a hedge or a warning. The brands that have fought EWG ratings publicly without resolving the underlying ingredient disclosure issues have generally lost the AI citation battle, regardless of who is technically correct about the formulation.

The practical implication for brands is that third-party database presence is now a top-priority AEO action item. Submitting product information to INCIDecoder and EWG is not optional infrastructure — it is the equivalent of being listed in Google's product index for retail SEO. Brands that have not done this submission, or that have not updated their submissions since reformulating, are forfeiting a meaningful share of relevant queries.

The same dynamic plays out in adjacent categories where ingredient databases serve as primary evidence sources. The CPG and food and beverage AEO playbook around recipe and ingredient recommendations covers the parallel play in food and beverage, where USDA databases and Open Food Facts serve the role that INCIDecoder and EWG serve in beauty.

Real PDP Audit Data: 50 Brands, 12 Queries

We ran a structured PDP audit across 50 beauty brands — 20 prestige, 15 DTC, 10 drugstore, and 5 clinical — testing each brand's top three PDPs against 12 representative beauty queries on ChatGPT shopping mode, Perplexity Shopping, Claude, and Gemini. The patterns from that audit are the cleanest data we have on what is actually working in beauty AEO right now.

The 12 queries spanned skincare, makeup, and haircare, and were selected to represent the most common beauty discovery intents: best vitamin C serum for oily skin, retinol for sensitive skin beginners, foundation for mature skin with rosacea, hyaluronic acid serum that works under makeup, sulfate-free shampoo for color-treated hair, niacinamide serum for hyperpigmentation, pregnancy-safe skincare routine, clean mascara that does not flake, peptide moisturizer for fine lines, salicylic acid cleanser for acne, mineral sunscreen that does not leave white cast, and bond-repair treatment for chemically damaged hair.

The top-line findings:

Brands with full INCI disclosure in HTML appeared in 62% of relevant query responses. Brands with image-only INCI appeared in 14% of relevant query responses. The disclosure format alone — same ingredient list, different encoding — was a 48 percentage point swing.

Brands with explicit concentration disclosure on the lead active appeared in 71% of relevant query responses where concentration was a likely query factor (vitamin C, retinol, salicylic acid, niacinamide, hyaluronic acid). Brands without concentration disclosure appeared in 23% of those queries.

Brands with a clinical study citation on the lead product appeared in 79% of relevant query responses where clinical evidence was a likely factor (anti-aging, dark spot, acne, sensitivity). Brands without clinical citation appeared in 31%.

Brands present on INCIDecoder with a complete and current analysis appeared in 68% of ingredient-compatibility queries. Brands not present, or present with stale analyses, appeared in 22%.

Brands with EWG Skin Deep ratings of 1, 2, or 3 (the green range) appeared in 74% of clean beauty queries. Brands rated 7 or higher (the red range) appeared in 11% of those queries and were typically flagged with a hedge when they did appear.

Brands with substantive owned-editorial content (a brand-published ingredient glossary, application guide, or skin-concern explainer) appeared in 64% of educational-intent queries. Brands without owned editorial appeared in 28%.

The data points to a coherent operational priority: the brands winning beauty AEO have done the ingredient-disclosure, clinical-citation, and third-party-database work, and the brands losing have not. The marketing budget difference between the two groups is essentially noise compared to the infrastructure difference.

The Beauty AEO Playbook

The 90-day operational playbook for a DTC beauty brand or a beauty retailer that wants to move citation share in the next two quarters:

1. Audit your current citation rate. Run 50 to 100 category and ingredient queries across ChatGPT shopping mode, Perplexity Shopping, Claude, and Gemini. Document which products appear, which competitors appear, and which ingredient claims are quoted. The baseline citation rate is the foundation of every other decision.

2. Convert image-only INCI lists to HTML. This is the single highest-leverage change available to most beauty brands. The work is genuinely cosmetic — same content, different encoding — but the citation-rate uplift in the audit data is consistently in the 30 to 50 percentage point range. Many brands resist this change to protect formulation IP from competitor extraction. The data is clear that the citation-share cost of resisting exceeds the IP-protection benefit by a wide margin.

3. Add concentration disclosure on lead actives. For every PDP where a percentage matters (vitamin C, retinol, salicylic acid, niacinamide, peptides, hyaluronic acid), state the concentration explicitly in the product description and in structured data. The brands that disclose concentration are quoted directly in concentration-specific queries. The brands that hide concentration are absent from those queries entirely.

4. Publish clinical study summaries with methodology. If your brand has clinical study data on any product, publish a substantive summary page describing the study design, sample size, duration, and results. Link the relevant PDP to the study summary, and structure the summary with extractable claims (after 12 weeks, 89% of participants reported X) rather than narrative copy. AI models cite clinical study summaries as primary evidence in efficacy queries.

5. Submit to INCIDecoder and EWG Skin Deep. Both databases accept brand submissions and updates. Verify your top 30 products are accurately listed, with current formulations and complete INCI data. Brands that have not updated their submissions since reformulating are typically listed under outdated ingredient profiles, which damages citation share for the current product.

6. Build a brand ingredient glossary. Publish a brand-owned ingredient glossary with substantive entries on the actives in your product range. Each entry should be 200 to 400 words covering what the ingredient is, how it works, who should use it, who should avoid it, and what concentration ranges are typically effective. AI models cite brand-owned glossaries as authoritative when the glossary is substantive and accurate. The glossary also creates internal-link opportunities from every relevant PDP.

7. Restructure PDPs around skin concern and ingredient. The retailer-template PDP organized around brand voice and feature carousel is not the structure AI models reward. The structure that works is concern-led — open with the skin concern the product addresses, follow with the lead actives and concentrations, follow with the application guide, follow with the full INCI list, follow with clinical citation if available. Brands that have made this PDP restructure see citation-rate improvements within 60 to 90 days.

8. Instrument citation tracking. Sign up for one of the AI citation tracking tools that has beauty category coverage — Profound, Athena AI, or one of the emerging beauty-specific platforms. Build a weekly dashboard tracking share of category by ingredient, share of category by skin concern, and share of comparison citations against your top five competitors. Without measurement, the playbook above is guesswork.

For brands also navigating the broader shift toward AI agents that complete purchases on the buyer's behalf, see agentic commerce and the buy-on-behalf brand decision shift, which covers the transaction-layer dynamics that intersect with the discovery dynamics covered here.

What Beauty AEO Looks Like in 2027

The patterns we are seeing in 2026 will compound in 2027 in three directions.

Concentration disclosure becomes table stakes. The brands that disclose concentration today are gaining citation share. By late 2027, concentration disclosure will be a baseline requirement to appear in any ingredient-led query at all. Brands that have not made the change will be invisible in their categories.

Clinical citation becomes the primary differentiator at the prestige tier. Drugstore brands compete on price and accessibility. Prestige brands have historically competed on brand storytelling and packaging. In the AI-shopping era, prestige brands are competing on clinical substantiation, because that is the differentiator AI models reward at the price point. Brands without clinical study programs are at a structural disadvantage that marketing spend cannot close.

Retailer surfaces evolve into editorial properties. Sephora and Ulta will increasingly look less like retail destinations and more like editorial properties with transaction capability attached. The retailers that build the deepest ingredient editorial, the most rigorous clean beauty standards, and the most extractable category content will retain relevance as AI shopping agents become the primary discovery surface. The retailers that try to defend the search-and-filter model of 2018 will lose share to brand-direct discovery.

The window for brands to build the AEO infrastructure that will define category positions through 2028 is open right now. The audit data is unambiguous: the brands that ship the playbook in the next two quarters will compound citation share that competitors with larger marketing budgets cannot close. The brands that wait will spend the next three years buying their way into category conversations that the AI models have already settled.

Takeaway: Beauty AEO is an ingredient-disclosure, clinical-citation, and third-party-database problem before it is a content marketing problem. The brands winning AI citation share in 2026 — Glossier, Rare Beauty, Drunk Elephant, The Ordinary, Paula's Choice — have rebuilt their PDPs around extractable ingredient claims, concentration disclosure, and clinical evidence, and they have ensured presence on INCIDecoder and EWG Skin Deep. The retailers winning — Sephora and Ulta — have restructured their taxonomies and editorial properties to compete as evidence sources rather than just transaction destinations. The brands that ship the disclosure and substantiation playbook in the next 90 days will own category citation defaults through 2028. The brands that protect 2018-era marketing copy will lose share to disclosure-first competitors regardless of budget.