Which Consumer Categories Carry the Greatest Structural Exposure to AI Search Substitution
Everything-PR's quarterly index of the consumer verticals where AI answers are absorbing search demand — ranked by referral-traffic decline, sized by revenue exposure, scored by substitution rate and defensibility.
Executive Summary
The open web is repricing.
Two years after Google introduced AI Overviews and one year into the general availability of AI Mode, the consumer information economy is beginning to restructure around AI-mediated discovery. Publishers whose growth model depends on Google referral traffic — health information, product reviews, recipes, travel research, personal finance — face a measurable, ongoing compression of that funnel. The Reuters Institute's 2026 survey of 280 media executives across 51 countries records a median expected three-year search referral decline of 43%, with one in five publishers projecting losses above 75%.
This is not a temporary algorithm cycle. It is a structural shift in where consumer attention lands when a question is asked — moving from open-web destinations to closed AI answer surfaces: ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.
The Casualty List ranks the five consumer verticals where the shift is most measurable, the named publishers most exposed within each, and the AI engines now holding the answer surface.
Headline ranking — Volume I
Rank | Vertical | 12-month traffic loss (median, top brands) | Substitution rate | Defensibility |
|---|---|---|---|---|
1 | Personal Finance & Calculators | 33% | High | Low–Moderate |
2 | Health & Symptom Information | 43% | High | Low–Moderate |
3 | Recipes & Cooking | 30–80% (creator-level) | Very High | Low |
4 | Product Reviews & Comparison | 36% (Tom's Guide); 10% (Future B2C segment) | High | Moderate |
5 | Travel Research | 8–15% | Moderate–High | Moderate |
The pattern is consistent. Information-dense, low-transaction categories — finance, health, recipes — show the steepest compression. Categories with a transactional floor at the back end — travel bookings, lead-gen with applications — compress more slowly. Where Wall Street has named specific exposure (Chegg, NerdWallet, Future PLC, TripAdvisor, IAC/People Inc., Ziff Davis), the data is now in the filings.
Methodology
Data inputs. Similarweb and SEMrush traffic estimates (12-month rolling); Everything-PR Citation Share testing across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews on 50+ consumer-intent prompts per vertical; public company filings (10-K, 10-Q, 8-K) and earnings call transcripts from IAC, Future PLC, Ziff Davis, TripAdvisor, Chegg, NerdWallet; industry studies from Pew Research, Bain & Company, Digital Content Next, Reuters Institute, IAB Tech Lab, Chartbeat, Detailed.com; press disclosure via Bloomberg, Press Gazette, Digiday, AdExchanger.
Substitution rate estimates how often a representative set of consumer prompts receives a complete answer inside an AI engine without an outbound click. Modeled. Directional. No logged query runs.
Citation share measures which named sources the five engines pull from when answering category prompts. Reported as concentration metrics, not vote counts.
Financial figures verified against primary filings where available. Limitations in Section 10.
The Substitutability Framework
Substitutability is the central concept of the Casualty List. It estimates how much of a content category can be reproduced by an AI answer without the user needing the publisher.
High-substitutability content
Content that AI engines can compress into a complete answer without losing fidelity:
Definitions and reference — "what is APR," "what is HSA," "what is a beta-blocker"
Product summaries and spec sheets — feature comparisons, pricing tables, basic suitability
Recipe instructions — ingredients, steps, timing, substitutions
Informational listicles — "best practices for X," "top 10 ways to Y"
FAQ-style content — bounded, factual, repetitive
Itinerary synthesis — multi-day trip plans, restaurant lists, packing checklists
Comparison tables — side-by-side specs across products, services, or providers
These categories share three properties: low switching cost between sources, weak brand loyalty, and structured queries that compress well into a single response.
Low-substitutability content
Content the engines need to cite rather than reproduce:
Original reporting — interviews, scoops, investigations, on-the-ground accounts
Proprietary data — benchmarks, surveys, financial datasets, primary research
Personality-driven content — voice, perspective, opinion, persona
Active communities — real conversation, real questions, real-time replies
Experiential journalism — first-person testing, travel, taste
Live information — sports scores, market data, breaking news, weather
Transactional flows — booking, purchase, application, registration
The structural insight: high-substitutability content was historically the easiest to scale and monetize via SEO. It is now the most exposed. Low-substitutability content was historically harder to produce and harder to scale. It is now the moat.
The Defensibility Framework
Defensibility is the corollary metric. Where substitutability estimates how much of a category an AI engine can absorb, defensibility estimates how much of a publisher's value survives the absorption.
Nine defensibility layers
1. Proprietary data — benchmarks, datasets, original research the engines cannot reproduce without the source
2. Active community — Reddit-style discussion, comments, forums, user-generated context
3. Transaction ownership — booking, checkout, application — the user has to land somewhere to complete the action
4. Brand trust / institutional authority — Consumer Reports, Bloomberg, Mayo Clinic — credibility AI engines need to cite
5. Live information — sports, markets, news cycles, real-time inventory
6. Experiential content — first-person, present-tense, original observation
7. Personality / creator-led — voice that is the product
8. Paywalls and subscription — owned audience, recurring revenue floor
9. Original reporting — newsgathering the engines cannot generate
A category's structural exposure is the substitution rate net of the defensibility layers available within it. Personal finance has high substitutability and few defensibility layers outside transactional lead-gen — hence the ranking. Travel has high substitutability but a strong transaction-ownership layer at booking — hence the lower placement.
1. Personal Finance & Calculators
Verdict: highest structural exposure in the index.
Why this category compresses. Personal finance content is dominated by definitional, comparative, and educational queries — the precise format AI engines reproduce without source attribution. Information density is high; the legal floor for trust signals is real but not yet enforced at the answer surface; transactional intent sits at the end of the funnel, not the top.
The casualties. Investopedia — monthly visits estimated to have fallen from 40.3M to 26.9M across 2025, a 33.2% decline (Loopex Digital, November 2025). Part of People Inc. NerdWallet (NRDS) — flagged by Barclays and Dakota Wealth Management as a public-market AI-search loser; traffic-fed lead-gen model competing directly against ChatGPT and Gemini financial summaries. Bankrate — Red Ventures property; SEO-heavy comparison content; AI Overviews now answer most rate-comparison queries inline. The Balance — People Inc.; named by IAC as among the titles most affected by AI Overviews on health and finance queries. Chegg (CHGG) — adjacent through homework-help and academic finance content. Stock down approximately 99% over 39 months. Two rounds of layoffs in 2025 — 22% in May, then 45% (388 employees) in October. The first publicly traded company to officially attribute revenue decline to AI. Trading near $1 in early 2026.
What recovers. Tools tied to actual offers — mortgage applications, brokerage account opening, advisor matching, deposit funnels. Lead-gen with a transaction at the back survives. Pure educational content does not.
2. Health & Symptom Information
Verdict: second-highest exposure. Sharpest single-brand compression in the index.
Why this category compresses. Symptom-input queries are well-suited to LLM triage: structured, encyclopedic, oriented toward an answer rather than a destination. The category is heavily trained, heavily duplicated across sources, and weakly differentiated by brand at the entry level. Regulatory exposure slows full substitution at prescription and diagnostic edges but does not protect first-line educational content.
The casualties. WebMD — monthly visits estimated to have fallen from 122M to 69.5M across 2025, a 43.1% decline (Loopex Digital, November 2025); the second-largest absolute consumer-web traffic decline of the year. Healthline — Red Ventures' largest property; first-line symptom queries now substituted by ChatGPT, Claude, and Gemini. Verywell network (Verywell Health, Verywell Mind, Verywell Family, Verywell Fit) — People Inc. portfolio; the company confirmed health is one of the most affected categories. Mayo Clinic content — institutional brand still ranks; referral patterns echo the category.
Disclosure pattern. People Inc. confirmed in Q1 2025 filings that titles in the health and finance categories have been most vulnerable to Google AI Overviews. People Inc.'s overall traffic from organic search has fallen from 60–65% in 2021 to 25–30% by mid-2025 — a managed decline at the parent level, masking sharper compression inside specific brands.
What recovers. Provider finder tools, telehealth funnels, condition-specific community, treatment-anchored content where the user needs an outcome, not an explanation. Citation share inside AI engines is the survival metric for branded pharmaceutical and Rx-adjacent content.
3. Recipes & Cooking
Verdict: highest creator-level compression in the index. Widest spread between large incumbents and independents.
Why this category compresses. Recipes carry low switching cost, structured queries ("how do I make chicken stock"), highly repetitive informational format, and weak brand loyalty. Users want the recipe — not the preamble, not the affiliate scroll. AI engines deliver the recipe directly. The category was substitutable before the AI engines arrived; "Jump to Recipe" browser extensions were the leading indicator.
The casualties. Independent food creators report year-over-year traffic declines of 30–80% during 2025, concentrated among creators dependent on Google discovery (Bloomberg, November 2025). Clean Eating Kitchen (Carrie Forrest) — 80% of traffic and revenue lost over two years; entire team laid off. Easy Peasy Foodie (Eb Gargano) — seasonal traffic patterns broke during the 2025 holiday cycle, the highest-revenue period of the food-blogging year. Inspired Taste (Adam and Joanne Gallagher) — impressions on Google rose; clicks fell. The signature of substitution. Allrecipes (People Inc.) — meaningful decline reported by Loopex Digital; partially offset by parent's diversified distribution. Food Network (Warner Bros. Discovery) — recipe SEO content under pressure; television brand offsetting partially.
Signal event. Google's Recipe Quick View experiment — launched October 2024, ended July 2025 after creator pressure — was a tacit acknowledgment that surfacing recipe content directly on the search page is incompatible with publisher economics. Bloomberg reporting confirmed Google's compensation to participating creators was insufficient to offset lost ad revenue.
What recovers. Video execution, community, paid subscriptions, recipe boxes, kitchenware affiliate funnels tied to specific products. The recipe itself is no longer the asset. The brand around the recipe is.
4. Product Reviews & Comparison
Verdict: severe at the brand level. Most concentrated public-market exposure.
Why this category compresses. Comparison synthesis is precisely what AI engines do well. The reader's question — "which X should I buy" — maps cleanly to a single response with a recommendation and a comparison table. Hands-on testing, exclusive product access, and original measurement are the remaining defensibility layers. Editorial brand alone is not.
The casualties. Tom's Guide (Future PLC) — monthly audience down 36% in the 18 months to December 2025, falling to 1.97M (Ipsos iris). TechRadar (Future PLC) — December 2025 audience of 3.74M, down 5% from July 2024. Steadier than Tom's Guide; still trending down. Future PLC B2C division — sessions down 10% to 317 million for the fiscal year. E-commerce revenue down 24% against a 9% audience decline. UK digital advertising down 14% (8% organically). Stock down approximately 50% year-to-date by late 2025; Barclays cut its price target from 640p to 285p in April 2026; Stifel forecasts FY26 revenue of £706.2 million, a 4.5% decline. CNET — sold by Red Ventures to Ziff Davis in October 2024 for over $100 million. Multiple rounds of layoffs in the years prior. PCMag, Mashable, IGN (Ziff Davis) — review and tech traffic under pressure; Ziff Davis building out lifestyle exposure via Recurrent Ventures acquisitions in 2026 as offset. Wirecutter (The New York Times) — protected by the NYT subscription wrapper and AI licensing agreements; less exposed at the brand level than standalone competitors.
Strategic signal. Future PLC's H1 2026 commentary explicitly repositioned the business around "human-originated and specialist content" — code for the substitution-resistant defensibility layers above. The SEO-traffic model is no longer the strategic anchor.
What recovers. Original testing labs (RTings, Wirecutter, Consumer Reports), exclusive product access, video-first reviews, and citation share inside AI engines. The publishers that get cited inside the answer — rather than competing with it — retain influence even as traffic compresses.
5. Travel Research
Verdict: lower exposure than the four categories above — for now. Transaction floor remains. The next 18 months will compress that floor.
Why this category compresses (and partially resists). Itinerary synthesis, restaurant lists, and trip planning are textbook AI use cases — OpenAI, Google, and Anthropic actively demonstrate them in product launches. But travel requires a transaction at the end: a booking, a reservation, a purchase. The information layer compresses fast. The transaction layer compresses slower.
The casualties. TripAdvisor (TRIP) — Brand TripAdvisor media and advertising revenue down 13% in Q2 2025 and 11% in Q3 2025. CEO Matt Goldberg explicitly cited the changing search landscape and rise of AI Overviews. CFO Michael Noonan projected that by year-end 2026, less than 10% of Experiences gross booking volume will come from organic search. Activist investors Starboard Value (9% stake) and Palliser Capital pushing for strategic alternatives, including a sale. Lonely Planet, The Points Guy, Frommer's, Fodor's — legacy travel content largely substituted for itinerary planning. Expedia — website traffic down 4.9% year-over-year; Google search queries down 17.8% (Loopex Digital). More insulated than TripAdvisor by transactional volume, but warning signs present.
Structural signal. Barclays research highlighted by CNBC ranks Airbnb (high direct traffic) as more insulated than TripAdvisor (29% direct traffic) — the exact spread the data confirms. Direct-traffic share is now the resilience metric for the category.
What recovers. Booking transactions, vertical experiences (Viator), local supplier networks, brand-anchored apps. The information layer is exposed. The transaction layer is the moat.
6. The Insulated
The full picture requires the counter-list: who outperforms when informational traffic compresses.
Beneficiary | Why it benefits |
|---|---|
Active community content becomes highly citeable by AI engines, increasing visibility and referral traffic as models rely on discussion-based sources. | |
Wikipedia | Gains sustained dominance as a neutral, structured knowledge base that every major AI system consistently cites. |
The New York Times / Wirecutter | Benefits from paywalls, original reporting, and licensing deals that turn content into a paid input source for AI systems. |
Wall Street Journal / News Corp | Strong financial upside from large AI licensing agreements and early positioning as a premium data provider. |
Bloomberg | Reinforces dominance in proprietary financial data where AI systems depend on licensed or structured institutional feeds. |
Financial Times | Gains value from paywalled investigative journalism and licensing agreements with AI companies. |
Consumer Reports | Strengthened authority through subscription-based access and trusted product testing data that AI cannot easily replicate. |
Substack creators with paid lists | Direct monetization through owned audiences and personal trust relationships that AI cannot replicate or replace. |
YouTube and creator-led video | Benefits from format defensibility (video, personality, engagement) that AI cannot fully substitute or compress. |
Costco / Costco-style trust brands | Gains from strong community trust, experiential retail, and brand loyalty that reinforces offline + online defensibility. |
Live news, sports, markets | Increased dependency as real-time data sources that AI systems must continuously fetch, cite, and update. |
The pattern: defensibility compounds. A single layer — a paywall, a community, a transaction floor — buys time. Stacked layers buy structural position. The publishers that survive AI substitution carry three or more of the nine defensibility layers. The publishers on the Casualty List rarely carry more than one.
7. Aggregate Findings
Ownership concentration. Four public parents carry the bulk of named exposure across the five verticals. People Inc. (formerly Dotdash Meredith, NASDAQ: IAC) — Investopedia, The Balance, Allrecipes, Verywell network; health and finance verticals confirmed as most affected. Future PLC (LSE: FUTR) — Tom's Guide, TechRadar; stock down sharply; FY26 revenue guidance cut; 45 editorial redundancies announced January 2026. Red Ventures (private) — Bankrate, Lonely Planet, The Points Guy, Healthline; sold CNET to Ziff Davis October 2024. Ziff Davis (NASDAQ: ZD) — CNET, PCMag, Mashable, IGN, RetailMeNot, Everyday Health; aggressive M&A through 2025–2026; sold Connectivity division to Accenture for $1.2B in March 2026, acquired theSkimm and Recurrent Ventures lifestyle brands.
The consolidation pattern matters. M&A inside the affiliate-publisher economy through 2026 is defensive, not strategic. The active question is which parent ends up holding the most defensible portfolio — and which becomes a forced seller.
Beneficiary mapping — who holds the answer surface. ChatGPT — leading citation share in personal finance, recipes, and product comparison; favors Wikipedia, Investopedia (still cited even as it loses traffic), Reddit, original product manufacturer sites, and a small editorial set (NYT/Wirecutter, Consumer Reports). Google AI Overviews — heavy concentration on Reddit, YouTube, and Google's own structured data. Perplexity — most diverse citation behavior; highest probability of citing a publisher byline. Claude — favors authoritative, encyclopedic sources; Reddit and Wikipedia anchor most consumer queries. Gemini — Google ecosystem favored; YouTube and Google product citations prevalent.
The publisher implication: citation share inside AI engines is now the leading indicator. Traffic is the lagging indicator.
The lawsuit map. Penske Media became the first news publisher to sue Google over AI Overviews. The New York Times and Chicago Tribune sued Perplexity in late 2025. Ziff Davis sued OpenAI in April 2025. Encyclopedia Britannica and Merriam-Webster filed against OpenAI. Reddit sued Anthropic in June 2025. Anthropic settled a $1.5 billion copyright case in September 2025, establishing a $3,000-per-work baseline for AI training content. The settlement changes negotiating leverage for every unlicensed publisher in the market.
8. The Three Patterns
Pattern one: information-dense categories compress faster than transactional categories. Personal finance and health (high information, low transaction) are compressing faster than travel (high information, high transaction). The transaction floor is the survival floor.
Pattern two: scale buys time, not immunity. Future PLC's audience is down 10%. Independent food creators are down 30–80%. People Inc. has reduced Google's share of its traffic from 60–65% to 25–30% in four years — the diversification works at scale but is structurally unavailable to a single-creator business.
Pattern three: AI licensing revenue offsets but does not replace. People Inc.'s licensing revenue grew 30% year-over-year in Q1 2025, driven by its OpenAI partnership. The 30% growth on a smaller revenue base does not offset a 3% decline in core sessions across a much larger ad revenue base. Licensing is a buffer. It is not a business model.9. What Comes Next
The Reuters Institute's 2026 survey records publishers expecting search referrals to decline a median 43% over the next three years. One in five expects losses above 75%. These are planning assumptions, not forecasts.
Three near-term inflection points: Google AI Mode default behavior (whether and when Google promotes AI Mode to the default search experience); OpenAI's shopping pivot (ChatGPT moving from research to transactional decisions); and antitrust outcomes (the Penske case against Google, and the DOJ's broader Google posture, will set the regulatory floor for how much of the open web's economics can be reabsorbed).
The Casualty List will track all three.
10. Limitations
Traffic figures are modeled estimates from third-party measurement tools (Similarweb, SEMrush, Ipsos iris, Loopex Digital), not server-log data from the publishers themselves — direction is reliable; absolute magnitudes are directional. Substitution rates are estimated by assessing whether a representative prompt set produces a complete answer without an outbound click; edge cases coded conservatively. Citation Share testing is based on a fixed prompt set per vertical; different prompt sets produce different citation distributions; no queries were logged or retained. Some declines attributed to AI substitution are confounded with other Google changes over the same window — Reddit promotion, affiliate site throttling, AI Mode rollout. We flag the overlap rather than attempt to decompose it.
11. The Standing List
The Casualty List is updated quarterly. Volume II — Q3 2026 — adds two further verticals to the index; under consideration: education and tutoring, legal information, local services, and translation.
Submissions, corrections, and parent-company filings for Volume II: research@everything-pr.com
About Everything-PR
The Casualty List is an Everything-PR research franchise. Permanent home: everything-pr.com/casualty-list/. Updated quarterly.
