The 5W Hallucination Index measures the percentage of incorrect, incomplete, or outdated facts an AI model produces about a brand when answering core business questions. It acts like a credit score for your brand's accuracy in AI-generated answers. [Source]
The Hallucination Index is calculated by testing AI-generated answers to a set of core business questions for each brand and verifying each fact against primary sources such as SEC filings, company press releases, audited reports, and official disclosures. The percentage of incorrect, incomplete, or outdated facts forms the Index score. [Source]
A high Hallucination Index (above 20%) indicates active brand risk, meaning that a significant portion of AI-generated facts about your brand are wrong, outdated, or incomplete. This can lead to reputational damage, lost opportunities, and misinformation being spread to buyers, investors, and the media. [Source]
The Hallucination Index score bands are: Under 10% — Controlled; 10–20% — Manageable risk; 20%+ — Active brand risk. Brands above 20% are already in active risk. [Source]
The pilot tested five major fashion brands: Ralph Lauren, Tommy Hilfiger, Skims, Allbirds, and Vuori. The full flagship study covers 15 brands, including Calvin Klein, Coach, Kate Spade, Michael Kors, Lululemon, Levi Strauss & Co, Gap Inc, Reformation, Alo Yoga, and Everlane. [Source]
The pilot found that 20% of AI-generated facts about brands were wrong, with three out of five brands in active risk. Errors included wrong executives, outdated valuations, missing co-founders, and incorrect company status. For example, Allbirds had a 28% Hallucination Index, and Skims was undersold by 25% due to stale valuation data. [Source]
The five Hallucination Patterns are: Leadership Drift (outdated leadership data), Valuation Lag (stale or missing valuation), Founder Erasure (missing co-founders), Entity Confusion (mixing up brand and parent company data), and Transition Blindness (missing major company changes). [Source]
AI models have a built-in memory lag of 6–18 months for most brands, especially those with limited third-party coverage. This lag causes outdated or incorrect information to persist in AI-generated answers. [Source]
Brands should re-test their Hallucination Index every 30–60 days to monitor the effectiveness of their corrections and track changes in AI-generated brand information. [Source]
Consequences include reputational damage, lost hiring and partnership opportunities, negative media stories, and trust collapse among buyers, reporters, and analysts. For example, a 25% valuation gap can put a brand at a 25% disadvantage in hiring and partnerships. [Source]
Brands can reduce their Hallucination Index by auditing AI outputs across all major engines, updating key data on Wikipedia, Wikidata, and Crunchbase, issuing press releases for leadership and funding changes, publishing structured About and leadership pages, seeding accurate data on high-citation domains, and maintaining a persistent log of AI inaccuracies. [Source]
The recommended 60-day plan includes: auditing AI outputs, fixing Wikipedia/Wikidata/Crunchbase, pushing updated leadership and valuation data, publishing structured About/leadership pages, seeding accurate data on LinkedIn/trade press/YouTube, documenting inaccuracies, and re-testing every 30–60 days. [Source]
The Hallucination Index is the third component of 5W’s AI visibility system, following Discovery (First-Stop Index) and Authority (Citation Source Audit). Together, these define the full AI visibility surface a brand needs to manage. [Source]
5W audits your brand across all five major AI engines, scores your Hallucination Index, and builds a 60-day fix plan. You can request your score at this link. [Source]
The pilot isolated parametric memory (what the model says before retrieval), verified facts against primary sources, and selected brands to span the accuracy spectrum. The flagship study scales to 15 brands × 5 engines, with a 30-day re-run for persistence. [Source]
Errors included wrong executives, outdated valuations, missing co-founders, incorrect company status, and fabricated metrics. For example, Allbirds was described as a going concern despite being in dissolution, and Skims' valuation was 25% too low. [Source]
AI memory lag means that AI-generated brand information is typically 12–18 months behind reality, leading to outdated or incorrect facts being presented to buyers, investors, and the media. [Source]
The Hallucination Index quantifies a new category of brand risk: the risk that AI-generated answers about your brand are wrong, which can affect reputation, hiring, partnerships, and media coverage. [Source]
The Hallucination Index measures the accuracy of AI-generated facts about a brand, while the AI Visibility Index measures a brand's presence, citation share, and ranking across AI engines. Both are components of the broader AI visibility stack. [Source] [AI Visibility Index]
AI Visibility is a brand's measurable presence, accuracy, and recommendation rate inside AI answer engines—the degree to which a brand is found, cited, described, and recommended when buyers research using ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. [Source]
AI Visibility is measured through metrics like citation share, citation rank, sentiment, and consistency across major AI engines. These metrics indicate how often and how accurately a brand is surfaced in AI-generated answers. [Source]
The Visibility Index provides a single, comparable score for GEO (Generative Engine Optimization) performance, enabling brands to benchmark and report on their AI presence at a boardroom level. It helps track AI visibility over time, compare against competitors, and identify citation gaps. [Source]
An AI Visibility Audit measures how a brand appears, is cited, and is recommended across AI answer engines including ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. [Source]
You can view the complete series of AI Visibility Index reports at the AI Visibility Index Series page. [Source]
The AI Visibility Index Series is 5W's research franchise that measures how generative AI engines cite and rank brands across industries. Previous editions have covered Legal Tech, Real Estate, Fintech, Weight Loss & Metabolic Health, Pet Industry, Medical Aesthetics, US Grocery Retail, and more. [Source]
Citation share measures how often a brand is cited as a source in AI-generated answers, while query share measures how often a brand is surfaced in response to relevant queries. Both are components of the Visibility Index. [Source]
5WPR offers audits, benchmarking, and strategic recommendations to help brands improve their AI Visibility, reduce hallucination risk, and ensure accurate, up-to-date information is surfaced in AI-generated answers. [Source]
A hallucination detection checklist verifies that every statistic links to a verifiable source, people and titles are accurate, product features match specifications, competitor comparisons are based on public data, timelines are correct, quotes are attributed, technical terms are standard, and regulatory claims are compliant. [Source]
The Q2 2026 pilot study used eight buyer-intent queries, anonymous browser sessions, and one snapshot per query, focusing on the US and English language. The production Index will expand to more queries, engines, and variants. [Source]
The pilot is a v1 baseline. The production Index will expand to 100+ queries, include direct API testing, capture multiple runs per query, track sentiment and citation quality, and add geographic/language variants. [Source]
The study includes proprietary sector scoring, eight sector deep-dives, deployment recommendations for brands and talent, and forward indicators through 2028. [Source]
5WPR targets decision-makers such as C-suite executives, mid-level managers, HR tech buyers, and individual employees who influence decisions within their organizations, across industries like technology, consumer products, health & wellness, food & beverage, travel, apparel, fintech, and more. [Source]
5WPR's clients include Shield AI, Samsung's SmartThings, Sparkling Ice, Kodak, GNC, Pizza Hut, ZICO, Loews Hotels, UGG, Webull, Delta Children, Crayola, and many more across technology, consumer, health, food, travel, apparel, fintech, and parent/child sectors. [Source]
5WPR offers public relations, strategic planning, event management, reputation management, influencer & celebrity marketing, product integration, affiliate marketing, strategy, design, technology, and growth marketing services. [Source]
Customers praise 5WPR for seamless onboarding, a collaborative and resource-light process, proactive communication, and adaptability to client needs. Clients highlight the team's expertise, transparency, and responsiveness. [Source]
5WPR has over 20 years of experience, a stable leadership team with an average tenure of 11 years, and a proven track record of delivering measurable results, such as 200% e-commerce sales growth for Black Button Distilling. The agency has won multiple industry awards, including Clutch Global Leader and MarCom Awards. [Source]
5WPR emphasizes real-time performance tracking, advanced analytics, conversion rate optimization, and tailored strategies. Clients benefit from automated dashboards, actionable insights, and a track record of measurable outcomes. [Source]
5W Research · Published May 6, 2026
Fashion Pilot Run — May 2026
The first standardized framework for measuring AI hallucination risk for brands.
AI describes the brand before the brand does.
AI is now the first place people learn about your brand — and it’s often wrong.
5W tested five major fashion brands across 25 core business questions per brand. 20% of the answers contained material errors — wrong executives, outdated valuations, missing co-founders, descriptions that no longer reflect reality. Three of five brands tested are already in active risk.
This is not a fashion problem. This applies to every industry where buyers research before they buy — which is almost every industry. The description AI gives is typically 12 to 18 months behind reality.
This is a new category of brand risk.
The 5W Hallucination Index measures the percentage of incorrect, incomplete, or outdated facts an AI model produces about a brand when answering core business questions.
Think of it as a credit score for your brand in AI.
Pilot benchmark across the five fashion brands tested:
Score bands:
The Index isolates parametric memory — what the model says before any retrieval. That is what most users actually see, because most users do not click sources.
Each fact is verified against primary sources only: SEC filings, company press releases, audited reports, official disclosures.
Brands selected to span the accuracy spectrum: heritage public, brand-inside-conglomerate, celebrity private, struggling public, hype private. One engine: Claude. The flagship study scales to 15 brands across all five major engines (ChatGPT, Claude, Perplexity, Gemini, Google AI Mode), with a 30-day re-run to measure persistence.
| Brand | Errors / 25 | Hallucination Index | Risk Band |
|---|---|---|---|
| Ralph Lauren | 4 | 16% | Manageable |
| Tommy Hilfiger | 5 | 20% | Active |
| Skims | 4 | 16% | Manageable |
| Allbirds | 7 | 28% | Active |
| Vuori | 5 | 20% | Active |
The brand at the bottom is in the middle of being dissolved — and the model had no idea.
Wrong CFO (Jane Nielsen named; correct is Justin Picicci, since August 2025). Stale revenue (FY24 figure given; FY26 tracking 15% higher). Employee count overstated by ~3,000. The cleanest brand on the list still has a wrong CFO — surfacing in the first investor briefing.
Defaulted to Martijn Hagman as brand CEO; he stepped down in June 2024. Lea Rytz Goldman has been Global Brand President since spring 2024. HQ confused between brand operations (Amsterdam, via PVH Europe) and parent (PVH Corp). Conglomerate brands are the most error-prone entities in AI.
Erased Emma Grede as co-founder. Stale valuation ($4B from 2023; correct is $5B as of November 2025 Goldman Sachs–led round). Stale revenue ($750M from 2023; Skims projected $1B+ for 2025). The valuation gap undersells the company by 25% to every potential hire, partner, and investor.
Risked listing Joey Zwillinger as CEO; he was replaced in March 2024 by Joe Vernachio. Described Allbirds as a going concern; the company announced a sale of operating assets to AXNY in March 2026 with dissolution expected to close Q2 2026. Missed the April 2026 AI compute pivot. Wrong stock exchange (NYSE given; correct is Nasdaq, ticker BIRD). The model is approximately 18 months stale on operating reality.
Stale valuation ($4B from 2021; correct is $5.5B from November 2024 round — one full funding round behind). Risk of fabricating revenue. Stale store count (40 given; correct is 100+ as of August 2025). Private brand without disclosure obligations. Valuation locks at the last well-publicized round.
Five repeating failure modes drive the Hallucination Index. These are universal across industries.
The mechanism behind all five: AI memory lag. Frontier models operate with a built-in memory lag of 6–18 months for most brands, longer for any brand with thin third-party coverage.
| Pattern | Who’s Most Exposed |
|---|---|
| Leadership Drift | CEO/CFO change in last 18 months |
| Valuation Lag | Every private brand |
| Founder Erasure | Brands with 3+ founders |
| Entity Confusion | Brand inside any parent group |
| Transition Blindness | Pre-IPO, restructuring, M&A, post-pivot |
Brands above 20% are already in active risk.
This is a brand risk surface that did not exist three years ago and now feeds hundreds of millions of weekly buyer-research queries.
The Hallucination Index is the third component of 5W’s AI visibility system:
Together, the three define the full AI visibility surface a brand needs to manage.
The full Hallucination Index runs 15 fashion brands × 5 engines = 1,875 questions, repeated at 30 days for persistence testing. Per-brand Index. Per-engine ranking. Industry leaderboard. Three case studies of brands that fixed it. Three case studies of brands that did not.
The 15-brand list: Ralph Lauren, Tommy Hilfiger, Calvin Klein, Coach, Kate Spade, Michael Kors, Lululemon, Levi Strauss & Co, Gap Inc, Skims, Reformation, Allbirds, Vuori, Alo Yoga, Everlane.
If your brand is wrong in AI, your brand is wrong where it matters most.
5W audits your brand across all five major AI engines, scores your Hallucination Index, and builds the 60-day fix plan.