Perly Consulting │ Beck Eco

The State of Play

A living index of AI adoption across industries — where established practice meets the bleeding edge
UPDATED DAILY

The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.

The Daily Dispatch

A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.

AI Maturity by Domain

Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail

DOMAIN
BLEEDING EDGEESTABLISHED

Visual search, product matching & virtual try-on

LEADING EDGE

TRAJECTORY

Stalled

AI that enables visual product search, similarity matching, and virtual try-on experiences for apparel, accessories, and cosmetics. Includes image-based shopping and augmented reality fitting; distinct from retail analytics which analyses behaviour rather than enabling purchase.

OVERVIEW

Visual search and virtual try-on occupy an unusual position on the maturity curve: commercially deployed and generating measurable ROI for forward-leaning retailers, yet far from mainstream adoption. Visual search has commoditized into platform infrastructure -- Google Lens alone processes billions of monthly queries -- but virtual try-on remains sharply bifurcated. In accessories, eyewear, and beauty, VTO delivers proven conversion lifts and return reductions that justify investment. In apparel, the story is different. Technical realism gaps, 3D asset creation bottlenecks, and the stubborn economics of free returns have kept most retailers on the sidelines. The practice sits at leading-edge: a handful of named deployments demonstrate clear value, but the majority of the market has not started, and independent practitioners report that vendor ROI claims often fail to replicate in production. The central tension is not whether the technology works in controlled settings -- it does -- but whether it can scale beyond accessories and beauty into the far larger apparel market where fit complexity and returns economics resist it.

CURRENT LANDSCAPE

Google Lens has quintupled visual search volume over two years to 20 billion monthly searches, with e-commerce platforms reporting 25-30% conversion improvements from visual search integration. Q1 2026 saw Google expand virtual try-on to footwear across Australia, Canada, and Japan following strong US adoption, indicating commoditization of VTO at major platform level. That side of the practice is effectively solved infrastructure. Virtual try-on is where the action and the friction concentrate. Perfect Corp dominates the vendor ecosystem with 800+ brand partners, 120M daily user events, and continued API expansion into jewelry, watches, and personal care. Eyewear VTO has reached 67% adoption among online shoppers, and jewelry platforms report 94% higher conversion from AR interaction. Shopify app launches like Slash AR signal early democratization for SMBs. Major brands (Sephora, Gap, Walmart, ZARA) deployed VTO solutions in Q1 2026 to address the $365B returns problem.

However, the critical gap persists. Shoptalk 2026 (10K attendees) revealed that while major retailers deployed VTO, real-world limitations surfaced: dataset bias in training data reduces accuracy for underrepresented demographics, and deployment quality varies dramatically. Independent technical assessments documented persistent barriers: accuracy limitations across commercial tools (sizing recommendations vary by 2+ sizes between platforms), pattern fidelity failures (complex geometries and textures hallucinate), and adoption friction (technical capability outpaces consumer trust). Taobao's high-profile AI VTO launch sparked consumer backlash when marketing claims (personalized fit) diverged from delivery (generic face-swap outputs); analysis identified that successful VTO appears in seller product photography workflows, not consumer-facing try-ons. Most VTO vendors struggle to move beyond pilot stage due to 3D asset pipeline bottlenecks and per-asset manual fixes. BIPA litigation remains unresolved, and for many apparel retailers, free returns remain cheaper than VTO investment. The market valued at $7.25B in 2025 continues growing, but growth concentrates in verticals where the product is rigid (eyewear, jewelry) and the try-on is simple, with apparel adoption constrained by technical and economic barriers despite vendor maturation.

TIER HISTORY

ResearchJan-2017 → Jan-2017
Bleeding EdgeJan-2017 → Jan-2018
Leading EdgeJan-2018 → present

EVIDENCE (129)

— Official Q1 2026 SEC filing shows Perfect Corp revenue growth, profitability, and scale serving 800+ brands with 120M daily user events on AWS.

— Market analyst report quantifies VFR market growing 19.26% CAGR from $6.24B (2025) to $21.45B (2032), documenting technology advancement and adoption trends.

— SIGGRAPH 2026 research introducing FIT dataset (1.13M triplets) with physics-based garment simulation. Addresses critical fit-prediction gap limiting commercial adoption.

Guidance for Virtual Try-On on AWSProduct Launches

— AWS official guidance using Bedrock Nova Canvas for VTO. Major cloud vendor product GA signals infrastructure maturity enabling broad adoption.

— Google Research + University of Washington released FIT dataset (1.13M samples) accepted to SIGGRAPH 2026, directly addressing VTO's core unsolved problem—fit prediction vs. visual aesthetics.

— Comprehensive report: 92% Gen Z adoption, 60% US AR users, 98% satisfaction. Documents shift from novelty to strategic priority across enterprise retailers.

— Critical analysis documenting decade-long failure of virtual try-on to solve returns crisis, providing balance to vendor adoption optimism.

— Google India expands VTO across Gemini shopping, AI Mode, Circle to Search with 50B product Shopping Graph. Demonstrates platform consolidation and unified infrastructure.

HISTORY

  • 2017: Visual search launched at scale (Pinterest Lens, ASOS) with strong engagement but mixed technical maturity. Virtual try-on emerged as research-backed consumer need, though implementation relied on subjective measures rather than scientific fit prediction.
  • 2018: Visual search reached measurable scale (Pinterest Lens 600M/month searches, Alibaba Pailitao 10M+ daily users) with mainstream retailer adoption (Home Depot, Target, Macy's). Virtual try-on showed positive purchase intent impact in peer-reviewed studies but remained technically limited (~70% accuracy), still relying on subjective fit rather than body reconstruction or garment simulation.
  • 2019: Visual search accelerated to mainstream adoption with Pinterest expanding to 2.5B detectable objects and ViSenze securing Series C funding backed by ecosystem of major retailers (Rakuten, ASOS, Urban Outfitters, Zalora, Uniqlo). Virtual try-on remained in early adoption despite peer-reviewed advances (ICCV 2019 research) and emerging vendor momentum, with interactivity identified as key adoption driver; implementation barriers and subjective fit reliance persisted.
  • 2020: Visual search continued scaling with Pinterest Lens searches tripling year-over-year, A/B testing confirming 160% relevance gains, and mainstream integration (2.5B detectable objects, doubled fashion attributes, 50% accuracy improvement). Virtual try-on remained research-focused: EU-funded eTryOn project developed XR prototypes and algorithms, academic advances continued (AIST unsupervised learning), but no consumer-facing commercial breakthroughs; consumer awareness plateaued at 35% despite market forecasts predicting 500% growth through 2030.
  • 2021: Visual search expanded globally with Pinterest's Lens rolling out to Australia, Canada, France, Germany; ViSenze remained active as provider across retailer ecosystems. Virtual try-on crossed from pure research into vendor-led pilots and brand adoption (Vertebrae case studies documented revenue impact); systematic reviews synthesized 69+ studies on VTO adoption drivers, indicating maturation of research foundation for commercial deployment.
  • 2022-H1: Visual search reached 8B monthly searches (Google Lens, 2.5x YoY growth), while major platforms accelerated entry (Google Cloud Retail Search GA with 20% search-refinement reductions; Amazon AR try-on launch for shoes). ViSenze operated at 1B+ monthly queries from major retailers with measured 10-29x ROI. Virtual try-on saw continued algorithmic research addressing technical challenges (occlusion, misalignment, garment deformation) while empirical studies identified interactivity and consumption vision as key adoption drivers; Amazon and other major players signaled commitment despite acknowledging VTO remained experimental.
  • 2022-H2: Visual search deployments proliferated with Flipkart's Shopsy visual search in production using multi-task learning, ViSenze extending ecosystem to 900+ clients. Virtual try-on crossed into mainstream deployment: Walmart's Be Your Own Model scaled to 270,000+ items; 3DLOOK demonstrated quantified ROI (47% return reduction for TA3 SWIM). Systematic review synthesized consumer behavior research documenting adoption drivers; privacy concerns and technical barriers identified as persistent resistance factors despite proven purchase intent impact.
  • 2023-H1: Visual search surpassed 10B monthly searches (Google Lens), confirming mainstream adoption and commoditization across platforms. Virtual try-on accelerated with Google and Amazon launching production features and 3DLOOK showing 48% return reduction, yet BIPA litigation against major brands (Estée Lauder, Louis Vuitton, Christian Dior) exposed legal barriers; market forecasts projected $20B industry by 2030 despite regulatory and consumer trust headwinds.
  • 2023-H2: Visual search adoption continued scaling: ViSenze processed 175M searches in India (Oct 2023, 30% YoY growth) and served 1B+ queries globally across 900+ retailers; infrastructure maturation documented in Pinterest's DERM embedding model. Virtual try-on vendor ecosystem expanded: ModiFace powered 84 beauty brands with 6x engagement and 2x conversion uplift; retailer case studies (Big W, ASOS, Zara) confirmed ROI ($30% return reduction); market research documented sustainability and personalization drivers. Open-source momentum accelerated (VStar 694 GitHub stars). Yet BIPA litigation precedent from 2023-H1 remained unresolved, creating deployment friction despite quantified ROI.
  • 2024-Q1: Virtual try-on advanced algorithmically with ECCV 2024 research on diffusion-model authenticity improvements (IDM-VTON), while commercial deployments expanded: Fittingbox documented eyewear VTO ROI across SmartBuyGlasses, Specsavers, Fielmann; Perfect Corp reported 50% conversion increase and 4x engagement uplift; luxury brands (Gucci, Chanel, Luxottica) confirmed AR try-on adoption with 65% conversion lift. Visual search adoption metrics stabilized with 36% consumer penetration and 62% millennial preference. Physical retail expansion signals emerged with Zero10's AR Mirror rollout beyond e-commerce, indicating hybrid deployment trajectory.
  • 2024-Q2: Virtual try-on advanced with systems-level literature reviews documenting engagement and return-rate benefits, while fundamental computer vision limitations remained documented (30% error rates on abnormal imagery). Pinterest expanded API tooling for visual commerce (14% CPA improvements claimed), and regional metrics showed sustained adoption: visual search in India grew to 1.4B searches with 72% fashion concentration. Academic research synthesized persistent technical challenges (fabric simulation, body modeling) and ethical concerns (privacy, data security) alongside proven ROI in eyewear and beauty verticals.
  • 2024-Q3: Virtual try-on expanded into new product categories: Perfect Corp + TIMEZ launched AR watch VTO with hand-tracking (53K+ models) and Meesho deployed saree VTO on GCP with 2D/3D pipelines. Visual search infrastructure continued scaling: Pinterest Lens searches tripled since 2019 with 50% accuracy improvements and integrated Shop tab (June 2024); feature updates (product pins, prices, stock info) advanced commerce integration. However, underlying AI visual reasoning constraints surfaced: research on GPT-4o and Gemini 1.5 Pro found failures at basic tasks (18% accuracy on overlapping shapes), indicating foundational limitations in multimodal model visual understanding despite continued algorithmic advances in VTO fidelity.
  • 2024-Q4: Perfect Corp achieved continued scale: 977k+ paying subscribers (17% YoY growth), $44.3M Q1-Q3 revenue (12.5% growth), serving 700+ brands globally. Academic validation accelerated: peer-reviewed studies from European and Asian institutions documented VTO's positive impact on purchase intention in beauty (L'Oréal) and regional platforms (Shopee, Indonesia), but also confirmed intrusiveness and adoption friction as persistent barriers. Critical gap surfaced: Microsoft's Bing Visual Search API reported real-world product matching failures in production, indicating ongoing infrastructure brittleness. Industry analysis emerged warning that VTO's adoption remained surprisingly limited in mainstream retail despite technical maturity, with free returns often more economical than VTO investment and social media platforms shifting focus from product discovery to content consumption—signaling that commercial viability remains contingent on returns economics and consumer behavior shifts, not technology maturity alone.
  • 2025-Q1: Visual search adoption continued ecosystem scaling with ViSenze processing 35% higher volumes on Myntra and Pinterest Lens achieving Shop tab integration with inventory features. Virtual try-on expanded into physical retail: Amazon's Milan Parafarmacia store deployed self-service AI skin analysis mirrors (Feb 2025); cruise lines (Royal Caribbean, Carnival) expanded fleet-wide tools. Perfect Corp reported $60.2M annual revenue (+12.5% YoY), 1M+ subscribers, 732 brands served. Eyewear VTO adoption doubled since 2022: 29% of glasses shoppers used VTO, with 67% adoption among online eyewear shoppers and 215M annual virtual try-ons (+49% YoY). Market maturation indicators: McKinsey 2025 showed 68% of shoppers prefer VTO-enabled retailers; estimated 25-30% of major apparel retailers now offer VTO. Yet regulatory headwinds continued with unresolved BIPA litigation, and underlying AI visual reasoning remained constrained.
  • 2025-Q2: Visual search adoption accelerated with 300% surge in retail/logistics deployment over two years; Home Depot integrated visual search for part identification, and Belcorp deployed Perfect Corp VTO across 13 countries with increased conversion rates and basket sizes. ViSenze's Multi-Search platform reached 300+ global customers (Meesho 1B annual searches, Zalora 10-15% engagement/AOV lifts). Market research confirmed maturation: virtual try-on market $12.09B (26% CAGR growth), 60% of US consumers frequent AR users, 94% higher conversion with 3D/AR content. However, technical barriers remained unresolved: research identified computational demands, API documentation gaps, garment misalignment, and basic 2D overlay limitations; manufacturing practitioners confirmed VTO's poor accuracy for complex apparel fit. Critical gap remained unaddressed: underlying computer vision systems still failed basic visual reasoning tasks.
  • 2025-Q3: Infrastructure maturation and vendor validation dominated Q3: Perfect Corp processed 120M daily user events on AWS with 99.99% global availability across 800+ brand partners; Tarte Cosmetics reported 200% sales uplift and triple-digit session time growth from VTO deployment. VC investment accelerated with Veesual raising $7.5M and Doji $14M for photorealistic avatar technology. Google expanded VTO rollout across billions of products while Sephora and Warby Parker reported 11% and 85% conversion lifts. Market synthesis showed 27-65% increased purchase likelihood after AR interaction and sector-wide 200% engagement gains. However, critical limitations persisted: deployment required careful orchestration (expert warning of poor-execution risks), privacy and tactile-feedback concerns offset ROI metrics, and BetterMirror analysis emphasized that accuracy gaps and free returns economics often undercut VTO adoption drivers.
  • 2025-Q4: Platform consolidation and market forecasting dominated Q4. Google extended Try-On feature to shoes and new geographies (Australia, Canada, Japan) with depth perception visualization. Perfect Corp maintained scale with 800+ brand partners processing 120M daily events and 300M API calls at sub-3-second latency; Ulta and Watsons reported double-digit sales growth. Market research forecasts quantified trajectory: VTO market $12.5B (2024) expected to reach $48.8B by 2030 at 25.5% CAGR; Smartphones/Tablets segment projected 26.8% CAGR to $42.9B. Consumer adoption metrics showed 10% regular visual search usage, 36% trial penetration, 42% interest in the U.S.; Google Lens reached 20B monthly searches (4B shopping-related). Practitioner analysis aggregated deployment metrics: 18-25% conversion lifts and 20-40% return reductions documented; case studies from Onix Systems, Mocky.ai, and Zalando confirmed ROI in multiple verticals. However, adoption remained bifurcated with visual search commoditized (low strategic value, massive volume) and VTO adoption-constrained in apparel by technical limitations and returns economics despite high ROI in eyewear and beauty.
  • 2026-Jan: Vendor ecosystem expansion and production deployment consolidation. Perfect Corp launched nine new modular APIs expanding fashion VTO to watches, bracelets, rings, earrings, necklaces, scarves, hats, shoes, and bags; global VTO market sized at $7.25B (2025) with 15.36% CAGR projection to $17.09B (2033). Regional production deployments documented: Zara activated 'probador virtual' in Spain/Italy with 2-minute processing, Doppelgänger deployed photorealistic avatars on mobile, Google Doppl expanded to US. Jewelry segment showed strong adoption signals: 60% consumer digital preference, 94% higher conversion from AR interaction, 40%+ session time increases. Google Lens sustained massive scale at 12B+ monthly visual queries. However, industry assessment highlighted persistent scalability challenges: VTO vendors reported difficulty scaling beyond pilots due to 3D asset creation bottlenecks, catalog integration gaps, and per-asset manual fixes; underlying computer vision limitations and free returns economics continued constraining apparel adoption.
  • 2026-Feb: Ecosystem democratization and critical gap surfacing. Shopify app launch (Slash VTO) signaled SMB accessibility expansion; Perfect Corp partnerships expanded to personal care (LashLovr with True Beauty Lashes). Google Lens quintupled visual search queries to 20B monthly searches over two years, with e-commerce platforms reporting 25-30% conversion improvements. However, independent practitioner assessment (Alle fashion startup) documented persistent gap: VTO showed weak adoption (15-25%), did not reduce returns contrary to vendor claims, with high per-image costs ($0.50-$2) and latency (>15 sec) limiting real-world utility. Academic validation continued (Vision-DeepResearch Benchmark) but underlying tension remained: documented case studies (StyTrix) reported 36% return reduction and 94% conversion lifts, yet independent deployments failed to replicate ROI. Overall signal: vendor partnerships and ecosystem growth continue, but practitioner reality diverges from marketed capabilities.
  • 2026-Mar: Major brand deployments met critical backlash. Google expanded VTO to footwear globally; Sephora, Gap, and Walmart deployed VTO solutions at Shoptalk 2026 (10K attendees), where dataset bias for underrepresented demographics emerged as a documented limitation. Taobao's high-profile VTO launch sparked consumer outrage when personalized-fit promises delivered generic face-swap outputs, with analysis identifying seller photography workflows -- not consumer-facing try-ons -- as where VTO actually functions reliably. Independent technical assessments confirmed that misaligned fits can increase returns rather than reduce them; practitioner and industry analysis converged on the conclusion that consumer adoption barriers (accuracy, latency, representation) remain unsolved despite seven years of vendor investment.
  • 2026-Apr: Major platform consolidation and research validation. Google completed rollout of virtual try-on into all product search results (April 30, 2026), enabling tens of millions of daily virtual fitting sessions across 5B+ daily searches with documented 18-32% return reduction and 15-25% conversion lift. Perfect Corp filed Q1 2026 SEC report showing continued profitability and scale at 800+ brand partners with 120M daily events. Critically, Google Research and University of Washington released FIT dataset (1.13M samples, accepted to SIGGRAPH 2026) directly addressing VTO's core unsolved problem—fit prediction accuracy. Independent critical assessments persisted: academic and vendor research documented that current VTO limitations (realism gaps, accuracy variations by 2+ sizes between platforms) mean that purchase confidence depends on factors beyond visual fidelity. Market research quantified growth trajectory at 19.26% CAGR ($6.24B in 2025 to $21.45B by 2032), with adoption concentrated in verticals where garment complexity is low (eyewear, jewelry, beauty) and apparel segment adoption remaining constrained by technical fit accuracy barriers and free returns economics despite vendor maturation. Startups (Catches, Genlook, Amiri) documented deployment outcomes (2.3x conversion lift, 38% return reduction), yet decade-long failure to solve returns crisis remained evident, indicating that technology maturity alone is insufficient for broad market adoption.
  • 2026-May: Google's VTO expansion to all product search results and Perfect Corp's Q1 2026 profitability report confirmed platform-scale consolidation, while Google Research and University of Washington's FIT dataset (1.13M samples, accepted to SIGGRAPH 2026) directly targeted the core unsolved fit-prediction problem. Independent critical assessment documented that visual realism alone does not drive purchase confidence despite vendor claims of 40% conversion lifts, and that a decade of VTO investment has not resolved the retail returns crisis — underscoring that mainstream apparel adoption depends on fit accuracy and consumer behaviour shifts, not technology maturation alone.