E-Commerce
Last reviewed
May 9, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v5 ยท 3,659 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 9, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v5 ยท 3,659 words
Add missing citations, update stale details, or suggest a clearer explanation.
E-commerce, short for electronic commerce, refers to the buying and selling of goods and services over the internet. The arrival of large language models, computer vision systems, and recommendation engines has reshaped nearly every layer of the online shopping stack, from how shoppers search for products to how merchants set prices, write product descriptions, and resolve customer service tickets. By 2026, conversational shopping assistants, agentic checkout protocols, and generative product imagery have moved from pilot programs into mainstream deployment at retailers including Amazon, Walmart, Shopify, and Etsy.
This article surveys the AI tools, platforms, and protocols used across the e-commerce industry, including shopping assistants, personalization engines, search and discovery systems, dynamic pricing platforms, customer service automation, virtual try-on, generative product imagery, returns management, marketing copy generation, and the broader concerns around hallucinated product details and AI-generated fake reviews.
See also: E-Commerce ChatGPT Plugins
| Plugin | Image | Model | Release Date | Description | Available | Working |
|---|---|---|---|---|---|---|
| BuyWisely (ChatGPT Plugin) |  | GPT-4 | May 20, 2023 | Compare Prices & Discover the Latest Offers from thousands of online shops in Australia. | Yes | Yes |
Conversational shopping assistants combine large language models with retailer catalogs, review data, and pricing feeds to help buyers research products, compare options, and complete purchases inside a chat interface. The category grew rapidly in 2024 and 2025 as platform owners launched first-party assistants and frontier model providers added shopping features to their consumer apps.
OpenAI added shopping features to ChatGPT in stages during 2025. In April 2025 the company introduced product recommendations inside chat for queries like "best running shoes for flat feet," with image carousels and links out to retailers. On September 29, 2025, OpenAI and Stripe launched Instant Checkout for U.S. ChatGPT users, allowing direct purchases from Etsy merchants inside the chat interface, with over a million Shopify merchants including Glossier, SKIMS, Spanx, and Vuori announced as coming soon. Instant Checkout is built on the Agentic Commerce Protocol and is available to ChatGPT Free, Plus, and Pro tiers in the United States. In November 2025 OpenAI rolled out a separate Shopping Research mode, powered by a GPT-5 mini variant, which asks clarifying questions and produces a tailored buyer's guide; OpenAI noted that Shopping Research can still make mistakes and is not connected to Instant Checkout.
Perplexity launched a free AI shopping experience for U.S. users in November 2025, with conversational search, contextual memory of past shopping conversations, and a checkout flow built on a partnership with PayPal. Under the partnership, retailers remain the merchant of record, handle returns, and keep the customer relationship. The shopping feature initially launched on desktop and the web for U.S. users with iOS and Android arriving shortly after.
Rufus is Amazon's generative AI shopping assistant, launched in beta in February 2024 and rolled out to all U.S. Amazon app users later that year. Rufus runs on Amazon Bedrock and answers shopping questions ranging from broad research ("what to consider when buying running shoes?") to comparisons ("what are the differences between trail and road running shoes?") and specific product attributes. Throughout 2025, Amazon added features including price history tracking with 30, 90, and 365 day windows; account memory for individual shopping activity; and Scheduled Actions for recurring tasks such as monthly snack restocks. In November 2025, Fortune reported that Amazon expects Rufus to drive about $10 billion in incremental sales, with active users up 115% year over year.
At Google I/O 2025 in May, Google announced a new AI Mode for Search that combines Gemini reasoning with the Shopping Graph, which indexes more than 50 billion product listings. Features include a personalized price tracker with agentic checkout ("Buy for Me"), and a virtual try-on tool that lets shoppers upload a full-length photo to preview shirts, pants, skirts, and dresses. In December 2025, Google updated the try-on feature to work with a single selfie. According to Google, virtual try-on listings receive 60% more high-quality views than standard product listings.
Klarna launched a customer service assistant built on OpenAI models in February 2024. In its first month the assistant handled 2.3 million conversations, equivalent to the workload of roughly 700 full-time agents, while matching human agents on customer satisfaction scores and reducing average resolution time from 11 minutes to under 2 minutes. The assistant operated in 23 markets and over 35 languages and was projected to add about $40 million in profit improvement to Klarna in 2024. In 2025 Klarna walked back its AI-only stance and reintroduced human agents for complex cases, framing the new approach as "AI gives us speed, talent gives us empathy," though the assistant still handles roughly two-thirds of customer inquiries.
Walmart launched Sparky, a generative AI shopping assistant, in June 2025. Sparky helps shoppers find items, synthesize reviews, and make recommendations conversationally, and Walmart began testing ad placements inside Sparky responses in 2026. Etsy released a ChatGPT app in 2026 and tested its own conversational gift-search agent, joining the wave of marketplace-specific assistants alongside Rufus and Sparky.
Personalization platforms tailor product recommendations, content blocks, search results, and promotions to individual shoppers using machine learning on behavioral, contextual, and catalog data.
| Platform | Founded | Key Capabilities | Notable Customers |
|---|---|---|---|
| Algolia | 2012 | NeuralSearch, AI Personalization, A/B testing, recommendations | Lacoste, Decathlon, Under Armour |
| Bloomreach | 2009 | Loomi AI for search, merchandising, content, marketing automation | Albertsons, Puma, Bosch |
| Constructor | 2015 | NLP-driven product discovery, KPI-based ranking, merchandising | Sephora, Backcountry, Petco |
| Dynamic Yield | 2011 | Recommendations, personalized content, A/B testing | Acquired by Mastercard from McDonald's in 2022 |
| Klevu | 2013 | Self-learning search, personalization, merchandising for SMB and mid-market | Paul Smith, Uniform Wares |
Dynamic Yield has had a particularly winding ownership history. McDonald's bought the platform in 2019 to power drive-thru and kiosk personalization, then sold it to Mastercard in 2022 as part of Mastercard's Data and Services organization. The platform serves over 400 brands across retail, restaurants, and financial services.
Algolia's NeuralSearch combines vector search with traditional keyword search to handle natural-language queries and typos, while its AI Personalization layer reranks results based on individual affinity. Bloomreach's Loomi extends across the company's Discovery (search and merchandising), Engagement (marketing automation), and Content products. Constructor differentiates with KPI-based optimization, where merchandisers can choose to maximize revenue, conversion, or other metrics rather than relevance alone. Klevu targets Shopify, BigCommerce, and Magento merchants in the SMB to mid-market range.
Search and discovery in e-commerce has shifted from keyword matching toward semantic understanding, vector embeddings, and intent prediction. The leading platforms blend structured catalog data with unstructured signals such as reviews, images, and clickstream data.
Algolia operates a hosted search API that powers more than 17,000 customers across web and mobile apps. Its NeuralSearch product, released in 2023, uses neural hashing to make vector search affordable at scale, and supports hybrid lexical and semantic ranking, federated search, and dynamic re-ranking based on user signals.
Coveo is a Quebec City based search and recommendation vendor that went public on the Toronto Stock Exchange in 2021. Its commerce platform offers AI-powered relevance ranking, personalized recommendations, and a unified index across product, content, and community sources. Coveo emphasizes enterprise B2B and B2C deployments and has integrations with Salesforce Commerce Cloud, SAP Commerce, and Adobe Commerce.
Bloomreach Discovery is the search and merchandising arm of Bloomreach's commerce experience platform. It uses Loomi AI to learn from session-level behavior and historical data and presents merchandisers with both autonomous and rules-based controls. Bloomreach Discovery is delivered on AWS Marketplace and through native apps for Shopify and Salesforce.
Lily AI takes a different angle from generic search vendors by enriching product attributes themselves. The platform analyzes catalog images and copy and adds shopper-centric attributes ("flowy," "date night," "breathable") that customers actually use in queries. Lily AI announced a partnership with Bloomreach in which Lily-enriched product data feeds directly into Bloomreach Discovery search and recommendation models.
Dynamic and AI-driven pricing tools monitor competitor prices, demand signals, and inventory levels to recommend or automatically adjust list prices, promotional discounts, and markdowns.
| Platform | Approach | Target Segment |
|---|---|---|
| Pricefx | Cloud-native price optimization with deep learning models and an LLM chat interface for natural-language pricing queries | Mid-market and enterprise B2B and B2C |
| Competera | Competitor monitoring plus deep learning recommendations across thousands of SKUs | Mid-market and enterprise retailers |
| Prisync | Competitor price tracking and rule-based dynamic repricing | Small to mid-market e-commerce, especially Shopify |
Pricefx, founded in Germany in 2011, offers price optimization, CPQ (configure, price, quote), and rebate management modules. Competera markets case studies in fashion, electronics, and cruise lines and emphasizes deep learning models trained on customer demand elasticity. Prisync starts at $99 per month and is the most common entry point for small Shopify merchants who want competitor monitoring without a full ML pricing platform.
AI customer service platforms automate ticket triage, knowledge base answers, returns, refunds, and order status questions, reducing response time and freeing human agents for complex cases.
| Platform | Focus | AI Capabilities |
|---|---|---|
| Zendesk AI | Enterprise help desk | Pretrained intent models, agent copilot, generative replies |
| Gorgias | Shopify-first help desk | Auto-resolution, order actions, AI macros |
| Ada | Multilingual enterprise automation | Reasoning agent, support across 100+ languages |
| Intercom Fin | SaaS and B2C messaging | Fin AI Agent, knowledge-grounded answers, tone control |
| Tidio Lyro | SMB chat | Lyro AI, FAQ automation, visual flow builder |
| Klarna assistant | BNPL and shopping | OpenAI-powered global support across 35+ languages |
Gorgias is purpose-built for Shopify merchants and integrates tightly with order management so the AI can issue refunds or edit orders without handing off to a human agent. Intercom's Fin agent answers questions grounded in a merchant's knowledge base and product documentation, with conversation analytics and team routing. Tidio's Lyro is positioned at the SMB end of the market and bundles a visual flow builder with the AI agent.
Virtual try-on combines computer vision, generative imagery, and 3D body modeling to show shoppers how an item will look on them before purchase.
Walmart acquired the Tel Aviv based virtual try-on company Zeekit in May 2021. Zeekit's underlying technology uses real-time image processing, computer vision, and deep learning to simulate how garments will fit and drape on a shopper's body. Walmart launched the Zeekit-powered virtual fitting room in March 2022, then expanded it with the Be Your Own Model experience in September 2022, in which shoppers upload a photo of themselves rather than picking from a model gallery.
Google's virtual try-on, available across Google Search, Shopping, and Images, supports apparel from retailers including Macy's, Kohl's, Walmart, and Nordstrom. The selfie-only flow released in December 2025 made the feature accessible without a full-body photo.
Vue.ai's Dressing Room uses generative adversarial networks (GANs) to render product images on diverse model body types, sizes, and skin tones. Customers including Lane Crawford, Picard, and Showpo deployed Dressing Room and reported a 1.5x lift in add-to-cart and 2x lift in time on site for shoppers who engaged with the experience.
Other active vendors in the visual try-on category include Doddle, which provides AR-based try-on widgets for fashion and beauty merchants, and a long tail of glasses and cosmetics try-on tools embedded by brands such as Warby Parker, Sephora, and L'Oreal.
Generative product photography tools replace, augment, or simulate traditional studio shoots by placing product images into AI-generated backgrounds, scenes, and lifestyle settings.
| Tool | Specialty | Pricing Notes |
|---|---|---|
| Pebblely | Single-product lifestyle scenes from one image plus theme prompts | Free tier 40 images per month, Basic $15 for 1,000 images at 2,048 by 2,048 px |
| Booth.ai | Studio-quality product photography from a single product image | Subscription tiers for solo creators through agencies |
| Flair.ai | Drag-and-drop scene builder with props, lighting, and virtual cameras | Subscription, with collaborative team workspaces |
| Photoroom | End-to-end e-commerce workflow including background removal, batch editing, and Instant Background | Free tier plus Pro and Business plans |
Photoroom's Instant Background preset can produce 20 production-ready images in under 3 minutes in published benchmarks. Flair.ai sits at the creative-control end of the market: the canvas knows the product's perspective and scale, and the user composes the scene rather than relying on a one-shot prompt. Pebblely is popular with solo founders and boutique brands who handle their own marketing.
Returns are a structural cost in online retail, with U.S. return rates routinely above 15% in apparel and electronics. AI-driven returns platforms aim to reduce that cost by approving returns intelligently, routing returned inventory to the highest-value resale channel, and capturing data to prevent future returns.
Returnly, founded in 2014, was acquired by Affirm in 2021 for around $300 million and integrated into Affirm's post-purchase suite. Returnly used machine learning to make real-time return policy decisions, for example offering instant store credit for low-risk shoppers.
Optoro operates a Returns Management System and the SmartDisposition platform, which uses machine learning to route returned items to the most profitable next channel, including resale, liquidation, donation, or destruction. Optoro customers include Gap Inc., American Eagle Outfitters, and Best Buy. The platform combines NLP on customer return comments with computer vision for damage assessment and reinforcement learning for inventory routing.
Generative AI copy tools help merchants and agencies produce product descriptions, ad copy, email campaigns, blog posts, and on-site landing pages at scale.
| Tool | Strength | Notes |
|---|---|---|
| Jasper | Long-form marketing content, brand voice, agency workflows | Templates for blog posts, email sequences, ads |
| Copy.ai | Short-form copy, social posts, product descriptions | GTM AI platform for sales and marketing teams |
| Anyword | Predictive performance scoring for ad copy | Models predict click-through and conversion before publishing |
| Mutiny | Website personalization for B2B and DTC | AI-driven page variants by audience segment |
Anyword is the most measurement-oriented of the group, scoring each variant against historical campaign data before it ships. Mutiny is less a copy tool than a personalization layer that swaps headlines, hero copy, and calls to action based on audience segment, firmographic data, or referral source.
The major marketplaces and commerce platforms have built first-party AI products that touch every part of the merchant workflow.
Amazon Bedrock is the foundation model platform that powers Rufus and many third-party retail assistants. Amazon offers Bedrock-based reference architectures for product description generation, e-commerce recommendation chatbots, and seller assistance. Project Amelia, launched on Bedrock in 2024, is a seller-facing assistant that answers questions about inventory, forecasting, and account issues. In 2025 AWS introduced Bedrock AgentCore Payments, built with Coinbase and Stripe, to let agents transact on behalf of buyers.
Shopify Magic is the merchant-facing brand for Shopify's AI features, available free across all subscription tiers. Sidekick is the conversational agent that sits inside the Shopify admin and was launched broadly with the Winter Edition 2025 release in December 2024. Sidekick supports text, voice, and screen-sharing input. It can analyze sales data, edit products, generate marketing copy and images, set up discounts, and run multi-step tasks in the background. The May 2025 Sidekick update added multi-step reasoning, advanced analytics, and integrated image generation.
Etsy has rolled out a series of AI features for buyers and sellers, including image-based search, AI-driven gift recommendations, and a tested conversational gift-search agent. In 2026 Etsy launched a ChatGPT app and was among the first U.S. marketplaces to plug into Instant Checkout via the Agentic Commerce Protocol.
The Agentic Commerce Protocol (ACP) is an open standard for agent-initiated checkout, codeveloped by OpenAI and Stripe and released under the Apache 2.0 license on September 29, 2025. ACP defines how AI agents discover products, present them to a buyer, and complete a payment without exposing the buyer's underlying payment credentials to the agent.
A key primitive in the protocol is the Shared Payment Token (SPT), issued by Stripe, which lets a host application like ChatGPT initiate a charge on behalf of the buyer through the merchant's existing payment infrastructure. Merchants retain control over what is sold, how their brand is presented, and how orders are fulfilled, and they remain the merchant of record for tax, fraud, and chargeback purposes.
Launch partners and supporters announced alongside ACP include Salesforce Commerce Cloud, commercetools, Etsy, and Shopify. The protocol is intended to be payment-provider agnostic, and the GitHub specification at agentic-commerce-protocol/agentic-commerce-protocol invites third-party payment networks to implement compatible token formats. Adoption beyond ChatGPT picked up through 2026, with Perplexity's PayPal-based checkout taking a different architectural path while pursuing similar goals.
Generative shopping assistants can fabricate product specifications, prices, availability, or even entire products that do not exist in a catalog. OpenAI explicitly warned that ChatGPT's Shopping Research can still make mistakes, and similar caveats apply to Rufus, Gemini, and Perplexity. For sellers, hallucinations can lead to disputed orders or trademark conflicts; for shoppers, they can produce overconfident buying advice that is wrong on basic facts. Retrieval grounding against verified merchant catalogs and structured product feeds is the standard mitigation, but coverage is uneven across long-tail SKUs.
Research by Pangram Labs found that approximately 3% of 30,000 front-page Amazon reviews analyzed were AI-generated, with nearly three-quarters of those rated five stars and 93% carrying the "verified purchase" label. App store data tracked by DoubleVerify showed apps with AI-powered fake reviews tripled in 2024 compared with 2023. The U.S. Federal Trade Commission's final rule banning fake reviews took effect in October 2024 and explicitly covers AI-written reviews, and the FTC took action against AI writing tool Rytr for offering services that could be used to flood marketplaces with fake reviews. Amazon and Trustpilot permit AI-assisted reviews if they reflect a genuine experience, while Yelp requires reviewers to write their own text. Studies indicate that human readers struggle to distinguish AI-generated reviews from human-written ones, raising concerns about consumer trust in review systems generally.
Generative imagery makes it cheap to fabricate convincing photos of products that do not match reality, including counterfeit goods presented in legitimate-looking lifestyle scenes. The same generative tools that legitimate merchants use for product photography can be used by bad actors to mislead buyers, raising ongoing questions about disclosure and watermarking standards.