How does Myntra use generative AI? From MyFashionGPT to AI sizing and try-on
India's Myntra launched a ChatGPT-based shopping assistant in 2023 and has since added AI sizing, try-on and seller tools. What it has published, what is still piloting, and the lessons.
KEY TAKEAWAYS Summary by the editors
- Myntra launched MyFashionGPT, a natural-language shopping assistant built on Microsoft's Azure OpenAI Service, in its app in late May 2023, timed with its End of Reason Sale.
- Microsoft reported that Myntra shoppers who used the assistant were three times more likely to make a purchase and added products from 16% more categories on average.
- In July 2026 Myntra said its size recommendation engine covers 85% of its eligible apparel catalogue, responds in under two seconds and explains why a size was suggested.
- Myntra reports that AI extracts up to 40 attributes from a single seller product image, with catalogue experts reviewing the output, and that seller go-live time fell from 10 to 15 days to about one to two days.
- Features such as the fit visualiser and 'Try it on Me' were still being scaled or rolled out in July 2026, and most performance figures are company-reported.
Myntra, India's largest online fashion retailer, uses generative AI in three places: conversational search and styling for shoppers, size and try-on tools to improve fit confidence, and catalogue and support automation for sellers and operations. It was an early mover, launching the ChatGPT-based MyFashionGPT assistant in May 2023, and by mid-2026 it described a broader programme covering sizing, try-on, seller onboarding and customer service.
What is MyFashionGPT and how does it work?
MyFashionGPT lets shoppers search in natural language, for example asking for outfits for a festival, a cricket match or a trip, instead of typing catalogue keywords. According to Microsoft, Myntra built it on Azure OpenAI Service after a February 2023 hackathon produced a working prototype in two days; teams then spent about five weeks refining it before launch in the Myntra app in late May 2023, timed with the End of Reason Sale.
The design matters. DailyO's review at launch explained that the request goes to a ChatGPT model and Myntra's own search engine then uses the response to show products. Microsoft reported that the team trained the system on Myntra's catalogue and added guardrails so results stay within it, and that Hindi queries such as sardiyon ke kapde (winter clothes) were understood. Gender filtering was a problem the team had to work on.
Microsoft reported, citing Myntra's product team, that shoppers who used the assistant were three times more likely to make a purchase and added products from 16% more categories on average. These are company-reported figures without published methodology, and self-selection (engaged shoppers are more likely to try new features) may explain part of the difference.
Which other AI features does Myntra use for shoppers?
Myntra had used machine learning well before generative AI. iTnews Asia reported in 2023 that its My Stylist feature personalised the home screen and recommended outfits, sizes and designs using deep convolutional neural networks on catalogue data, alongside image search, fashion object detection and outfit recommendations. In 2023 Myntra also introduced Maya, first presented as a virtual fashion influencer on social media; Microsoft and YourStory now describe Maya as a conversational shopping assistant.
At its mynnovAIte showcase in Bengaluru on 16 July 2026, Myntra set out the current shopper-facing features, as reported by YourStory:
| Feature | What it does | Status reported |
|---|---|---|
| Size recommendation engine | Suggests a size with an explanation; covers 85% of eligible apparel; responds in under 2 seconds | Live |
| Personalised search | Personalised results for over 90% of monthly active users | Live |
| Maya assistant | Conversational shopping help; 25% repeat interaction rate among engaged users | Live |
| Fit visualiser | Shows how garments may drape on a similar body type | Being scaled |
| Try it on Me | Shows an outfit on the shopper's own photo | Rolling out |
| Styling layer | Mix and match outfits on one model | Reported, scope not detailed |
The explainability layer on size recommendations is worth noting. Telling a shopper why a size was suggested (for example, based on past purchases or brand sizing) gives them a reason to trust or override it, which matters in a category where fit drives returns. Myntra has not published return-rate effects.
How does Myntra use AI for sellers and operations?
Myntra is a marketplace, so catalogue quality depends on thousands of sellers. YourStory reported that a seller can now upload one product image and AI extracts up to 40 attributes such as colour, fabric, neckline and sleeve style, which catalogue experts then review. Myntra says seller go-live time fell from 10 to 15 days to about one to two days, and that AI image generation cut a roughly one-day task to four hours.
- Product video: 400 to 600 AI-generated product videos per day.
- Seller guidance: a redesigned partner portal shows ranked daily actions, and a voice platform called Saarthi calls sellers about operational issues or pricing gaps.
- Customer support: an assistant called Meera resolves 31% of queries, according to Myntra.
- Internal analytics: a natural-language analytics platform, BIRA, which Myntra says cuts analysis from days to minutes.
The pattern is human review on top of automation. Catalogue experts check extracted attributes, which is a sensible control because attribute errors propagate into search, filters and size advice.
What are the limits of Myntra's approach?
Early reviews were frank. DailyO noted at launch that MyFashionGPT returned similar suggestions for a 21-year-old and a 50-year-old shopper, did not handle elaborate descriptions or luxury wear well, and was limited by the catalogue it searched. Those are typical limits of grounding a general model in a retailer's own assortment.
Several flagship features (the fit visualiser, 'Try it on Me' and a seller copilot) were still being scaled, rolled out or developed in July 2026. Virtual try-on on a shopper's own photo also raises consent and image-handling questions that any retailer must address before scaling.
What data and infrastructure sit behind Myntra's AI?
Myntra's AI features rest on a data platform built for machine learning at high volume. At the Aerospike Summit 2023, senior architect Narayana Pattipati described, as reported by iTnews Asia, a platform that consolidates diverse data streams into one database for the entire machine learning lifecycle, using a hybrid memory architecture that combines flash storage and RAM. It serves online feature stores, state stores and inference caches, and helps with problems such as ranking the widgets on the app's home screen.
Myntra also said it planned to run most new features through A/B testing. That discipline matters more than any single model: without controlled tests, a retailer cannot separate the effect of a new AI feature from seasonality, promotions or the fact that engaged shoppers try new tools first.
What can other fashion e-commerce players learn from Myntra?
- Ground generative AI in your own catalogue and search engine rather than letting a model answer freely.
- Support local languages early; Hindi queries were part of Myntra's design from the start.
- Invest in seller-side AI on marketplaces, because catalogue quality limits every shopper feature.
- Keep humans reviewing AI-extracted attributes before they go live.
- Make AI recommendations explainable, especially for size and fit.
Frequently asked questions
What is MyFashionGPT?
MyFashionGPT is Myntra's natural-language shopping assistant, launched in the Myntra app in May 2023. It sends a shopper's request to a ChatGPT model via Azure OpenAI and then uses Myntra's own search engine to show matching products from the catalogue.
Who is Maya on Myntra?
Myntra introduced Maya in 2023 as a virtual fashion influencer on social media. Microsoft and YourStory now describe Maya as Myntra's conversational shopping assistant, which Myntra says has a 25% repeat interaction rate among engaged users.
Does Myntra have virtual try-on?
Myntra reported in July 2026 that 'Try it on Me', which shows an outfit on the shopper's own photo, was rolling out, and that a fit visualiser was being scaled. Neither had published performance data at that point.
How accurate is Myntra's AI size recommendation?
Myntra has not published accuracy figures. It says the engine covers 85% of its eligible apparel catalogue, responds in under two seconds and includes an explanation of why a size was suggested.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
SOURCES
- Microsoft Source Asia: India's Myntra innovates with generative AI to help shoppers put the right look together
- YourStory: Inside Myntra's AI playbook: Faster seller onboarding, smarter shopping, leaner operations
- DailyO: Myntra is making ChatGPT your personal stylist in sleek new AI update
- iTnews Asia: Myntra leverages AI/ML to level up customer experience
- Microsoft: AI First Movers: Myntra
