Back Market
Designing Trust for Humans and LLMs
Type
Bootcamp final project
Timeframe
12 days
Toolkit
Figma & AI agents
Year
2026
Problem
Back Market is Europe’s leading marketplace for refurbished electronics, but a new type of traffic now arrives from AI assistants instead of Google. These visitors trust AI to compare options, yet they often hesitate at the moment of purchase. AI tools already recommend Back Market as a “best option,” but the existing PDP was not optimized to reassure AI-referred users unfamiliar with refurbished products or to make the trust story structured and machine-readable so AI keeps ranking Back Market highly.
Solution
Over a two-week solo project with Back Market during my Ironhack bootcamp, I redesigned the PDP end to end: research, strategy, information architecture, UX, UI, and design system. I focused the experience on clearly addressing first-time refurbished concerns and structuring key reassurance signals for both humans and AI. This included surfacing warranty and trust content in a way that reduces hesitation for users, while also improving how LLMs can read and recommend Back Market as the smart choice
I redesigned Back Market’s Product Detail Page for AI-referred traffic, focusing on first-time refurbished buyers who arrive from tools like ChatGPT and need strong reassurance before buying.
Research
I combined secondary research on AI search behaviour with 35 surveys and 6 user interviews. The main fear was after-sales service and warranty, ahead of device reliability and battery health. Competitor PDPs barely addressed these fears, relying on similar configuration layouts without clear proof of refurbishment quality. From this, I built “Luna,” a 38-year-old, price-conscious but risk-averse buyer who wants a good deal, a real warranty, and a fast, confident decision.
Key insight: hidden warranty details, unclear grades, and no visible refurb process create comparison paralysis and last-minute doubt.
Concept
I explored wide ideas such as 3D product scans, “refurbish it yourself” flows, AI shopping agents, and quizzes. After feedback from the Head of Design a Product Design Manager and a Senior product designer, I narrowed the scope using MoSCoW.
The core experience became:
– Warranty and returns clearly visible above the fold
– Immediate proof of a professional refurb process
– Simple, focused configuration with trade-in surfaced at key moments
This led to a mental model and page structure: Prove → Price → Trust → Buy.
Design
I iterated from low to high fidelity prototypes, mobile-first and responsive. Low-fi tests with 6 users validated the Prove → Price → Trust → Buy pattern and above-the-fold warranty placement. Mid-fi prototypes introduced micro-interactions such as a trade-in overlay and refurb video auto-play. High-fi tests with 8 users surfaced two fixes: animations were too fast and some colors felt inconsistent; I slowed motion and tightened the palette.
The design system includes reusable components, an accessible color palette (WCAG 2.1 AA), a responsive type scale, and an 8px spacing grid.
AI & SEO Strategy
To keep AI agents recommending this PDP, I structured key trust elements as semantic blocks, lists, and hard numbers instead of pure marketing copy. I recommended placing the Back Market Promise at the top of the DOM, strengthening E-E-A-T signals, and implementing an llms.txt file so LLMs can parse a clean, Markdown-like version of the page.
Outcome
In user testing, participants reduced time-to-purchase from around 8 minutes to 3 minutes. They reported a clear confidence boost after seeing the refurb process video and trust blocks, and described the layout as intuitive, modern, and trustworthy. These results are from usability tests only; live A/B data would be the next step before full validation.
Next
After this project goes live, I would:
– Monitor conversion rate, cart abandonment, and AI referral performance
– Iterate on reassurance content, trade-in placement, and LLM-oriented structure based on real traffic behaviour.
Video demo
Read the full detail process on medium
