Making a Brand Discoverable in the Age of LLMs - Adobe Hackathon 2026 Case Study Round

Problem Preview

"# Case Study: Making a Brand Discoverable in the Age of LLMs ## Live Proctored Assessment — Brand Content & Strategy Manager ### Format This is a live proctored exercise, completed in a single supervised session. Work independently — no outside resources, collaborators, or AI assistants (ChatGPT, Gemini, Claude, Copilot, etc.). We are assessing how you reason. ### Response Bullet points are welcome — we're grading your reasoning, not your prose, so don't worry about polished English or framing full sentences. Aim for a focused answer across Q1–Q4 combined (roughly 400–800 words total, or equivalent bullets); a sharp short answer beats a long rambling one. --- ## 1. Problem Statement You have joined a consumer brand as its Brand Content & Strategy Manager. For two decades, “helping people find us” meant Search Engine Optimization. That world is shifting. People increasingly ask an assistant (ChatGPT, Gemini, Claude, Perplexity, Copilot) instead of searching. The assistant replies in one paragraph, names a few brands or none, and the user rarely clicks a link. If your brand isn’t named in that paragraph, for that user it doesn’t exist. And even when a user does click through, many bounce off the site within seconds — unable to quickly find what they came for, they leave without ever really engaging. This is now a technical problem wearing a marketing costume — which is why we're handing it to engineers. --- ## 2. The symptoms we're hitting ### Invisible in AI answers Ask an AI app such as ChatGPT or Gemini about our category and we're not named at all — it lists competitors and omits us entirely. ### Stale facts Where we do appear, the details are sometimes two years old (old logo, discontinued product, wrong price). The fresh facts don't seem to be what the assistant finds and repeats. ### Visitors don’t stay engaged Even when users do land on our site, they bounce quickly. They arrive without the context an assistant’s summary gave them, feel lost navigating the site, and can't easily find the specific information they were looking for — so they leave. The site also treats every visitor identically: it doesn’t retain or adapt to the context and intent a person arrived with, so there's no continuity from the assistant's answer to the page and no personalization to keep them engaged. A and B share a common underlying reason: machines are now the first readers of our content, and our content was written for humans, not machines. C stems from a related but distinct gap: our site itself isn't designed to orient a visitor who arrives mid-journey, already primed with a specific question. > The Appendix at the end explains the general concepts you'll need — how discovery and AI citation actually work — with no reference to this scenario. Read it before answering. --- ## 3. The example: Garuda Footwear (fictional) Picture a large, fictional Indian shoe brand, **Garuda Footwear** — founded 1994, 400+ stores, flagship Garuda Velocity running shoes (new model: Velocity X9), plus cricket, kids, and ethnic lines. It rebranded in 2024 (new logo, tagline **“Built to Fly”**, recycled soles). ### Its failures, as leadership sees them: **(1) Invisible.** “best Indian running shoes under 6000” on AI apps such as ChatGPT and Gemini lists rivals and omits Garuda entirely — it simply isn't named. **(2) Stale facts.** Where mentioned, an assistant repeats the old 2019 logo, the dead tagline, and a discontinued shoe's price — the new facts aren't repeated widely enough across the web, so what gets found and cited is still the old version. **(3) Visitors bounce.** A user asks an assistant about the Velocity X9, clicks through to Garuda's site, and lands on the generic homepage — no trace of the specific question that brought them there. They can't quickly find the X9's page, price, or availability, feel lost, and leave within seconds. The site shows the same generic page to everyone: it doesn't recognize or retain the visitor's context, so a user primed with a specific question (or a returning one) gets no personalized continuity. --- ## 4. Scope — what we want, and don't (read before answering) We want **TECHNICAL and CREATIVE ideas:** crawlers, indexing, structured data/markup, site rendering, retrieval-friendliness, machine-readable content, entity identity, and monitoring; plus creative tactics for making facts consistent, corroborated, and quotable across the web. We do **NOT** want business/commercial ideas. No paid partnerships, sponsorships, or influencer deals; no paying an LLM vendor for placement/ads; no agency retainers, hiring, M&A, or co-branding. If your main lever is “spend money so someone else features us,” it is out of scope. Reason like an engineer who read the manual, not a sales team booking ad slots. --- # 5. Your task Look at the failures listed above. Answer the questions below to help Garuda’s leadership. Each of Q1–Q3 has two parts — (a) identify the root cause and (b) suggest a fix. Bullet points are fine. ## Q1 — Visibility ### Invisible in AI apps such as ChatGPT and Gemini **(a) Root cause.** Explain the underlying reasons why Garuda is absent from AI apps such as ChatGPT and Gemini. Reason from how these systems actually work — don't just re-narrate the symptom. **(b) Fix.** Propose concrete technical and creative steps that would make Garuda discoverable and named by these assistants. --- ## Q2 — Staleness ### Outdated / wrong facts **(a) Root cause.** Explain why the few facts that do surface (e.g. on a direct-name query) are stale — old 2019 logo, dead tagline, discontinued shoe's price. Reason about what gets found and repeated across the web, and why the fresh 2024 facts don't. **(b) Fix.** Propose concrete technical and creative steps that would make the current facts the ones assistants find, trust, and repeat. --- ## Q3 — Engagement ### Visitors bounce; no context retention / personalization **(a) Root cause.** Explain why visitors who do land on the site bounce off without engaging. Reason about what happens (or doesn't) between an assistant's answer and the on-site experience — including whether the site retains and adapts to the context and intent a visitor arrived with, or treats every visitor identically. **(b) Fix.** Propose concrete technical and creative steps that would help visitors orient quickly and find what they came for, reducing bounce — including how the site could retain and use a visitor's context (e.g. the question or intent they arrived with, or prior visits) to personalize the experience and keep them engaged. --- ## Q4 — Spot a symptom we didn't The three symptoms above (**Invisible / Stale Facts / Visitors Bounce**) are not exhaustive. Identify another distinct way Garuda fails in an assistant-first world — one that isn't a restatement of visibility, staleness, or bounce — name it in one line, explain its root cause from how these systems work, propose a concrete technical/creative fix, and note how you'd measure it. One symptom, done well, beats several shallow ones. Same scope rules apply (no business/ad levers). For each Q1–Q3, briefly note how Garuda would know the fix is working — how you'd monitor visibility, factual accuracy, and on-site engagement across multiple assistants and visits over time. A focused answer with a few excellent, well-targeted ideas beats a shallow list of fifteen. State any assumptions in one line. No code or tools required. --- # 6. Rubric | Criterion | Looking for | | --------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ | | **Root-cause understanding (part a of Q1–Q3)** | Correctly explains why the symptoms happen; reasons about retrieval, crawlers, page structure, and corroboration | | **Technical measures (part b of Q1–Q3)** | Specific, plausible, correctly targeted; names real mechanisms and the causal link to the root cause | | **Creative problem identification & solution (Q4)** | Identifies a genuinely distinct, in-scope symptom beyond Invisible / Stale / Bounce, with a sound root cause and a non-obvious, correctly targeted fix | | **Measurement & rigor** | A sane way to test/monitor visibility & accuracy across assistants over time | | **Use of example & clarity** | Grounded in Garuda; clear enough for a non-technical leader | | **Scope discipline** | Stays technical/creative; avoids business/ad-buying levers | We care less about whether you know the buzzwords, and more about how clearly you reason from how these systems work to what a brand can concretely do about it. --- # Appendix — Background concepts This appendix explains the general ideas behind the exercise, with no reference to any specific company. It describes how these systems behave — not what to do about any particular problem. Working out the fixes is your job. A glossary of key terms appears at the end of this document. --- ## A. How search visibility works (the basics) Search engines, and increasingly AI assistants, discover content the same basic way: an automated program (a “crawler”) visits a page, reads what's there, and stores it so it can be found later. For a page to be visible at all, three things have to su"