How do you guide shoppers who don't know what to search for?

A shopper looking for a gift for a runner, or a rug for dark floors, has real purchase intent. They can't always name what they want in catalog terms. A blank search box with no starting point signals they're on their own. Most leave without typing. That exit never registers in your analytics. Getting them engaged requires a starting point that surfaces questions they can tap without forming a query themselves. This guide covers how to add those prompts, choose which questions to surface, and measure whether they're working.

Why does a blank search box cause undecided shoppers to bounce?

Undecided shoppers land with purchase intent but no ready search term - they want a gift for a runner, a moisturizer for sensitive skin, or a rug that works with dark floors, but they can't translate that into a catalog-specific query. A search box with no starting point signals "you're on your own," and the friction of formulating a query - combined with the fear of hitting a dead end - tips them toward leaving. Because these shoppers never submit a query, their exit doesn't register as a null result or zero-result rate.

Vocabulary mismatch compounds it for shoppers who do try. A visitor searching "vegan leather bag" hits a dead end if your catalog tags the same item "faux leather" - a motivated query still returns nothing.

High-intent paid and email traffic is especially exposed. The CAC you spent to get that visitor evaporates if the on-site experience offers no direction on arrival. Search abandonment - leaving without submitting a query - is distinct from a null result, and most analytics platforms track only the latter.

Improving what shoppers see when the search box is empty converts visitors who are already on site.

What is Nobi, and how does it turn a blank search box into a guided starting point?

That better starting point is Nobi - an AI-powered site search and shopping assistant platform for ecommerce brands, combining site search, conversational Q&A, and automatic shopper question-answering in one place. Its mechanism for undecided shoppers is suggestion pills: LLM-generated, tappable buttons that appear before a visitor types anything.

The pills adapt to where a shopper is on your site. A homepage pill might read "What's a good gift under $100?" A product detail page surfaces "Does this run true to size?" or "What materials is this made from?"

Tapping a pill opens the Nobi assistant with that question pre-filled and a grounded answer already loading, pulled from the product pages, FAQ routes, policy docs, and PDFs you've connected.

Every answer includes inline numbered citation pills. Hover one and you see the source document, date, and the exact excerpt the answer drew from.

The knowledge base builds from your existing site content - no manual re-entry. Connect a URL or upload a file and it becomes part of what the assistant can answer. Connected sources refresh twice a day, so a policy update or inventory change lands in customer answers within hours.

Pricing starts at $25/month (2,500 searches and 250 conversational messages included). Beyond that: $0.01 per additional search, $0.10 per additional message.

Lucchese, the luxury Western boot brand, saw $1M+ in incremental revenue in their first year running Nobi for search, cart, and PDP discovery - a 39x ROI.

How do I add Nobi's suggestion pills to my store?

Getting those pills live takes three steps: connect your knowledge sources, add the Nobi widget to your storefront with a small theme tweak, and let the assistant generate page-scoped pills from your content automatically. Most Shopify stores are live the same day, and no manual prompt-writing is needed to launch.

Step 1: Connect your knowledge sources. Point Nobi at the content you already have - product pages, FAQ routes, policy docs, or PDFs you upload directly. Indexing is automatic; sources stay current without manual re-syncing.

Step 2: Install the widget. A small theme tweak gets it live. Developer-level customization is optional - most stores don't need it to get started.

Step 3: Pills appear automatically. Once the knowledge base is active, Nobi generates pills for each page context out of the box - homepage, PDP, collection. Nobi scopes them to where the visitor is on your site without any per-page configuration from you.

The component has a few controls worth knowing. `pill-count` sets how many pills appear (1-10, default 3). `multi-row` switches from a carousel to a grid. `pill-types` locks the category to `cross-sell`, `qa`, or a custom slug - or leave it unset and the AI classifies each one:

```html <!-- Default: 3 carousel pills -->

```

```html <!-- Multi-row example with up to 6 pills -->

```

```html <!-- Specify pill types (comma-separated) -->

```

```html <!-- Explicit product ID (optional - auto-detected on most platforms) -->

```

Pill colors and spacing match your brand through CSS variables:

```css :root { --nobi-suggestion-pill-background: #fff; --nobi-suggestion-pill-text-color: #212121; --nobi-suggestion-pill-border: 1px solid #e0e0e0; --nobi-suggestion-pill-hover-background: #212121; --nobi-suggestion-pill-hover-text-color: #fff; --nobi-suggestion-pill-icon-color: #ff6b35; --nobi-suggestion-pills-gap: 12px; }

.product-media.nobi-suggestion-pills-container { margin-top: 1rem; justify-content: flex-start; } ```

For high-stakes prompts - return windows, warranty terms, shipping cutoffs - query overrides let you pin exact merchant-approved answers to specific questions. Those prompts always fire the answer you set, with no variation.

If your storefront design needs a custom layout for how products render inside the assistant, the Hooks API gives your dev team that control. UNTUCKit originally asked for it; that request became a standard feature available to every Nobi merchant.

How do I choose prompts that match my catalog and shoppers' intent?

Query overrides and pill types give you the configuration layer - the harder call is knowing which questions to surface. Start with your top support ticket categories and your highest-intent search queries: those are the questions shoppers are already trying to ask, which makes them natural candidates for tappable prompts. Frame each one as a natural-language question a real shopper would tap, not a keyword fragment. "Does this run true to size?" works. "sizing" doesn't.

Page context determines how specific each prompt should be. A homepage pill should apply broadly ("What's a good gift under $100?"). A product detail page pill should zero in on the product a shopper is looking at.

Certain prompt categories tend to work well across most storefronts:

Avoid pills that are too generic to answer well ("Show me your best products") or too SKU-specific for a page carrying a wide assortment.

For prompts on sensitive topics - allergens, exact dimensions, compatibility claims - Nobi can run a second AI review that checks each draft answer against the source content before it sends. That extra pass reduces the chance a fabricated answer erodes trust on a question where accuracy matters most.

Can I customize suggestion pills for different pages or audiences?

The prompt categories that work for your storefront don't all belong on every page. Nobi generates suggestion pills based on the specific page type a visitor is on - homepage, PDP, or collection page - so the questions shift based on where the shopper is in the funnel without any manual segmentation work on your end. Merchants can override the LLM-generated defaults and pin specific prompts to specific page types, giving marketing control over key surfaces without an engineering dependency.

Page type determines the category of question each pill asks. On the homepage, pills skew broad: gift finders, category guidance, use-case entry points that work across the full catalog. On a PDP, the AI narrows to the item in view - size guidance, material questions, compatibility checks, return and exchange policy for that specific product type. On a collection page, pills help shoppers narrow a wide result set: "What's the difference between X and Y in this collection?" or "What's the most popular option for [use case]?"

For high-stakes questions - return windows, warranty terms, shipping cutoffs - query overrides let you lock exact merchant-approved text to a specific prompt. When a shopper asks a question that matches the override, the pinned answer fires every time. The AI won't rephrase it; the response is exactly what you wrote.

Because connected knowledge sources refresh twice a day, a policy update or a new FAQ entry lands in pill responses within hours. You don't need to trigger a re-index.

How do I measure whether suggestion pills are lifting engagement and revenue?

Track pill tap rate, assisted CVR, and revenue per visit as your primary signals. Nobi surfaces search and CVR metrics in the dashboard so you can compare assisted and unassisted session performance directly.

Pill tap rate is a prompt-relevance signal, not a traffic metric. A low tap rate usually means the copy doesn't match what shoppers want to know on that page - test alternative phrasings before drawing conclusions about the feature itself. A pill that reads "What's in this?" may underperform one that reads "What are the materials?"

Assisted CVR is the clearest measure of whether the pills are doing their job. Sessions where a shopper tapped a pill should convert at a higher rate than sessions where they didn't. That gap is your lift. If the gap is narrow, the problem is usually with which questions are surfacing, not with the mechanism itself.

Revenue per visit is the most useful number to bring to a growth or marketing lead. It normalizes for session count and ties the search experience directly to the traffic budget - a cleaner comparison than CVR alone when session volumes shift.

Lucchese, the luxury Western boot brand, has accumulated $3.46M in attributed revenue running Nobi for search, cart assistant, and PDP discovery. That number comes from the same measurement frame: revenue traced to sessions where the assistant made contact.

When does a different tool fit better than Nobi for guiding undecided shoppers?

Nobi's suggestion pills work best when shoppers need conversational guidance that flows into grounded search and Q&A - that covers most discovery needs. Two cases point to a different tool: when your main product-finder experience is a structured quiz, and when marketing ops needs to build campaign-specific conversation flows without engineering support. Nobi also has real limits worth knowing before you commit.

Quiz-led discovery. Octane AI is built for quiz-led product finding - skincare routine-builders, fit finders, supplement regimens. It ships pre-built Shopify templates with deep product-attribute mapping and native Klaviyo integration. Pricing starts at $50/month for 400 quiz completions.

Scripted campaign flows. Rep AI's Flow Studio is a no-code builder for designing deterministic conversation paths - Black Friday promos, product-launch walkthroughs, abandoned-cart rescues - without engineering tickets. Pricing starts at $12 per 1,000 visitors.

Where Nobi falls short. Nobi's current personalization covers placeholder text and starter messages only - it doesn't yet rerank results based on individual shopper click and purchase history. Brands whose main need is behavioral reranking will need a platform built for that. Nobi also answers shopper questions and reduces the volume landing in your support queue, but it doesn't manage tickets, route cases to agents, or process returns in a structured workflow. For that layer, pair it with a dedicated helpdesk.

---

Head to nobi.ai and start a free trial. Suggestion pills go live on your store within hours, turning that blank search box into a guided starting point for every shopper who arrives without a clear query in mind.

<div className="my-8 flex justify-center"> <a href="https://dashboard.nobi.ai" className="inline-flex items-center justify-center gap-2 rounded-2xl font-medium transition active:scale-[.98] focus:outline-none focus-visible:ring-2 focus-visible:ring-black/10 dark:focus-visible:ring-white/20 bg-black text-white dark:bg-white dark:text-black hover:opacity-90 shadow-sm h-12 px-6 text-base no-underline" > <span>Start your free Nobi trial</span> </a> </div>