Kernvel Image SearchIQ — User Guide

Shoppers can't always put what they want into words. Image SearchIQ lets them search your store by photo instead — snap a picture, upload an image, or use their phone camera, and instantly see the matching products in your catalog. Fewer dead-end searches, better discovery, more sales.

This guide covers everything Image SearchIQ does today: how to get it live, what shoppers experience, and how to read the demand signals that come back. Every feature described here is shipping in the app — nothing aspirational.

What Image SearchIQ does

Shopify's built-in search matches words, not pictures. A shopper who saw the perfect piece on someone, or has a photo saved on their phone, but can't name the style, colour, or cut, hits "no results" and leaves — even when that exact item is sitting in your catalog. Image SearchIQ closes that gap.

Here's the whole loop in one breath: when you install, Image SearchIQ builds a private visual index of your product images. A shopper then snaps or uploads a photo on your storefront, the app turns that photo into a visual "fingerprint," compares it against your index, and shows the products that actually look like it — ranked by similarity, in real time. Every search is also a signal: you learn what shoppers hunt for, which searches come back empty (your catalog gaps), and which looks they want that you don't stock yet.

What you get out of the box:

• Search by photo — shoppers find products by picture, not keywords. Native camera capture on mobile.
• Search by words too — the same widget is a unified search box: on Starter and above shoppers can type a query (with search-as-you-type suggestions), matched against your product-image index by the same visual engine.
• Crop-to-region — a shopper drags a box over the one item in a busy photo (just the bag, just the shoe) and searches only that. Free on every plan.
• Photo + words (hybrid search) — refine a visual search with text: "this style, but in blue," "similar, but cheaper."
• AI auto-tagging & filters — Image SearchIQ reads each product photo and labels it with attributes (colour, category, material, style), which become storefront filters with no hand-tagging.
• Visual recommendations — "more like this" and "complete the look" on product pages, driven by the same visual engine.
• Visual demand intelligence — the looks shoppers searched for and couldn't find, grouped into named demand clusters you can stock toward.
• Merchandising controls — pin the products you want seen first, bury the ones you don't.
• Search analytics — volume, catalog gaps, the search-to-purchase funnel, and speed.
• Near-duplicate finder — a daily scan that flags product images that look almost identical, so you can clean up duplicate listings. On every plan.
• Privacy by design — a shopper's search photo is used to find a match and then discarded. It is never stored.

Some surfaces are tier-gated — the table below shows exactly what's on each plan. Core visual search, crop-to-region, merchandising, and analytics are on every plan, including Free.

Getting started

Setup is a short wizard plus one step only you can do (placing the widget on your theme). It takes about three minutes, and your storefront search goes live automatically once your catalog finishes indexing — you don't have to sit and wait.

1. Index your catalog

The moment you install, Image SearchIQ starts building a private visual index of your product images. Each product is fetched once, analyzed by the AI engine, and added to your index. New and updated products are re-indexed automatically as you edit your catalog — there's nothing to maintain.

You can watch progress on the Catalog index page:

• A live progress bar and percentage, with an estimated completion time.
• Counts of products, images embedded, images pending, and any that failed.
• A Rebuild index button to re-run the whole catalog (for example after a big catalog change). It's safe to click even if a build is already running — duplicates are guarded.
• A failed-images retry if any image couldn't be embedded (usually a temporary fetch blip — retrying re-embeds just the failures).
• A recent jobs log so you can see every full re-index and incremental update with its outcome.
• A near-duplicate products finder — a daily scan of your indexed images that flags product pairs that look almost identical, with the similarity score for each pair. Handy for spotting accidental duplicate listings or tightening product photography. It's on every plan.

The 80% live gate. Visual search reveals itself to shoppers automatically once your catalog is 80% indexed. Until then, the storefront widget stays hidden and shows "Visual search is being set up" instead — so a shopper never gets empty or wrong results on a half-built index. Large catalogs (tens of thousands of products) can take a couple of hours; you can keep configuring the app while it finishes.

2. Add the visual search page (optional)

From the dashboard checklist, one click creates a dedicated /pages/visual-search page on your storefront that customers can browse to and search from. (If you'd rather, create it manually under Online Store → Pages — name it "Visual search.")

3. Place the search widget

This is the one step Image SearchIQ can't do for you. In your theme editor:

1. App embeds → turn on "Image SearchIQ" to add a camera icon beside your theme's search bar. This is the fastest way to go live.
2. (Optional) On the Visual search page from step 2, use Add section → Apps → "Image SearchIQ" to place the full search experience on a dedicated page.
3. Set the widget's accent colour and placement under Settings → Widget branding.

You don't have to manually confirm anything — Image SearchIQ detects the widget as live the moment a shopper loads a page with it, and the dashboard checklist updates itself.

The onboarding wizard

First run walks you through five steps with a progress strip: Welcome → Catalog index → Configure search → Add the widget → Review & launch. The "Configure search" step lets you set your match strictness, results-per-search, and out-of-stock handling up front (all changeable later in Settings). You can launch the dashboard even before indexing finishes — search simply turns on for shoppers the moment the index crosses 80%.

The storefront visual search experience

This is what your shoppers see once the widget is live. Depending on your widget placement setting, the entry point is either a camera icon beside your search bar (header) or a floating button on your storefront pages.

Snap or upload. A shopper takes a photo, uploads an image from their device, or uses their phone camera right from your storefront. The widget resizes the image client-side before sending, so even a big phone photo is fast.

Instant matches. Image SearchIQ embeds the photo and surfaces the products from your store that look most like it, ranked by visual similarity — in real time. Storefront visual search targets sub-2-second responses.

Crop to the exact item. Photos are usually busy — a whole outfit, a full room. The shopper drags a crop box over the one item they care about and searches just that. This is included free on every plan.

Refine with words (hybrid search). On Starter and above, shoppers can add a phrase to steer a photo search — "this dress, but in blue" or "similar, but cheaper" — combining what they can show with what they can say to land on the right product faster.

Find similar from an on-page image. A "find similar" camera can appear on your product images so a shopper can instantly pull up visually similar products from your catalog.

Type a query, not just snap one. Beyond refining a photo, the widget is a full unified search box: a shopper can search with words alone. It offers search-as-you-type suggestions (matching product titles appear as they type) and a zero-state of your most-popular products to tap before they've entered anything. Typed words run through the same visual engine — matched against your product-image index — so a shopper can lead with a photo, a phrase, or both. Like photo-plus-text refinement, standalone text querying rides on the hybrid-search entitlement, so it's Starter and above; on Free the widget is photo-and-crop only.

Honest fallbacks, never a dead end. If nothing crosses your match-strictness threshold, the shopper sees a gentle "closest matches" row rather than a blank "no results" page — and that miss is quietly logged as a catalog-gap signal for you.

Visual recommendations on product pages

Available on Growth and above.

The same image engine that powers visual search also drives your product-page recommendations — so suggestions are based on how products actually look (pattern, colour, shape), not just shared tags or co-purchase history.

Two storefront surfaces, both placed via theme blocks on your product pages:

• More like this — visually similar products to the one a shopper is viewing.
• Complete the look — complementary picks from other categories that go with the item, nudging shoppers toward bigger carts and higher average order value.

There are no rules to write and no manual merchandising — the engine surfaces the right picks automatically and improves as your catalog grows. Recommendations run off each product's already-indexed visual fingerprint, so they're fast and add no extra cost.

Preview before you ship. The admin Recommendations page lets you pick any product and see exactly what shoppers would get — both the "more like this" and "complete the look" rows — so you can sanity-check the picks against your live catalog before turning the storefront blocks on.

Auto-attribute tagging & filters

Auto-tagging available on Starter and above; storefront auto-filters on Growth and above.

Image SearchIQ reads each product photo and labels it automatically with rich attributes — category, colour, pattern, material, style, occasion, season — without you tagging a single product by hand. Those attributes do double duty:

• Storefront filters. On Growth and above, the discovered attributes become clean filters shoppers can narrow by (colour, style, material, and more).
• Smarter search. The same tags power hybrid "this, but in blue" searches and the complete-the-look picks.

On the admin Auto-filters page you can see every attribute the tagger discovered across your catalog — each facet, its values, and how many products carry each one (for example, "color → green across 42 products"). A Re-tag catalog button re-runs tagging across the store, which is handy after you add a lot of products or change your catalog significantly.

If you're on Starter, your products are being auto-tagged and you can re-tag any time — upgrading to Growth turns those tags into the storefront filters and unlocks the full facet breakdown.

Merchandising controls (pin & bury)

On every plan.

Visual search ranks results by how closely products match what a shopper photographed. Merchandising is your thumb on the scale on top of that automatic engine — override the ranking only where it matters and let the engine handle the rest.

• Pin — the product surfaces near the top of any visually relevant search, ahead of the automatic ranking. Use it for hero products, new drops, or high-margin lines.
• Bury — the product is pushed to the bottom of results (or hidden) even when it's a strong visual match. Use it for clearance, discontinued lines, or items you don't want promoted.

Everything you don't touch is ranked automatically by visual similarity — there's no manual upkeep. Pinned and buried products are collected on the Merchandising page so you can manage them in one place, with counts of how many are pinned, buried, and merchandised. (Out-of-stock items are handled separately by your out-of-stock setting — see Settings below.)

Visual demand intelligence

Available on Scale and above.

Every time a shopper snaps a photo and your catalog has no good match, that's a buying signal most stores never capture. Image SearchIQ clusters those unmet searches into named demand patterns — "velvet co-ord set," "chunky gold hoops" — ranked by how often shoppers wanted them. You stock toward real, demonstrated demand instead of guessing.

On the Demand intelligence page you'll find:

• Unmet searches (last 30 days) — the total volume of visual searches that found nothing above your match threshold.
• Demand clusters found — distinct look-alike groups discovered from those misses.
• Top demand clusters — each labelled in plain language, tagged High demand once at least 10 shoppers searched for it (or Emerging for an early pattern), with the example attributes that describe the look ("category: co-ord set, colour: green, material: velvet").

It's a genuine data moat: the signal comes from your own store's visual traffic and grows the more your store is searched — competitors structurally can't replicate it. And it's privacy-safe — only an anonymized pattern and label are stored, never the shopper's photo and never a shopper identity.

Tip: Pair this with the StockIQ connection to push these gaps straight into your sourcing workflow — see the Integrations guide for "Assortment gaps from unmet search."

Search analytics

On every plan; the lookback window grows with your plan — 7 days on Free, 30 on Starter, 90 on Growth, full history on Scale and above.

The Analytics page turns visual search into numbers you can act on:

• Searches — total visual search volume in your selected window, plus a daily trend line.
• No-result rate — the share of searches that found nothing above your threshold (your catalog gaps), trended over time.
• Click-through — how often a search leads to a product click.
• Search → purchase funnel — searches → clicks → add-to-cart → purchases, with the conversion rate at each step.
• Top styles shoppers convert on — the visual attributes of the products shoppers actually click after searching, so you can see what's working in your catalog.
• Search health — average, p50, and p95 latency. A healthy index keeps the slowest 5% of searches comfortably under two seconds.

Pick your date range at the top; on capped plans you'll see an honest "up to N days on your plan" badge with a link to extend the window.

Privacy: embed-and-discard

Shopper trust is built in, not bolted on.

A shopper's search photo is never stored. When a shopper searches by photo, the image is decoded, turned into a numeric embedding (a visual fingerprint), used to find matches, and then immediately discarded. It is not written to disk, not written to cloud/object storage, and not attached to any saved record. This is true for every storefront search path and for the in-admin Search playground.

What is stored is your own merchant data: your product-image embeddings, search analytics, result-click analytics, and the aggregate (anonymized) demand signals — everything the app needs to operate and improve visual search for you.

Recovery connections (see the Integrations guide) are the one nuance: if a shopper explicitly asks to save an item to a wishlist or get a back-in-stock alert from a no-result search, Image SearchIQ may retain the Shopify customer id as a delivery anchor for the connected Kernvel app. A customer-redact request removes it.

Compliance. Image SearchIQ honours Shopify's compliance webhooks — customer data requests, customer redaction, and shop redaction (store-scoped data is deleted after uninstall per Shopify's flow). Store data stays isolated to your store and encrypted at rest. A Data Processing Agreement is available on request.

Settings

The Settings page tunes every Image SearchIQ surface. Everything ships with sensible defaults — only change what you care about.

Visual search engine

The AI engine that powers your store's visual search is managed by Kernvel and included in your plan — there's nothing to configure. It's pinned to your catalog when your index first builds, so results stay consistent over time. (Switching engines would require a full re-index, so it's intentionally locked once set.) You'll see it as Active once your index is live, or Initializing before then.

Index & search behaviour

• Minimum match strictness — how close a visual match must be before it's shown. Higher = fewer, more confident matches; lower = more, looser matches. 62% is the recommended default. Below this line, shoppers see the gentle "closest matches" fallback and the miss is logged as a catalog gap.
• Results per search — how many matching products to show per visual search (4–48).
• Out-of-stock matches — what to do when a strong visual match is currently sold out: demote it, hide it, or show it normally.

Widget branding

• Accent colour — the colour used for the storefront camera button and modal accents.
• Camera icon placement — in the header beside search, or a floating button on storefront pages.

A reminder lives here too: the widget stays hidden until your catalog index is 80% built, so shoppers never see empty results.

AI surfaces

Master on/off switches per AI surface — visual search & playground, AI auto-tagging, hybrid search, visual recommendations, and visual demand intelligence. Tier-gating still applies: a row marked for a higher plan won't run until you upgrade, even if switched on, and shows a one-click path to billing.

Your plan limits

A read-only summary of what your current plan includes — products you can index and searches per month. Indexing cost and abuse protection are handled automatically behind the scenes; there are no usage surcharges.

Billing

Image SearchIQ is free to start — no credit card on the Free plan — and every paid tier includes a 14-day free trial. Annual billing saves about 17%. Charges are billed through your Shopify invoice; Kernvel never touches your card.

Plans at a glance

The Plans & billing page shows your current plan, usage meters (products indexed and photo searches this month), a monthly/yearly toggle, and the full plan grid. Upgrades take effect when Shopify confirms the new subscription.

What happens at your limits

• Searches. New searches over your monthly cap are gently rate-limited (the shopper sees a calm "check back later," not an error) — upgrade and they flow again immediately.
• Products. New products beyond your indexable limit queue and start the moment you upgrade. Your index is never deleted.

Payment trouble & downgrades

If a Shopify payment fails, you get a 3-day grace period — your search, widget, and index keep working. Update the card in Shopify billing and you're back to active automatically; after three days without payment the plan reverts to Free, and nothing is deleted. If you downgrade, your index keeps working and core visual search keeps running — AI surfaces above your new tier (recommendations, demand intelligence) simply pause until you re-upgrade.

FAQ

What counts as a photo search? Every visual search a shopper runs from your storefront widget counts toward your monthly cap. In-admin test searches you run yourself from the Search playground count too — Shopify doesn't distinguish.

What does "products indexed" mean? When you install, Image SearchIQ builds a private visual index of your product images. Each product counts once toward your indexed-products cap, no matter how many images it has. New products are indexed automatically as you add them.

Is the shopper's photo stored anywhere? No. Shopper search photos are analyzed to find matches and then discarded — never written to disk or cloud storage. Only your product images are indexed.

Why does Growth cost more than Starter? Starter covers the essentials: more products, basic hybrid search, and auto-tagging. Growth adds full hybrid search, AI auto-filtering, and the visual recommendations engine. Visual-search shoppers convert at a meaningfully higher rate, so these usually pay for the plan quickly.

What is Visual Demand Intelligence? On Scale and above, Image SearchIQ clusters the searches that returned no good match — the looks shoppers keep hunting for that you don't carry — and labels each cluster in plain language ("velvet co-ord set"). It's a stocking signal competitors structurally can't get, and it's privacy-safe: only an anonymized pattern and label are stored.

Can I switch between monthly and yearly? Yes. Yearly bills once per year and saves about 17%. You can upgrade tier at any time; plan changes take effect when Shopify confirms the new subscription.

Is there a free trial? Every paid tier includes a free trial. We don't ask for a card — Shopify holds the payment method. Cancel inside the trial and you owe nothing.

My search results feel too loose / too strict — what do I change? Adjust Minimum match strictness in Settings. Dial it up to show only near-identical matches; dial it down for broader recall.

Do shoppers ever see an empty "no results" page? No. If nothing crosses your match threshold, the shopper sees a "closest matches" fallback, and the miss is logged for you as a catalog gap (and, on Scale+, rolled into demand clusters).

Glossary

• Visual index — the private, store-scoped collection of "fingerprints" of your product images that visual search compares photos against.
• Embedding — the numeric "fingerprint" of an image (or a photo + text). Similar-looking things have similar embeddings, which is how matches are found. Shopper search embeddings are computed and discarded; product embeddings are stored.
• Embed-and-discard — the privacy model for shopper searches: embed the photo, find matches, drop the photo immediately.
• Hybrid search — a visual search refined with words ("this, but in blue").
• Crop-to-region — searching just the cropped part of a photo instead of the whole image.
• Match strictness (similarity threshold) — the minimum closeness a product must have to a shopper's photo to be shown as a confident match.
• No-result / catalog gap — a search where nothing crossed the match threshold; a signal that you may be missing what the shopper wanted.
• Demand cluster — a named group of look-alike unmet searches, surfaced as a stocking signal (Scale+).
• Pin / bury — manual merchandising levers that push a product up (pin) or down/out (bury) of visual search results.
• Auto-tagging / auto-filters — AI-discovered product attributes from images, used as storefront filters and to power smarter search.
• Incremental index — automatic re-indexing of just the products you've added or changed, so your index stays current with no manual rebuild.
• The 80% live gate — visual search reveals itself to shoppers only once your catalog is at least 80% indexed.