When a Customer Sends a Photo and Asks "Do You Have This?" — Reply Instantly With AI Image Search
A massive chunk of DMs to fashion, jewelry, home, and even restaurant businesses are exactly this: a screenshot of a product, followed by "do you have this?" or "how much?". Without AI image search, replying takes a member of your team five to ten minutes of squinting and searching. With AI image search, the answer is on screen in seconds, with the real price from your catalog. For a busy DM-driven business, it's the single highest-impact AI feature you can turn on.
- Customers screenshot products from Instagram, TikTok, or competitor sites and ask "do you have this?" — somewhere between 30% and 60% of DMs to a fashion, jewelry, or home brand contain a photo.
- AI image search recognises the photo and matches it to your real catalog in seconds, then quotes only your real price and stock.
- The same loop sends a payment link or books an appointment right after — turning "do you have this?" into a sale in one chat.
The business problem
Scroll the Instagram inbox of any active fashion, beauty, jewelry, or home-goods business. A huge share of the messages look like:
- "Do you have this?"
- "How much?"
- "Is this still available?"
Each one with a photo attached.
Without AI image search, your team opens the photo, opens your catalog, types a guess at what the product is called, scrolls, gives up, opens the product, copies the price, switches back to the DM, types a reply. Five to ten minutes for one message. Twenty messages a day = two to three hours of staff time gone — and the customer has usually moved on or bought from a competitor by then.
What goes wrong with the usual approach
Three patterns:
- Manual lookup, every time. Slow, error-prone, doesn't scale, only works during business hours.
- "Please share the product name or SKU." The customer doesn't know the name. Half drop off.
- A generic AI that pretends. Without a real link to your catalog, the AI guesses based on what the photo looks like and sometimes quotes a competitor's price or invents one. Big problem.
None of these win the sale. The customer arrived hot — they liked a product enough to screenshot it. They lose interest within minutes if you don't reply.
How to think about it
AI image search is the difference between "we'll get back to you" and "yes, we have it, here's the link to buy, want it tomorrow?".
Under the hood it's three steps, all hidden from you and the customer:
- The photo arrives on WhatsApp, Instagram, Messenger — anywhere.
- The AI recognises the product from your real catalog. It compares the photo to product photos you've uploaded, finds the closest match, and gets a confidence score.
- The AI replies with the real price and stock from your live data, and offers the next move — a payment link, a booking, or a clarifying question if nothing matched confidently.
The whole loop runs in seconds.
How UptoNova helps
UptoNova's image search is built into the AI sales agent. As an AI agent for Instagram and WhatsApp, it reads the photo, matches it to your catalog, and answers in the same DM. Because it's a grounded AI that never makes up prices, every quote comes from your live data — never a guess. Specifically:
- Connect your catalog once. Upload via CSV or connect Shopify / WooCommerce directly. With Shopify WhatsApp integration, UptoNova reads each product's photos and details and learns them, so a photo on WhatsApp resolves to the right SKU and price.
- Recognises customer photos automatically. When a photo arrives on any channel, the AI finds the best match in your catalog with a confidence score.
- Only quotes from your real prices. The matched product's real price and current stock come straight from your live catalog — never invented.
- Asks when it isn't sure. If the photo doesn't confidently match anything you sell, the AI says "we don't carry that exact piece — would you like to see similar?" instead of forcing a wrong match.
- Closes the sale right there. Matched and confident? The AI sends a Stripe payment link or books an appointment in the same chat.
- Works on every channel. WhatsApp, Instagram DMs, Messenger, Telegram, web chat. See the multi-channel inbox guide.
What this looks like in a real conversation
Here is what this looks like in a test scenario with a sample jewelry catalog:
Customer: [sends photo of a gold pendant on a chain]
Customer: Do you have this?
AI (less than 2 seconds later): That looks like our Aurora pendant in 14k gold — in stock at $189. Want me to send you a checkout link?
Customer: Yes please
AI: Here's your link: [payment link]
Zero team time. A staff member can review the trace afterward if they want, but the sale closed by itself.
Honest failure modes
| What can go wrong | Why | How UptoNova handles it |
|---|---|---|
| Wrong colour matched | The AI sees the silhouette better than the precise colour | The AI asks "did you want it in navy or charcoal?" |
| A competitor's product gets a closest-match reply | Confidence threshold set too low | Above a confidence floor we match. Below it, the AI asks instead of guessing. |
| Catalog photos are studio shots; customer photos are lifestyle | Visual style mismatch | Add a lifestyle photo per product — accuracy jumps noticeably |
| Heavily watermarked screenshots | Watermarks confuse the visual match | The AI asks the customer to describe what they're looking for |
Setup playbook for an online store
- Connect your catalog. Shopify, WooCommerce, or upload a CSV.
- Confirm photos. Every product needs at least one clear photo. The clearer the photo, the better the match.
- Connect your channels. WhatsApp, Instagram, Messenger — all in one inbox.
- Test with 10 real customer photos. Drag a few real ones in. Confirm the matches are sensible. Adjust the confidence threshold if needed.
- Switch on for live customers. Watch the first day in the trace view — every match shows its confidence so you can spot drift.
Frequently asked questions
How accurate is AI image search?
For most product catalogs with decent photos, the top match is right about 70-85% of the time, and the right product is in the top three suggestions 90%+. Accuracy drops on catalogs where many products look near-identical — like plain T-shirts in 20 colours — where the AI will ask the customer to pick a colour.
Does it work on screenshots with text overlays?
Mostly. Light text overlays — captions, story handles — are fine. Heavy stickers across the whole image hurt accuracy. When confidence is low, the AI asks a clarifying question instead of forcing a wrong match.
What about competitor products?
If the customer sends a competitor's photo and you don't carry it, the AI says so honestly: "we don't carry that exact piece — would you like to see similar?" That keeps trust intact and opens a cross-sell.
Does it work across categories — clothing, jewelry, home goods, food?
Yes for most. Apparel, jewelry, home goods, beauty, and even restaurant dishes work well. Harder categories: precise electronics where the model number matters more than the look, and regulated products like pharmaceuticals.
Can it handle photos with multiple products?
Partly. The AI focuses on the most prominent product and quotes that, then asks "I see X — anything else you wanted from the photo?" rather than guessing all of them.