Why Google Lens Can Identify It But Not Price It

Point your phone at a lamp, a jacket, a stack of records, and something like Google Lens will tell you exactly what it is in about two seconds. Brand, model, sometimes the specific edition. That part of the problem is basically solved. The part it doesn't solve is the one you actually opened the camera for: what do I do with this, and what is it worth.

Identification and pricing are two different problems that happen to look like one problem because they both start with a photo. Lens, and the dozen visual search tools built on the same idea, solves the first one well. None of them solve the second, because the second one isn't a photo problem at all.

Naming a thing is not the same as pricing it

Say Lens tells you the lamp is a Stiffel table lamp from the 1960s. That is genuinely useful information. It is also, by itself, worth nothing in dollars. A Stiffel lamp with a working three-way switch and its original harp sells for a real number. A Stiffel lamp with a cracked socket and a shade that doesn't match sells for a much smaller number, or for nothing, because nobody wants it enough to pay shipping. The name doesn't distinguish between those two lamps. Only a market does.

This is true of almost everything that gets photographed for pricing. A brand and model tells you the population of things it could be compared to. It doesn't tell you where in that population your specific item falls, and the spread inside one model number is often wider than the gap between two different models entirely.

Variant and condition do the work identification can't

Two objects can share a name and be worth wildly different amounts. The variant matters: a first-run pressing versus a reissue, a discontinued colorway versus the one still in stores, a limited run versus the standard model that a single photo has no way to tell apart. Condition matters even more, and it is where identification apps go quiet, because condition isn't a fact about what the object is, it's a fact about what happened to it. A scratch, a repair, a missing box, a smell. None of that shows up in a model number.

Price×Point treats identification as the first step, not the last one. Once an item is identified, it still has to be matched against completed sales of that same variant in comparable condition before a number means anything. How to check an item's condition walks through the grades that separate a listing that sells from one that sits.

A photo can start a price, but it can't finish one

Here is the honest version of what a photo-to-price flow can and can't do. It can identify the object. It can pull the sales history for that object, real completed transactions with real final prices, not asking prices from listings that never sold. It can show what a retailer charges for a new one, for comparison. What it can't do is inspect the object for you. It can't see the hairline crack in the base, or know that the zipper sticks, or notice the box is missing. You have to bring that part yourself.

That's also why an unpriced result is sometimes the honest result. If nothing comparable has actually sold, the right answer is that there isn't enough data, not a guessed number dressed up to look precise. A number with no sales behind it is worse than no number, because it looks like evidence and isn't.

Sold price and asking price are not the same question either

Even once you have real comparable items, there's a second gap that photo identification alone will never close: the difference between what something recently sold for and what people are currently asking for one like it. Asking prices skew high, because a seller can always come down later and can't go up once someone's agreed to buy. A tool that shows "similar items for sale" without separating those from actual completed sales is answering a different question than the one you asked. Sold price versus asking price covers why that gap exists and why it doesn't close on its own.

What actually closes the gap

The fix isn't a smarter camera. It's connecting identification to a market. Once the app knows what the object is, the next step is to look up real completed sales of that specific thing, filter by the condition and variant you're actually holding, and report the median of what people paid, not what people are hoping to get. Across Price×Point scans, roughly one in six items resells above what a retailer charges for a new one, which is the kind of fact you can only learn by comparing a specific item to actual sales, not by naming it correctly.

None of this makes the photo step useless. It's the fast, correct way to get from figuring out what something is to knowing what to compare it against. It's just not the finish line people assume it is when an app hands them a name and nothing else. If you're staring at a box of stuff after a photo told you what half of it is called and none of what it's worth, the next useful thing to learn is how to find out what your stuff is worth, which is the part identification was never built to answer.