Step 1
Identify the item
Google's Gemini AI reads the photo: shape, material, maker's marks, labels, model numbers, printed dates and patterns. It names the item as specifically as the photo supports, for example a Pyrex 401 mixing bowl in the Pink Daisy pattern rather than just a glass mixing bowl.
Step 2
Search the web for prices
Gemini then runs Google searches for what the same item, or the closest match, sells for second-hand in the US. It is told to prefer completed sales and auction results over asking prices, and to say when it could only find asking prices.
Step 3
Adjust for visible condition
Chips, wear, stains, missing parts and a missing box all lower the price. The estimate only accounts for what it can see in the photo.
Step 4
Give a range, a confidence level and sources
You get a price range in US dollars that a typical private seller could expect, a confidence level with the reason, and the web pages the estimate was based on.
Step 5
Suggest a next step
We suggest what to do based on the top of the range: donate, sell it yourself, check further, or get a second opinion. These thresholds are fixed by us, not chosen by the AI.
Where the price comes from today
Right now every estimate comes from one source: Gemini, with Google Search switched on. That means the number is an AI estimate checked against web pages, not a calculation from a database of sales. It's why every result lists the pages it used, so you can open them and judge the number yourself.
We plan to add direct price data, starting with a trading card price database and then marketplace sales. When a source is added, results will say which one each number came from.
Asking prices and sold prices
Listings show what sellers hope to get, and many of those items never sell at that price. What buyers actually paid is usually lower. Gemini is told to prefer completed sales and auction results and to tell you when it could only find asking prices. You'll see that in the “based on” line of your result.
What the confidence levels mean
| Level | What it means |
|---|---|
| High | The exact item was identified and several consistent prices were found. |
| Medium | A close match was found, or prices for it vary a lot. |
| Low | The identification is uncertain or few prices were found. Treat the range as a rough starting point. |
What we can't tell from a photo
| We can't tell | Who can |
|---|---|
| Whether a designer bag, watch, autograph or artwork is genuine | an authenticator or specialist dealer |
| Gold or silver purity without a visible hallmark | a jeweler with an acid or XRF test |
| Gemstone quality | a gemologist (GIA-trained) |
| Whether electronics or mechanical items work | you, by testing it |
| Hidden damage, repairs or smell | an in-person look |
| Provenance: who owned it and when | your paperwork, or an appraiser |
How to get a better result
- Shoot in daylight on a plain background, with the whole item in frame.
- Make marks readable: the bottom of a dish, the inside of a ring band, the back of a card, the label inside a bag.
- Use the note for what the photo can't show: size, weight, “works”, “box included”, where it came from.
- If the item name in your result is wrong, run it again with a close-up of the identifying marks.
Where does the price come from?
Google's Gemini AI identifies the item from your photo, then searches the web for what the same or similar items sell for second-hand in the US. The range is its estimate from what it found. Every result lists the web pages it used so you can check them yourself.
Why is my range so wide?
Either the item is hard to pin down from the photo, or similar items sell for very different prices depending on condition and version. A clearer photo of the marks or model number, plus a short note about condition, usually narrows it.
What if the value looks wrong?
Check the item name first. If we identified the wrong thing, run it again with a close-up of the identifying marks and a note. Then look at the sources listed under the result to see what the estimate was based on.