LET'S TALK GOOGLE IMAGE
I am trying to drag my butt into this AI stuff and thought I would see what was what in using Google Image.
Where do I begin to tell the story? I have been watching for a while the plethora of miss information and miss identifications that have hit the jewelry category here on eBay and other places. I now know why some of it is happening. Some of the identifications used by Google Images is miss leading and, in my opinion, just wrong.
I do like a lot of the descriptions, without the identifications. And in using the dive deeper mode it provided some options in the descriptions that I never thought of.
For us older collectors who have been collecting for years the miss identifications aren't a real issue as most of us understand what we have and our buyers know what they are looking for and what to look for. BUT for the newer sellers and buyers this could be a mind field, in my humble opinion.
I think some of the identifications could be useful in places to start your research. Unfortunately, it appears to being used as the final word by many sellers. What do you think?
hertzgirlz
·6 months agoScifi that is one of the best descriptions of google lens I have ever heard!!
tobaccocardyahoo
·6 months agoIMO Google Lens does as well as some antique dealers and appraisers. A lot of guess work because the data used to train the tool is not reliable or complete.
I have used it to get a rough approximation of what an item is, and then refined my input to it, so that it focuses on the segment of information I note is most likely to be reliable. The result is often right on the mark.
When we take stabs at a thread that an Ebay seller has created on this forum, there are many responses which are blind alleys or truly shots from the hip. All of those responses, here and on other sites contribute to the incorrect data AI tools must deal with.
As an individual whose past job experience has been in marketing, Ebay AI descriptions are in line with the output I received from many advertising copywriters who were describing products they knew nothing about. They and their agencies were replaced when I gained control of marketing for the companies I worked for. Ebay's AI descriptions cost a whole lot less to generate than these equally ignorant humans created.
scifidaughter64
·6 months agoMy 2cents worth. Google lens is like asking your drunk friend a question. Sometimes you get a decent answer and sometimes you wonder what they were looking at because it wasn't what you took a picture of.
dream*peddler
·6 months agoLOL @scifidaughter64 ! Good one!
southernmermaid8
·6 months agoSci - fi 😂🤣😂
christworks
·6 months agoI always start with eBay's image tool first when I'm out sourcing. If I do not get anything popping up on eBay, I will go to Google Lens. I use multiple other AI tools for deeper research at home. I have found ChatGPT to be the most dynamic and helpful for deep research and recognizing things that a simple Google Lens search won't pick up. I have ChatGPT set up to write eBay Keyword titles and condition descriptions based on rules I've set up for it to follow. I do not use it to write descriptions per se as I just copy/paste the condition into description. Buyers rarely read the descriptions, and they don't need me to sell them on the item with a bunch of subjective flowerly language.
flipping_treasures
OP6 months agoGreat information Dream.
I agree Sarah, it makes my brain hurt too some days . Things just seem to change too fast now days.
evelyb30
·6 months agoJust about half of the time I use Google image search I end up wanting to toss the phone or tablet across the room. When offering feedback, when they still allowed me to, I'd say "exact match or no match. Period. " or words to that effect. People rely on the garbage results. Me, I'm surprised when they actually get something right. It's a rarity.
vernon1959
·6 months agoI do utilize google lens-a little. as mentioned, miss identification drives me nuts! But it is good jumping off point for descriptives, especially if I'm at a total loss for an item. I sort of have a love/hate relationship with it 🙂
1940fc
·6 months agoSorry ya`ll
flipping_treasures
OP6 months agoAs far as descriptions goes, I like Google more than eBay. I too think eBay is too flowery. I do think Google does a fairly good job on dating.
mscurl
·6 months agoAgreed. I feel like people just don’t want to take the time to properly research anything anymore, so they just use google lens and go with whatever looks somewhat close. There might even be some who are intentionally picking what has the highest value and go with that. Caveat Emptor. That being said, it sure is a powerful and useful tool: I have this big old ornate chunky brass belt that was my mother’s from the ‘70’s and the name was worn off the leather. I’ve been unable to ID it for years and years. On a whim, I image searched it recently, and danged if it didn’t come up right away. I never thought about using some of the descriptions which is pretty brilliant imho. Whatever eBay is using for AI isn’t that great - I can always tell that a seller used it. Have a great weekend people.
bat197
·6 months agoAI is a double edged sword. It certainly makes mistakes. E-Bay's version is nothing to write home about. It's too artificial when writing descriptions. It sounds like 1950 advertising. That said it can be useful along with Lens in quickly identifying certain items. You have to have some sense in using it. which some people don't have.
dream*peddler
·6 months agoI have a love/hate relationship with Google Lens. I'll be at an estate sale, yardsale or thrift store and will look something up and see great resale prices. Do you think I can find those same prices when I get home? Anyway, as you've all said, it's a good tool to start the research process. Sometimes it will nail it and other times it won't. I agree with mscurl and bat197 in that I don't care for A I written descriptions. Way too flowery and verbose. Here is what it says about Google Lens:
🧠 How AI actually operates on information
AI models work by analyzing patterns in the data they were trained on. That training happened before you ever used the tool. Once the model is trained:
It does not keep or store your personal inputs to “learn” from them in real time.
It does not build a memory of your items, photos, or searches.
It does not improve its accuracy based on what you personally upload.
Instead, it uses the patterns it already learned during training to make its best guess about what it sees.
This is why Google Lens can be:
very good at describing shapes, colors, textures
very bad at identifying specialized items like vintage jewelry, art glass, or collectibles
It’s matching your photo to the closest thing it has seen in its training data, not to expert knowledge.
🔍 Why misidentifications happen so often
Image-recognition AI tends to:
prioritize visual similarity over historical or maker accuracy
match to popular items, not rare ones
confuse style with origin
treat keywords from similar images as facts
So if a brooch looks vaguely like a modern reproduction, Lens may confidently label it as such—even if your piece is a 1940s original.
This is exactly the issue you described in the eBay jewelry community thread: newer sellers assume the AI’s guess is authoritative, when it’s really just a pattern match.
📚 Does AI “collect” information from users?
There are two separate things to understand:
1. The model itself does not learn from your individual inputs.
It doesn’t update its knowledge or store your photos to improve its accuracy.
2. Some platforms may log usage data for product improvement.
This is a platform-level choice (Google, eBay, etc.), not the AI model “learning.” Even then, the data is used in aggregate—not as personal memory.
But the AI’s answers come from its pre‑training, not from anything it “collects” from you.
🧭 What this means for collectors and sellers
AI image tools are useful for:
generating descriptive words
spotting style similarities
giving you starting points for research
They are not reliable for:
maker identification
dating
authenticity
valuation
Experienced collectors can filter out the nonsense. Newer sellers often can’t—and that’s where the misinformation spreads.