Google parsing images?

[Dan_Vendel]

Ten minutes ago, I check the stats for a recent site. Noticed a hit through Google on the name of the photographer of one of the photos. Since I never, ever added that info as html, but only as pixelated text on the image itself, I was completely baffled.

To test, I made a search myself for the text I had on a couple of other photos, all jpgs, also with pixelated texts but only "" in the alt tags. Sure enough, they all showed up in Google with the pixelated text quoted as if it were html.

Can someone confirm that this is actually happening, or should I call my shrink?

Dan

[marketingmaniac]

It's probably part of the new Froogle beta,

[raptorix]

Is it possible that this information was stored as metadata in the JPEG?

[Dan_Vendel]

raptorix:

No. I made that image myself. Nothing what so ever hidden or otherwise available info in the fie itself. It was also found using the exact phrase on the image.

I think it's awesome if it's the new deal....

[raptorix]

Well its not very strange, maybe they just use some kind of OCR tool to parse Text, but must say its very funny, whats next in SEO land ;)

[whoisgregg]

[catalogs.google.com...]

Google's been working on analyzing text in images for a long time. I'd always expected it would find it's way to web graphics, sounds like it finally has. I suspect that was the original purpose of G catalogs, OCR training for Googlebot. :)

[whoisgregg]

Is it also possible that there are inbound links to that page with the anchor text, "Photographer Name"?

It's not unusual for a content-empty page (like a photo gallery page) to get it's keywords from off-page factors... worth investigating. :)

[vincevincevince]

My money would be on the photographer or similar having a page 'view my images' with his name on it.

[Dan_Vendel]

Nope. I have found out (after I started this thread) that Google actually have a gadget for scanning images and picking up words. As someone wrote earlier, it's a part of their Froogle, which they use for scanning e.g. catalogue images, etc.

(Maybe "scanning" isn't the correct term. I think they rather look into the code of the images and interpret that, as any image viewer do).