Google Reading Image Text?

BlackHat3

Regular Member
Joined
Sep 27, 2012
Messages
240
Reaction score
94
Hey guys,

I was doing some image searches and I found some meme's i was looking for by searching the text in the meme. Then i noticed that sometimes it is not in the filename or the text/code of the page it is on. Do you guys think google has some kind of OCR software in use? Thanks!
 
They started doing this a long time ago... lol.

http://techcrunch.com/2008/01/04/google-lodges-patent-for-reading-text-in-images-and-video/

do your research :P.
 
It could be OCR. It could also be that the image is associated by links or context with those words on other URLs, because that's more or less the basis for how google works in the first place.
 
It could be OCR. It could also be that the image is associated by links or context with those words on other URLs, because that's more or less the basis for how Google works in the first place.

They trademarked the technology behind being able to read the text in the image... if it is an "OCR", It's based around Google's own technology.
 
Yes they do. Proven 2012 on simple black/white image with text - jpeg ended up ranking for kw that was posted on it in images....
 
this google app is using some kind of ocr for recognizing text so they are probably also using this for their search engine.
 
Well, we tried to fool G on a site with "no ads" and added jpeg with advertisement - phone number, name, etc... that jpeg ended up ranking for those keywords... so YES... G knows what you write on naughty jpegs! :D not sure how good they are though, we had just white bg and black text - very readable.
 
Google does a lot more than just reading text, the analyze the image and find similar images. I have a reputation management client who has a bunch of pictures of himself in a suit and tie in a headshot style. All of the images that are ranking for this client are in a similar style (headshot with suit and tie). I've tested this with a couple of other things as well such as using a landscape for a nonsensical word and Google continues to prefer landscapes even when given pictures of people and animals.
 
Well, we tried to fool G on a site with "no ads" and added jpeg with advertisement - phone number, name, etc... that jpeg ended up ranking for those keywords... so YES... G knows what you write on naughty jpegs! :D not sure how good they are though, we had just white bg and black text - very readable.

I guess it would be like the captchas ... some is easy some not so much.
 
Even if they have been doing this for years, this is a great topic that is not touched on much. Great post OP
 
OP, why not share what you were searching for and what the result was so we can have some evidence of Googlebot OCR instead of these cool stories and Google intellectual property conjecture? Or anyone else, an actual Google SERP or Images search result with images that must have been OCR'ed.
 
I do know that I've tried searching for image macros based on what the caption on the image said, and not been able to find them so I don't know if Google is reading text from images and using it to rank files. If they are, I'm not sure they're doing it as well as they think they are. Still, I'd be interested in seeing some evidence
 
I used to work with some Google Apps ressellers in my previous work, and they made a demo with Google Docs. Guy was looking for an item on the search bar, and he could find a scan of a bill with the word in it. I guess that's a technology they use with Google Images too.
 
Back
Top