The below issue has now been fixed. The details in this post no longer apply.
This morning we began to receive complaints of almost exclusively false positives being returned in reports, as shown below:
Performing an search on the above image returns a variety of (non-matching) images:
The problem is with Google, our main data source
Concerned that our algorithms were getting the wrong data from Google, we tried the same search there, and found that they, in fact, are returning the wrong images:
Hope is not lost
While Google is showing the wrong image, the pages it is returning are correct (it typically returns the second or third image on the page, rather than the matching image.
That page contains the image we are looking for:
When will this be fixed?
Unfortunately, we have no control over this issue. We expect Google to fix it very quickly.
14/08: UPDATE: We’ve turned off the ‘discovered’ image for Google images (for now)
It’s been a week since this issue started, and as it still hasn’t been fixed, we have replaced the display of ‘discovered’ images with the one the user entered. In the vast majority of cases, the pages discovered do contain the input image, and showing an incorrect ‘discovered’ image was causing a lot of confusion.
We’ve pushed out on our ability to scrape alternative sources to counter this.
We will check the status of this issue on a daily basis and restore the ‘discovered’ images once it has been resolved.
27/08: Update: Issue fixed, everything is back to normal
This issue has been fixed and all systems have been reverted to their default behaviour. The details in this post no longer apply.
Over the next few days we will be rolling out a massive upgrade to the legacy monitoring system. This has been a complete rebuild of the underlying architecture, and will fix many of the long-standing issues with the system.
Rather than a “big bang” upgrade we will be rolling modules out slowly. To facilitate this, several small features have had to be disabled (as some functionality is changing). For instance, you will now need to log in before attempting to add any images to your account. We apologise for any inconvenience, although we’re very excited about these changes. We expect to have the entire system rolled out by 8pm GMT Wednesday the 1st of April.
We’re in the process of rolling Bing and Yandex image search results into our data. At the moment, you will only see a small subset of the available results until we verify the quality of their results. This means that you may now see a lot of new hits on images which have been dormant in your account for a while. Please let us know if you notice a rise in false positives.
This is an important change – not only will these engines increase the chance of getting all possible websites using your image, but it gives us a level of redundancy and means that we can still deliver some results when the parser breaks (i.e. every time Google updates its SERP layout).
The number of SEO agencies using Image Raider for large-scale projects has increased greatly over the past few weeks. This is great, but means that the processing queue is getting a little crowded and we’ve had to boost the capacity significantly. Continue reading