Google: Our Systems Cannot Determine Accuracy Of Content

Danny Sullivan from Google said on Twitter “Machines can’t tell the “accuracy” of content. Our systems rely instead on signals we find align with relevancy of topic and authority.” So Google cannot determine the accuracy of content, I guess.

Here is the tweet:

Machines can’t tell the “accuracy” of content. Our systems rely instead on signals we find align with relevancy of topic and authority. See: https://t.co/O65v1PTehr and https://t.co/cTveD8XNxp

— Danny Sullivan (@dannysullivan) September 9, 2019

Remember in 2017, Google launched the Google Trust Project that looks for signals to help determine accuracy of some content. Also Project Owl looks to promote authoritative content and an algorithm update that removes inaccurate results?

So Google cannot determine accuracy but they use other signals to do so. They do not use popularity:

No. It is not popularity.

— Danny Sullivan (@dannysullivan) September 9, 2019

That doesn’t say popularity = authority. Not, again, speaking as someone who work within Google Search, does it. That would be a far too simplistic signal to use & wouldn’t apply for the myriad of queries we handle, 15% of which are new….

— Danny Sullivan (@dannysullivan) September 9, 2019

I was asked how we verify accuracy. I replied we don’t and can’t but look to signals about authority. You then said that means popularity. I said it doesn’t. And, in fact, issues with featured snippet can happen with both “popular” and not-so-popular sites….

— Danny Sullivan (@dannysullivan) September 9, 2019

So again, popularity doesn’t equal authority and would be far too simplistic a signal to use. As for featured snippets, this covers how we try to keep improving and methods and policies when they fall short https://t.co/GPAtqNCuik

— Danny Sullivan (@dannysullivan) September 9, 2019

As for highlighting stats, we or any system that looks across the web can’t know if stats are accurate. We can identify things that appear to be stats, which appear to be from authoritative content & that often aligns with accuracy.

— Danny Sullivan (@dannysullivan) September 10, 2019

There are things machines & humans could do to try & verify accuracy, absolutely. I’m sure we’ll explore such more. But there’s so much content that isn’t even factually based, so AccuracyRank™ wouldn’t be that helpful. For now, we focus on trying to determine authority.

— Danny Sullivan (@dannysullivan) September 10, 2019

This is fun.

YES, a thousand times YES!

Thank you @dannysullivan. This is a myth that will not die.

It doesn’t help that G links to content that claims content accuracy is a ranking factor, but… we’ve already had that argument. ????

cc @schachin @RyanJones for interest in “ranking factors” https://t.co/5BnEJGxmYc

— Jenny Halasz (@jennyhalasz) September 9, 2019

Anyway – we all know Google’s algorithms are complex. Then when you start labeling specifics attributes of the algorithm, you can go down this long path that leads to no where good.

Forum discussion at Twitter.