Ash Nallawalla's blog

Creating a Google-proof persona

A persona is a fictitious person that has certain defined attributes. In product marketing, we create personas for major groups of users who will use the product. For example, a word processor’s set of personas might include a high school student, a university student, a generic office worker, a specialist author, a manager, and so on.

In the world of black-hat SEO or spamming, a persona is usually a very shallow person, with no thought given to its creation. Beyond a rather implausible Western name and Gmail address, there is no sophistication, perhaps because the only purpose of that persona is to send once-off spam. Since you can create billions of fake Gmail/Yahoo/Rediffmail accounts without any worry, you can create a new one for each email if you wish.

Aaron Wall has written an interesting flight of fancy (I mean it as a compliment) in Google+ and probably had a lot of fun speculating how Google could determine a persona to be a real person. You can almost hear the chuckles in Amit Singhal’s and Matt Cutts’ teams at the Googleplex.

Aaron speculates that the following behaviours help to brand a persona as a real person:

  • Quality of Gmail account and those of correspondents
  • Google Wallet and Checkout usage patterns
  • Google Maps use and travel patterns near credit card address
  • Use of YouTube
  • Use of +1 button

For the details you will need to see his Google+ post.

My take

Aaron has made a great start but IMHO other behaviours can be deduced. I spend most of my time with large corporate sites and reading the above with that lens made me shake my head. There is often no corporate Google account other than to create a WMT account at best. There wouldn’t be a credit card tied to that account. It wouldn’t use Maps to get directions. It wouldn’t watch YT. etc

Such a filter is fine for removing scraper sites from further evaluation, but I have a problem with his statement: “Of course no user will score super high on everything, but they can get probabilities & toss out usage data on anything below an 80% level of confidence.”

If this were so, then most corporate personas would fail, leaving their sites in peril.

I strongly believe that sites that pass a TrustRank (PDF) test with a high score are immune from checks that the rest of our sites have to endure.

Creating Google-proof personas

Let’s leave spammers out of this article. At my Australia/New Zealand Directory I see many SEO companies submitting links on behalf of clients with a fresh Gmail address that is probably not used after an initial round of link submissions. They might use that address to submit some articles to directories and the really inept agencies might use it for comment and forum signature spam. That’s it.

What’s wrong with this picture? Anyone in our industry can spot one of these Gmail addresses as a fake often by looking at them. I delete whole chunks of waiting submissions merely by looking at the address and not the actual submission. They are always a text string that ends with some digits.

I am not a retail SEO, so I don’t need to do this, but in the interests of improving the industry, here is how I would go about setting up a persona (leaving out details that might help the wrong people):

  • Create a spreadsheet with multiple columns and refer to it when using a persona. Place each persona on a new line.
  • Choose a realistic name that doesn’t draw attention. A “Mark Smith” will pass visual scrutiny, but a Barr. Wardt Wodelt (a real example in my junk folder today) looks suspicious.
  • Find a realistic photo of someone who isn’t a model. The plainer the better.
  • Fill the spreadsheet with the persona’s CV and various details. If they were born in New York and live in Los Angeles, then they need to be seen to write various things as if they currently live in LA. Their high school and university could be in one of those two places.
  • Open accounts at various online places with the same nick and same personal details, so that a web search for the nick will produce a lot of results pointing to the same full name and location. A real person would usually have a Facebook account, so ensure that it has some activity at regular intervals and performs things that real people do, e.g. add apps, Like articles, leave comments etc. Their LinkedIn account would need to show the same educational institutions and locations.
  • Create many more personas as needed, not all at once.

I won’t elaborate on how to make these personas more convincing, other than to say that they should have been created a long time ago, gradually, perhaps from different cities when you were visiting them. Creating 20 Gmail accounts from the same IP address in one session is a bad idea.

I don’t use many Google services, such as Checkout, Picasa, Gmail.com address, etc, so I might score low in Aaron’s list of checkpoints. However, I use addresses that were created in 1994 and 2002 and have left a vast trail all over the web since then. Spoofed spam has been sent from one of those, but I have not noticed any lasting damage to rankings, if any. I do participate in Google+, Groups, Orkut, WMT, Maps, and some other Google services, so my various accounts should look very human.

 

Ash Nallawalla

Search strategist experienced in large, complex websites. Ash's Google+ profile

Related Posts

President Obama’s whitehouse.gov pages archived

Ash Nallawalla

22 January 2017

Other, SEO

Feel free to share...After Mr Donald Trump became the president at noon today (USA EST), many reported that the whitehouse.gov website removed references to Climate Change and LGBT. That isn’t entirely accurate. The website up to that point has been archived and can be found at https://obamawhitehouse.archives.gov/. The LGBT URL was https://www.whitehouse.gov/lgbt but it redirects […]

Read More

Coding Australian 13 and 1300 “tel:” numbers – 404 errors

Feel free to share...Australia has six- and ten-digit local numbers that begin with 13 and 1300 respectively. The 13 xx xx numbers connect you to a local number in your city so it is a local call. The 1300 xxx xxx number is similar but is usually a single destination for the price of a […]

Read More

Older Posts