How to Remove a Word from Google AutoComplete

google autocompleteWhen a client asks for help with removing a negative search result, typically it’s just a typical ORM case of pushing up the good results over the one unsavory one. However, what is becoming more common is the discovery that even when unsavory results are long gone, Google, being ever so helpful, would like to remind people that they still might possibly want to see that unsavory result. I’m referring to Google Autocomplete.

You know when you’re typing in a search term and suddenly Google gives you a dropdown list of search terms to choose from? And, admit it, sometimes the suggested term was so bizarre or scandalous you just had to see what would pop up when you clicked it? Yeah, well it’s all fun and games until it happens to your name.

change google autocomplete

 

If you do a search out there on how to get this phenomena to stop happening, you will find a lot of talk but not much action. In other words, there is a lot of posturing on how Google arrives at these results and some theories about how to remove results from the autocomplete, but very little in the way of documented cases.  Since this blog is about helping the average Internet user do ORM for their name, I’ll skip the complex theories and ideas behind Google Autocomplete and try to get the heart of the matter: Removing Results from Autocomplete.

Basically, the idea of manipulating autocomplete will be the same as manipulating the SERPs for regular ORM. In other words, you won’t actually be removing the result so much as pushing it out with other more desirable results.

How many terms do you need?

At one time, Google was supplying as many as 10 autocomplete results for any given word searched. They seem to have cooled their boots a bit, though, and I’ve been hard pressed to find any word that comes up with more than 6 results at a time.  This means you should have at least 6 positive or neutral terms ready that you are wanting to be supplied at the end of your name or brand whenever anyone searches for you.

Match your prospective terms with corresponding content.

Google has to have a reason to supply these terms next to your name. So before you set off on your Autocomplete manipulation campaign, first you must supply some content about yourself to support these terms. If you are still in the infant stages of your ORM campaign consider starting with one the best guides I’ve ever come across for simple ideas on how to get content for your name out there.

Using Crowdsoucring to Manipulate Google Autocomplete.

As I said, there are many different factors on how the Autocomplete result populate.  A sizable contributing factor is the amount of independent searches for a term logged into a search query:

A Google spokesman…said that the Google Suggest function simply reflected the most common terms used in the past with words entered, so it was not Google itself that was making the suggestions.

Therefore, to play the game, you have to figure out ways to get crowds of people to search for the terms you want.  There are definitely some creative ideas out there when it comes to handling this, but for my case study, I chose to do it the American way: I paid for it… and I did it for 7 cents a click.

How to Manipulate Google Autocomplete with Mechanical Turk.

If you’re not yet familiar with Amazon’s Mechanical Turk, it is the leading marketplace for crowdsourcing tasks that are too intelligent to demand from computers but are too massive or, in my opinion, too menial to demand all from one person.  So, if you want something done en masse, why not get thousands of people from all over the world to do your task in 30 seconds?

I had a lot of fun setting up my experiment with Mturk. I spent all day tweaking my task copy to get more and more participants. (You have to find a balance between being the highest bidder and finding the most alluring headline to draw them into your task). Here are some tips to help you along if you’re just staring out with Mturk.

Google Autocomplete Manipulation Case Study Setup:

Using my own name for the case study seemed like the wisest route, but I still had to come up with a term. I would suggest if you are doing this for your own name, you come up with terms surrounding your niche or Internet persona. I could have very well done that here, but I really wanted to come out of left field to make sure there wasn’t already a Google bias going on. In other words, Google already has data for me in certain niches, like location, or reputation management, so I wanted to start fresh with this experiment. Remembering that Google needs to have some content to correlate to the term, I decided to start with a term from my previous post: Charity.

On 2/12 (roughly 3 weeks ago) when my name was searched, these were the autocomplete results that were shown:

lauren starling autocomplete

As you can see, these are the only results that pop up. Nothing related to charity. Infact none of these results have anything to do with me …time to change that!

Next, I headed over to Mturk to set up my project. Now remember, the key is we want people to search for a specific term.  Mturk has some guidelines on what they will and won’t allow so be sure to make your project a little more in depth than just “searh xyz plz”.  Plus, you want to be sure the workers are actually doing what they’re supposed to and not just saying they are. Also, you want them to click through on the content you previously provided to make sure Google starts to associate the your new term with your content.

The instructions for the workers form looked like this:

 

removing autocomplete

 

After submitting the project a few different times with varying pay outs, I settled on one that cost me $ 0.07  per worker/task. I was getting an about 25-35 task completed per day. I set the campaign to stop at 15 dollars (150 tasks).

Results:

I checked the SERPs every day for a while to see if there was any change. After about 2 weeks I got distracted ( and stopped checking) only to remember yesterday and discover:

removing autocomplete results

So somewhere in the second to third week of this campaign, Google updated their autocomplete to reflect the sudden influx of searches for this term.

Take Away:

  • ‘Lauren Starling’ is not a very heavily searched keyword. However, I’m assuming, neither is your name. Therefore, you should be able to accomplish the same type of results using roughly the same methods. If not, step it up a bit with more searches for a longer period of time.
  • Remember, in ORM in order to “remove” a result you have to overcome and push out the negative results with positive ones. Obviously, I only did one term, you will need to try this with at least 4- 6 (maybe throw in a few more for good measure) to knock off whatever it is you’re trying to eradicate.  I have no idea what these extra variables will do to the equation, if anything.  Sidenote: Interestingly enough, I did knock out a term (compare before and after pics) and seemed to have replaced it with the new one. So perhaps for lower search volume terms the max Autocomplete is 4.
  • I have no idea how long this search term will last in Autocomplete, now that no one is searching for it (although this article could serve as a self fulfilling prophesy and could sustain results indefinitely). You may want to think about extending the mturk campaign for as long as possible, perhaps with max tasks per day.

 


About

Lauren is an SEO consultant specializing in Online Reputation Management. Her passion is educating and assisting everyday people in controlling the content results in the SERPs for their name or brand.

  • http://blog.clayburngriffin.com/ Clayburn Griffin

    Oddly, Mechanical Turk started moderating these kinds of tasks. You haven’t had issues with that?

    Once I got the word in for my name, it stuck around for about two or three months. It dropped out, so I threw a few bucks at it again and it came right back. Now it’s been there several months without me doing anything.

    Also, I did a search for you! So maybe that will help it persist.

    • http://www.laurenstarling.org/ Lauren Starling

      I was fully expecting the possibility of being moderated, as I had seen others say the same, but no, I wasn’t. I even redid the project 17 times (tweaking the “sales copy” to get a better CTR). The project in mturk was presented as a quality control issue as to what people were seeing spread across the country when they searched for a particular term. I saw other similar projects. Considering just today I asked a group of colleagues in a skype chat to check the autocomplete results to be sure everyone was seeing the same results as I was, this may indeed be a valid request!

      • http://blog.clayburngriffin.com/ Clayburn Griffin

        I framed it as that way back in the day too, and they started getting moderated. Could have been that we were running such large volume on brand terms that it was obvious.

  • http://twitter.com/AdamGSteele Adam Steele

    Hey Lauren- this is great!

    I used Microworkers, and didn’t have much luck. The neg terms that were being reflected in Suggest/autocomplete, per GWT, were being plugged in a heck of a lot, so I needed to out do those numbers. Worse, this was in .CA and so I needed go gather CDN searchers, or someone with tons of CDN IPs to perform the duty. Unfortunately, it was very hard to do either, and I was unable to create more search queries for the positive terms than were occurring for the neg ones.

    Interestingly, these CDN actions (search + click) cost .30/per. I would imagine this has to do with the lack of supply of CDN workers.

    I like your thoughts on creating an attractive title – I never put much effort in doing so. I imagine I may have had more luck attracting workers had I done so. Split test titles?? haha.

    First I am seeing of your blog, and looking forward to digging through some older posts.

    Hope you are well!

    Adam

  • Adam Tal

    Aren’t the people on mturk mostly from Asia etc? I would have assumed Google could pick that up and only suggest your keyword in the countries where the searches came from. Am I wrong?

  • Sanjeet Kumar

    Thanks, for this helpful post. I have a query. Once i decided keywords i need to publish content in support of those keywords but it is not necessary that content i will publish will appear top in search result, so how those content will get click. or is it ok if we not get click and just get searched from different people ?

    Please help, Thanks a lot !

  • http://fiswebdesign.com/ Aaron Hemmelgarn

    This is like digging in your backyard and hitting a large rock which turns out to be gold. Great for SEO…! Thanks for this article…

  • http://fiswebdesign.com/ Aaron Hemmelgarn

    Oh and I see a bunch of your social media links in the sidebar, but where is your Google+?