Manipulating Google Suggest Results – An Alternative Theory

by rishil on March 8, 2011

Is SEO a gazetted Officer?

Is SEO a gazetted Officer?

Google Suggest is a Reputation Management Nightmare at times. A number of companies have been hit and hurt by results that show up with “Company Name + Scam” for example. The problem with those results is that when users see the suggestion, they are immediately tempted to click on them, as opposed to their original query.  In fact, google have been sued in France successfully and forced to remove some negative suggest results .

Is it Costing Businesses Money?

The other effect is financial. Ever seen Google Suggest results for businesses that run Promotional or Voucher Codes?

Currys Voucher and Discount

Currys Voucher and Discount

The obvious impact is that users are tempted to find a discount or voucher code , even though their original journey would have been to the brand.  This means that a number of Promo Code websites are profiting from Google Artificially showing those suggest phrases.

SO How Does Google Suggest Work?

Andy Beal who runs the awesome reputation management tool trackur, suggests (excuse the pun) that user behavior and search volume impacts google suggest:

In practice, it’s not quite that simple. After some testing, it appears that Google Suggest takes into account the following:

1. What was the first keyword search? For example, someone searches for my company, Trackur.

2. They then decide to refine those results by changing their search to “Trackur Free”–which was what they were looking for all along.

3. If enough people make that refinement, compared to simply searching Trackur, then it starts showing up in Google Suggest.

This is a theory adopted by a number of SEOs. In fact, a very good friend of mine, Brent Payne manipulated Google suggest just to show that sheer volume CAN change those suggest results. The first test worked and failed, while the result of the second is still live:

Brent Payne Manipulated This

Brent Payne Manipulated This

The Alternative Theory

The problem with the above method of manipulation is that it requires huge volume, and cost. It also sends some sort of manipulation signals, and can be caught. So how can you affect Google Suggest without the use of Mechanical Turk or other large scale manual manipulation techniques?

Let me take you to a post from 2009 that gave me a “tubelight” moment (bolding mine):

google suggest results are hurting my business. When typing in my domain (which people do for whatever reason in a search engine) it shows “MyDomain.com” as the 1st result but shows my domain with negative words about my domain as the 2nd and final suggestionMyDomain.com sucks“. Im needing to alter the google suggestions. I saw a competitor of mine had this “sucks” term listed on their site at one point but am not sure if thats the reason its now suggested and connected to my domain. I doubt many people would actually search for “www.mydomain.com_sucks”. Its all a bit odd. (Webmaster World)

The question is answered by Robert Charlton – and kinda hits a decent sized nail in the head (see my bolded portion):

suggest was a Labs project before it became a standard feature on the search box. In the Labs incarnation and in the early days of the search box implementation, numbers were returned along with the phrases. The numbers then reflected the number of pages returned, not search popularity.

Ultimately, google dropped the numbers, and I think the official word (or popular supposition) was that they did so because they were distracting and people weren’t really looking at them. But it’s also likely that google dropped these because they had factored other considerations into the personalized (and perhaps other) configurations of the tool, that the numbers therefore weren’t really correlating any more with what the tool was showing in the personalized setup, and it was too complicated to switch displays.

So Can Sheer Volume of Content Manipulate Google Suggest?

In fact a few weeks ago, Malcolm Coles and I were having a discussion on Twitter about this very phenomenon and he has put together a number of other people reporting exactly what I am – that Google Suggest does NOT only rely on search volume.

I decided to run a few tests. (lazy ones to be honest). In the first instance, I decided to use the power of twitter and RTs to drive indexed content for my two chosen terms, “ha ha ha” and “rofl” to be appended to my name in suggest.

The test with “ha ha ha” had a strange result, instead of going on suggest, it went on to appear in related searches:

Rishi Lakhani ha ha ha

Rishi Lakhani ha ha ha

So I pushed ROFL. And as you can see, as of last week:

Rishi Lakhani ROFL

Rishi Lakhani ROFL

Is Manipulating google Suggest Against their Guidelines?

Simply put yes. Therefore I warn you, you can and probably be penalised for the two suggestions below. This puts Google Suggest (now google Autocomplete) into the blackhat SEO category. Let me pull out a tweet from Danny Sullivan during SMX West

Google views manipulating suggestions as “abuse” & will take “corrective” action as needed #smx #1b2

Using Google Suggest Manipulation for Good

Source kandyjaxx on flickr

Source kandyjaxx on flickr

I ran a similar test for a brand, where they really wanted to push a product. The sheer volume of search on the brand was quite heavy, so we tried to get the product showing up for google suggest. (I can’t show the result as I have an NDA unfortunately).  It worked for a very short time however, and I need to hone the test a bit more.

But how does that help?

  • Allows you to force branded queries into common deep product searches (better conversion!)
  • Allows you to remove discount queries so as to claim back brand search share and discourage discounting
  • Allows you to reputation Manage more easily than using mechanical turk.

Using Google Suggest Manipulation for Profit

Source: http://www.flickr.com/photos/anilmohabir

Source: http://www.flickr.com/photos/anilmohabir

For lazy affiliates like me, (that’s what the Thinkvisibility Expert Panel call me!), this is probably a decent way to gear up search volume around a brand.

How? Well consider setting up manipulation around a range of words that users may get tempted to click through from Google suggest – scam is probably the best one. Craft a landing page optimized for the word, but pull a bait and switch in the copy.

  • User sees Brand + Scam
  • User Clicks Through
  • Content on page says that no, it’s not a scam!
  • Provide link to continue to brand
  • Profit!?

This can be geared in many, many other ways… I leave you to think about them independently.

Final Note

I have been berated that my subscribe button isnt big enough. So here you go. Subscribe.

Subscribe to my feed

Subscribe to my feed

Share and Enjoy:
  • Twitter
  • del.icio.us
  • Facebook
  • Google Bookmarks
  • FriendFeed
  • Sphinn
  • LinkedIn
  • PDF
  • StumbleUpon
  • Suggest to Techmeme via Twitter
  • Yahoo! Buzz

Rishi Lakhani is an independent Online Marketing Consultant specialising in SEO, PPC, Affiliate Marketing and Social Media. Explicitly.Me is his Blog. Google Profile

{ 4 trackbacks }

SMX – Search Marketing Expo: resumen en español del congreso sobre Google | Antonio González
March 10, 2011 at 9:10 am
What Really Impacts Google Suggest Suggestions? An Experiment - Technical SEO - State of Search
April 18, 2011 at 11:45 am
Reputation Management – Tactics that still work | | distilled
August 31, 2011 at 9:23 am
Monitoring Conversations: 10 Ways to Monitor Your Brand Online
January 8, 2012 at 7:40 pm

{ 23 comments… read them below or add one }

Richard Vaughan March 8, 2011 at 12:32 pm

Great post!

The theory of sheer volume of content impacting google suggest results in some very meaningful way definitely matches my experiences. That is why I’d currently suggest against hiding a rep problem with optimised junk content: You end up supporting the initial problem term and potentially making the situation worse. Obviosuly affiliates use this process to do the complete opposite: They WANT to rank for that oh-so-clickable “scam” term so they can make that sale!

Reply

Gordon Campbell March 8, 2011 at 12:44 pm

Great article, trying to work out how google instant works for a while, its interesting to see that it isn’t just search volume that influences suggestions.

Reply

Wouter Blom | Zoekmachine Marketing Consultant March 8, 2011 at 1:07 pm

NICE, I am trying to get a similar result: (on my own brand name)
http://bit.ly/h442Ez

Reply

Tyler March 8, 2011 at 1:32 pm

Curious. Are the ‘tests’ you ran exclusive to Twitter or did you send automated search queries to Google as well? Is there some part of the test you can’t reveal? This would be great to finally understand better for reputation management, and with some fun spare time, boost a keyword phrase maybe?

Reply

rishil March 8, 2011 at 1:51 pm

Nothing hidden at all. Purely twitter driven and no search pushed on to it.

Reply

Richard Vaughan March 8, 2011 at 1:58 pm

This should work with any type of content but with more meaty terms you need real changes in volume to see the affects. For promotional “manipulation” there are routes (use you imagination) for negative rep managemnet its about reducing the volume of content for that term: A lawyer can be a great tool in this regard ;)

Reply

Tyler March 8, 2011 at 3:11 pm

The train of thought I’ve been on regarding this is that if we can exercise some kind influence over what appears for Google’s suggestions, then we can potentially force out the negative keyword phrases so that people don’t get startled and look for our clients/businesses on ripoff report on their first search. The ‘imagination’ side is that if we really can influence what appears, then we can also ‘leech’ off of related keyword phrases to drive visitors to the ones we appear 1st for. That’s pretty cool, definitely manipulative, and not the kind of thing Google would like to happen, but I think it’s a pretty cool concept and worth more testing. Thanks for sharing your results rishil, it’s got me thinking…

Reply

Monica March 8, 2011 at 4:56 pm

Interesting! How Google suggest results come about is something I’ve always wondered about. Definitely a good find in analysing it in the Labs stage too.

Reply

Josh March 8, 2011 at 6:03 pm

How many retweets did you average in order to get the results to show up for your name?

Reply

rishil March 8, 2011 at 6:42 pm

To be honest, like I said, it was a lazy test. I just slammed twitter with funny Tweets that got loads of RTs

Reply

Trontastic March 8, 2011 at 7:23 pm

so THAT is what you were doing. For a second there I had thought you just lost your mind.

Reply

Sabine March 8, 2011 at 10:39 pm

This post is not a moment too soon. I just noticed a misspelled branded term thats suddenly coming in a lot, also thanks to suggest. How the hell it ended up there in the past 2 weeks beats me. Luckily the traffic converts just the same but it’s annoying. What other factors could influence this Rishi, you think anchortext might get weighed? I ask this because suggest shows an old brand name we havent used in a decade but is still top of mind an often used anchortext for inbound.

Reply

Shane March 10, 2011 at 2:15 am

I wa in the session at SMX West when this came up and it was pretty…uh…lively as the Google engineer was challenged with what in the world shoul people do when Google suggests negative, irrelevant, or even illegal terms.

The best method I’ve seen to “manage” Google’s suggestions was posted over at SEOmoz: http://www.seomoz.org/blog/our-online-reputation-management-playbook

Definitely a frustrating challenge for many website/business owners.

Reply

Tyson Bailey March 16, 2011 at 7:37 pm

Thanks. I am going to test this out this week. I will try and post the results

Reply

Turrist April 2, 2011 at 11:38 am

How long and how many false google search updated
its suggest inserting your “key”?

Reply

rishil April 2, 2011 at 11:49 am

I didn’t actually track that :)

Reply

XAVA April 28, 2011 at 12:38 pm

Hi Rishil! Great article! I am going to test it! I will keep you updated how it worked out and what I needed to do! Can you tell, how much content you needed to produce? Did you only youse Twitter? Thanks again!

Reply

Kristine S May 10, 2011 at 12:11 pm

Thank you for the info.. I would like to add one thing. Keep one thing in mind that this test did not. I did change results PURELY on volume as other have BUT where no matched SERP result exists the result took up to 3 weeks to appear. During that time Google created a SERP to somewhat match the query (though no real result did). IE if you search for Kristine (my last name) loves snow — you will find that it exists in Instant WITH NO exact page match in the SERPS. In fact Google rewrote descriptions to try to make a match.

Also note the volume needed to change the SERP was low – so some of the tests you have run may have simply not had time to materialize yet. I would offer the suggestion that you follow them for the next month or so and see if you have any additional changes not shown yet.

But thanks for the great post as always!

Reply

Michael Hayes October 11, 2011 at 1:59 pm

Interesting. Worth noting that the google suggestions are also heavily personalised – it will suggest navigational searches to sites after you’ve visited them, for example.

Which makes it more complicated to tests like this, as you have to remember to turn personalisation off (logging out doesn’t do the job because they still have cookies they can use to link you to some of your historical data).

Reply

Chaz December 26, 2011 at 7:31 pm

Really enjoy the read! Have you ever thought about creating a tool/bot that could do this for you? I do public relations here in Los Angeles and this would be absoluty amazing to tap into. Please if you ever create something related to a tool/bot, I would be so interested.

Reply

Olivia and Will Inc. December 27, 2011 at 8:35 am

Interesting. I am also trying to do this. Quite challenging for me. I am very excited to experiment this.

Thanks for sharing nice post.

Reply

Takeshi February 7, 2012 at 6:57 am

RSS Subscribe button needs to be bigger.

Reply

Mike February 23, 2012 at 6:39 pm

Great to see some interesting case studies.

Rishi, would you mind elaborating a bit more on the twitter strategy? You said you just posted a bunch of tweets with your keyword “ha ha ha”, is that correct? The RTs were from your original tweet or do you mean they were retweeted from the originator (not being yourself)? Would love to understand this better as my company is having a really hard time with suggested.

So far i’ve run bots and we have a program setup with our customers that is a new experiment, but would love to hear your actual strategy in more detail if you’re willing.

Reply

Leave a Comment

Previous post: Google Hints at Using User Signals via their Spam Extension.

Next post: Beat That Quote, Car insurance Aggregator!