Google Autocomplete Repair
What is autocomplete?
Autocomplete predicts what you may search for when performing a query on Google and offers suggestions to save users time. Google uses many factors to try to predict queries, including how the queries are trending on Google, your location, the language, and past searches you may have performed. We have found that the most relevant signal (except language) is trending queries. In other words, how often have people searched for various permutations of the query being searched?
The problem with autocomplete for brands
Sometimes Google instant predictions work against people and companies. Can they be changed? The answer is “sometimes.”
People suffer when Google autocomplete shows trending search terms that have a negative aspect to them. You’ve probably experienced this yourself – how many times have you begun to Google something, only to find the suggestions that pop up make a person or company look terrible? This often changes your mind, so you select the predicted search. When people see something negative in autocomplete they hadn’t thought of, and it’s negative, they often click on that query instead. When they do it can send a signal to Google that the search query is relevant, which in turn can cause the query to be shown more often.
To brands this can seem like a neverending downward spiral of negativity.
Negativity bias and autocomplete
People are often negative. We’re just wired that way. Not evil; just conditioned over time to look for something bad. When we see something negative, our senses ramp up. It’s a survival mechanism, but it also skews Google search results and autocomplete for with a waterfall of consequences once the new query is selected. You can read more about human negativity bias here.
Crowdsourcing
Rather than use one method to change autocomplete, we use many. This article outlines our approach. Before we move on, let’s be clear on what crowd-sourcing is.
Crowdsourcing refers to the process in which a business, individual, or website owner (the crowd-sourced) uses people (the crowd) to complete a particular task. These people may be professionals or ordinary people, paid or volunteering.
The platform will announce a project or problem online and invite a crowd to participate by getting involved, giving feedback, or offering a solution. Depending on the circumstance, the crowd will be rewarded, either by payment or recognition.
A very simple example of this would be Ben and Jerry’s Ice Cream and their “Do the World a Flavor” campaign, where they asked customers to come up with all new and original flavor ideas. It ran in 17 countries and had a massive amount of entries. Today, you can stop by Ben and Jerry’s and try a scoop of “Cherry Garcia” or “Chubby Hubby,” a direct result of the crowdsourcing method.
Changing autocomplete
In the past, reputation agencies could alter Google autocomplete using IP proxies and many people typing in search terms. Today, changing bad autocomplete results still use those methods, though less effectively, because the traffic used by automated platforms can seem (and be) non-organic.
The best method forward today is to use small private crowd-sourcing networks. Because these networks are not public, the activities of their participants leave a much smaller “footprint”. Private networks have a much higher success rate than public networks or traffic bots (even human-based traffic on a popular network). We have found that in most cases, this method works best for queries that do not have a high search volume. For example, changing autocomplete for “Apple Computer” would be impossible. But changing it for a local business has a much higher probability of success.
Reputation X has performed tests between public and private automated networks. We found that even if the participants in a public crowd-sourcing project are asked to use country-specific IP addresses (normally via proxies), they don’t actually do it effectively. In fact, in one test where the participants were asked only to use North American IP addresses, we found that 80% of the traffic came from IP addresses in the Philippines and India.
Private human networks are better
Private networks are more expensive, but of higher quality. Some people call these workers “mommy bloggers”. They are located throughout the target country, mainly the United States and Canada. They’re North American, so they’re paid more. The costs can add up, but we’ve found it works better than anything else.
Augmenting people with bots
We’ve also tested the use of bots that automatically perform searches as a human would. Some even use random duration and random proxy use. But, like public crowdsourcing, these leave a heavy footprint due to their overuse by low-end SEO firms. We believe these networks have been flagged. They do tend to work to some extent if the volume of searches matches the “normal” search volume for a given search phrase. However, they are not the answer by themselves.
Use multiple approaches
The best way to alter Google autocomplete search results, in our opinion, is to use a combination of private crowdsourcing networks very lightly augmented by bots and outsourced public crowd-sourcing (albeit at a very low level).
Of course, there is a lot more that goes into it, but we hope experienced search engine marketers find this information helpful in their endeavors to bring better outcomes for their online reputation management clients.
