In a recently published blog post, Google’s Danny Sullivan highlighted how autocomplete predictions are generated via the search engine’s algorithm.
Those who use Google Search has definitely experienced the autocomplete predictions. Google has also explained how they work in previous blog posts.
However, the recent publication is designed to make the user understand how the predictions are generated. Sullivan also touches on the scenario when predictions are not appropriate and how Google counteracts with such situations.
Typically, the algorithms look at 4 factors:
· Trending searches
· Language of the searcher
· The location of the searcher
· Freshness
Initially, Google considers trending queries when generating autocomplete predictions. However, it also combines the searchers' location and language when predicting queries. It may prioritize freshness as well if there is a ‘sudden’ interest in a particular topic.
Google handles such issues in two ways: Automatically or manually.
The automatic system is designed to prevent unhelpful and policy-violating predictions from appearing in the search results. At the same time, Google has a manual system or enforcement teams in place that takes over when the automated protocol doesn’t catch predictions that breach its policies.
Autocomplete is programmed not to provide suggestions that include violent, sexually-explicit, hateful, and dangerous content. It will also not make suggestions if a prediction returns unreliable content.
Unconfirmed content or rumors are not generated via predictions as well.
Those who use Google Search has definitely experienced the autocomplete predictions. Google has also explained how they work in previous blog posts.
However, the recent publication is designed to make the user understand how the predictions are generated. Sullivan also touches on the scenario when predictions are not appropriate and how Google counteracts with such situations.
How Are Autocomplete Predictions Generated?
Autocomplete predictions are automatically generated via real searches that people have conducted on Google.Typically, the algorithms look at 4 factors:
· Trending searches
· Language of the searcher
· The location of the searcher
· Freshness
Initially, Google considers trending queries when generating autocomplete predictions. However, it also combines the searchers' location and language when predicting queries. It may prioritize freshness as well if there is a ‘sudden’ interest in a particular topic.
When Autocomplete Will Not Provide Suggestions
Google has an alternative protocol in place for avoiding certain types of predictions. This includes queries that are unexpected, shocking, or less likely to generate reliable content.Google handles such issues in two ways: Automatically or manually.
The automatic system is designed to prevent unhelpful and policy-violating predictions from appearing in the search results. At the same time, Google has a manual system or enforcement teams in place that takes over when the automated protocol doesn’t catch predictions that breach its policies.
Autocomplete is programmed not to provide suggestions that include violent, sexually-explicit, hateful, and dangerous content. It will also not make suggestions if a prediction returns unreliable content.
Unconfirmed content or rumors are not generated via predictions as well.
Users Can Still Search
But despite of the prediction algorithms, the users are free to search for anything they want to – says Sullivan. As Google prevents autocomplete from predicting search results, but it will not prevent actual search results from appearing on the page.
