What is Google Autocomplete?
Google Autocomplete is the drop-down of predictions that appears under the search box. Type a few letters and Google offers the most likely completions, drawn from real searches, weighted by popularity, freshness, your language and location, and, if you are signed in, your own history. Google calls them predictions rather than suggestions, because the point is to guess what you were already going to type.
It exists to save keystrokes, but for anyone who works in search it is a window into demand. The predictions are what people actually ask, in their own words, ranked by how many of them are asking.
Why it matters
Why Google Autocomplete matters
Keyword tools report volume for queries you thought of. Autocomplete shows you the ones you did not. Typing a brand, a product or a topic followed by a space, then a letter, surfaces the questions, comparisons and complaints attached to it: the brand plus review, plus pricing, plus versus a competitor, plus a word nobody in the company wants to see. That last one is why autocomplete is also a reputation issue; a prediction that pairs a brand with a problem is seen by everyone who types the name.
It also shapes behavior. People click predictions, so the predictions decide which queries grow. A phrase that appears in autocomplete becomes a more common search by virtue of appearing, which makes the list worth understanding even if you never open a keyword tool.
In practice
How Google Autocomplete works in practice
- 01
For research, work the box by hand or with a tool that collects it. Type the seed term and cycle through the alphabet after it, try question words before it (how, what, why, is), and try it with versus, near, for and best. Do it in a private window with the location set to the market you care about, because signed-in history and geography both change the list. Collect the results, group them by intent, and map them to pages that answer them.
- 02
For reputation, search your own brand the same way and look at what completes. Google removes predictions that violate its policies (hateful, violent, dangerous or sexually explicit content, and certain predictions about named people), and you can report them, but it will not remove a prediction simply because it is unflattering. The lasting fix is the one that works everywhere: give the query a better answer. A brand-plus-complaint prediction loses its force when the first result is your own honest page about it.
Watch out
What people get wrong about Google Autocomplete
Treating predictions as volume data. They are ranked, not counted, and a prediction can appear because it is rising fast rather than because it is large. The second mistake is trying to game it. Services that promise to insert a phrase into autocomplete by running thousands of searches are selling a violation of Google's policies that is filtered out quickly and can bring attention you do not want. The third is reading the list as everyone's. Your predictions reflect your history and location; what the public sees needs a clean browser and the right market.
