Google Autocomplete works by predicting the rest of your search while you type, based on real searches that people have already done on Google. It used to be called Google Suggest: Google launched it in Labs in 2004 and announced the new name in October 2010, as part of its launch of Google Instant. It is one of the features covered in our guide to Google SERP features.
What Google Autocomplete is
Google’s help page on how autocomplete predictions work describes it as “a feature within Google Search that makes it faster to complete searches that you start to type.”
In a 2018 post on Google’s blog, Danny Sullivan, Google’s Public Liaison for Search, listed where it appears: the Google home page, the Google app for iOS and Android, the quick search box in Android and the address bar in Chrome. The same post gave some numbers:
- “On average, it reduces typing by about 25 percent”.
- “When you’re using Google on desktop, you’ll typically see up to 10 predictions. On a mobile device, you’ll typically see up to five, as there’s less screen space.”
Why it used to be called Google Suggest
Google announced the name change in its “This week in search” post of 16 October 2010: “We first launched Google Suggest in 2004 in Labs to help people enter their searches faster.” The post said the feature had not gone anywhere. Google had renamed it Google Autocomplete because the new name fit better with Google Instant and its “automatic queries and automatic results.” Two pages in the current top 10 for this search, from Mangools and SEOPress, still use the name Google Suggest for the same feature.
How are Google’s predictions generated?
Google’s help page says predictions “reflect real searches that have been done on Google.” To decide which ones to show, Google’s systems “look for common queries that match what someone starts to enter into the search box” and then also consider four things:
- The language of the query
- The location a query is coming from
- Trending interest in a query
- Your past searches
Google says these factors let autocomplete show predictions “unique to a particular location or time, such as for breaking news events.” Beyond whole searches, it can also finish single words and phrases, based on real searches and on “word patterns found across the web.”
Predictions, not suggestions
Google chooses its words here on purpose. Sullivan’s post says Google calls them “predictions” rather than “suggestions” because autocomplete “is designed to help people complete a search they were intending to do, not to suggest new types of searches to be performed.”
That is also why autocomplete is not a popularity chart. Google’s help page says it “doesn’t simply display the most common queries on a given topic,” and that it “shouldn’t be compared against Google Trends,” which is the tool Google offers for researching how popular searches are over time.
What shapes the predictions you see
Two people typing the same letters can see different lists.
| Factor | What Google says | What it means for you |
|---|---|---|
| Language | “The language of the query” is one of the four factors | The same topic can produce different predictions in another language. |
| Location | “The location a query is coming from” | Searches from another city or country can show different completions. |
| Trending interest | Lets predictions reflect a particular time, “such as for breaking news events” | Predictions shift when something is in the news, then settle again. |
| Past searches | Signed-in users may see “a history of their past searches” and “personalized predictions based on their activity on Google” | Your own list is partly about you, not just everyone else. |
Why some predictions are removed
Autocomplete is filtered. Google’s help page says its systems try to catch predictions “that are violent, sexually explicit, hateful, disparaging, or dangerous, or which lead to such content.” That includes predictions “unlikely to return much reliable content, such as unconfirmed rumors after a news event.” When the automated systems miss something, Google says its enforcement teams remove the prediction and “closely-related variations.”
The policies are a section of that same help page rather than a separate document. They cover the general Search content policies (dangerous, harassing, hateful, sexually explicit and terrorist content, violence and gore, and vulgar language) plus four rules specific to autocomplete:
- Elections. No predictions that read as a position for or against a political figure or party, or a claim about taking part in or the integrity of the electoral process.
- Health. No predictions about “potentially medically hazardous health claims.”
- Named individuals. No predictions that tie disparaging or sensitive terms to a named person.
- Serious malevolent acts. No predictions that accuse people or groups of serious wrongdoing without well-established or expert evidence.
Sullivan’s post adds that Google may also remove predictions that are spam, closely tied to piracy, or the subject of valid legal requests, and sums up the principle: “autocomplete should not shock users with unexpected or unwanted predictions.”
Two things follow. A missing prediction does not mean nobody searches for it: the post says popular searches in Google Trends “might not appear as predictions within autocomplete.” And filtering never stops a search. The help page says autocomplete “doesn’t prevent anyone from completely typing out a search query to get results.”
How to report a prediction
If you see a prediction you think breaks the policies, Google’s page on managing autocomplete predictions gives these steps on a computer:
- Go to google.com and type your search.
- Under the predictions, click Report inappropriate predictions.
- Choose the prediction and what your feedback is about.
- Give your feedback and click Submit.
On Android, the Android version of that page says to touch and hold a prediction, then tap Report this. Sullivan’s 2018 post said iOS users of the Google app could swipe left on a prediction to report it. Google is clear that a report is not a guarantee: it “looks at what you said, but might not remove the prediction.”
How to turn off trending and personal predictions
Google’s page on managing predictions says “You can turn off some of these predictions,” and covers trending searches and personal ones. One caveat first: the same page says Google is updating its Search settings, so what you see “may not match your current experience.”
Trending searches
- On your computer, go to google.com.
- At the bottom, click Settings, then Search settings.
- On the left, click Other settings.
- Under “Desktop,” turn off Autocomplete with trending searches.
Predictions based on your history
Google says that if you turn off Search personalization, it “doesn’t predict things based on what you’ve searched for before.” If you are signed in, you can also turn off Personalized Recommendations.
How do I clear Google Autocomplete?
You can remove your own past searches from the list. Sullivan’s 2018 post described how it worked then: on desktop, click “Remove” next to a past search; on mobile, tap the X. You could also delete past searches in bulk, by date or by term, in My Activity in your Google Account. The layout may have changed since, but Google’s page on managing predictions still links to its guide to finding and erasing your Search history.
Using Google Autocomplete for keyword research
Because predictions come from real searches, autocomplete is a quick, free way to see how people word a topic. Some keyword tools still label this data Google Suggest, but you can do the basic methods by hand in the search box.
- Start with seed wording. Type your topic slowly and note each prediction. These are the common ways people begin a search on it.
- Work through the alphabet. Type your topic followed by a space and “a”, then “b”, and so on. Each letter pulls out a different set of completions.
- Add question words. Put “how”, “why”, “can”, “is” or “what” before the topic to find the questions people type. Our guide to question keywords covers what to do with them.
- Research signed out. Google says past searches and activity shape what signed-in users see. Sign out, or use a fresh browser profile, and keep the same location each time so your results are comparable.
- Check demand somewhere else. Autocomplete shows no search volume, and Google says it is not a list of the most common queries. Use Google Trends to compare phrasings and a keyword tool or Search Console for numbers.
After you search, people also search for and related searches show other phrasings Google connects with your topic. For a full workflow that starts with the questions on the results page, see how to use People Also Ask for keyword research.
Autocomplete and People Also Ask
Autocomplete predicts the search you are typing, before you press enter, while People Also Ask shows related questions after you search, each with an answer. We compare the two side by side in what is the difference between Google Autocomplete and People Also Ask?
If you want those questions in a spreadsheet, PAA Box is our free Chrome extension: it opens the People Also Ask questions on the results page and exports them to CSV. It does not read autocomplete.
Not the same as address autocomplete or autofill
The search shows two other products with similar names. The address autocomplete that developers add to their own sites comes from a separate service, Autocomplete (New) in the Places API, which Google describes as “a web service that returns place predictions and query predictions in response to an HTTP request.” Autofill is a Chrome feature that can fill out forms with saved info like your name, addresses and payment details. Neither is the prediction list in Google Search.