Google autocomplete is the list that drops down beneath the search box as you type: search-box autocomplete inside Google Search, not the address or Places autocomplete used in forms and maps. It is available anywhere you find a Google search box: the homepage, the Google app for iOS and Android, the quick search box on Android, and the address bar in Chrome.
Google is deliberate about what it calls that list.
Predictions, Not Suggestions
Google calls them predictions, not suggestions, because they are meant to finish a thought you already started, not to propose a new kind of search.
They are best guesses of the query you were likely to continue typing, drawn from real searches that happen on Google: common and trending queries that match the characters entered, your location, and your previous searches. That framing comes straight from Google's own explainer of how autocomplete works in Search.
The list updates with every keystroke. Type "san f" and you get San Francisco completions like "san francisco weather"; add one letter to "san fe" and those disappear in favor of San Fernando completions. The extra character made it clear the first set was no longer relevant.
What Shapes the Predictions
No single signal decides the list. Google's own account and independent technical explainers name the same inputs:
| Input | What it does |
|---|---|
| Popular searches | Millions of other people's real queries are the base: common matches for what you typed |
| Location and language | Results shift so local queries match where you are and what language you search in |
| Signed-in search history | Past searches you did can resurface as personalized predictions |
| Trending and current events | Newly spiking queries can appear when they match the prefix |
| Fast prefix matching | Behind the scenes the match runs through a trie, a retrieval tree built for instant prefix lookup |
| Safety filters | Automated systems plus human review block violent, hateful, or explicit predictions |
Display changes by surface. On desktop you typically see up to 10 predictions; on mobile, typically up to five, because there is less screen space. On mobile or Chrome on desktop, Google may show extra detail below a prediction (dates, local weather, sports information). In the Google app, some predictions carry small logos or images where Google holds structured Knowledge Graph information.
Past searches get their own treatment. On desktop a past search in the list carries a Remove label; on mobile a clock icon on the left and an X on the right clears it. You can also bulk-delete past searches by date or term in My Activity in your Google Account.
Google reports that autocomplete reduces typing by about 25 percent on average and saves over 200 years of typing time per day cumulatively. That is Google's own self-measurement; no independent source corroborates it.
What Gets Removed, and How to Report It
Autocomplete only shows common and trending matches, but some matches are deliberately withheld. Google removes predictions against its autocomplete policies:
- sexually explicit predictions unrelated to medical, scientific, or sex-education topics
- hateful predictions against groups and individuals
- violent predictions
- predictions about dangerous and harmful activity
It may also remove predictions judged to be spam, closely associated with piracy, or subject to a valid legal request. The guiding principle Google states is that autocomplete should not shock users with unexpected predictions. That is also why a topic that is popular in Google Trends (a tool built for deliberately researching topic popularity) may still never appear as a prediction.
If you spot one that slipped through, you can report it with the Report inappropriate predictions link below the desktop search box, by long-pressing a prediction on mobile or the Android app, or by swiping left on iOS. A separate form handles legal complaints. Google reviews reports but may not remove the prediction, and when it does act it expands the fix to closely related predictions.
How to manage and report these behaviors is documented in Google's Search help on autocomplete predictions and in the same behind-the-scenes explainer.
Why a Suggestion Showing up Means Demand
For keyword work, the useful read is simple: if a term autocompletes, people actually search it.
Cloro's keyword-research guide puts it plainly: autocomplete returns suggestions ranked by real query frequency, and Google only suggests queries with real frequency behind them. Presence is the signal. A term that autocompletes has demand behind it.
That is why discovery workflows start here. In one programmatic-SEO system a practitioner described, step one of the whole operation is scrape Google autocomplete plus People also ask, then export thousands of long-tail variants from a suite tool, spot repeatable patterns like "best {tool} for {industry}", and group them into scalable page templates: 3,000-plus variations for one SaaS client in that account. Autocomplete is the fuel; the patterns are what get built on. The public catalog on qequals is that grouping done in the open on tracked terms: themes at crawl time, the raw list a click away.
One caution keeps the signal honest. Presence tells you a query exists and is surfaced; it does not hand you a monthly search count, competition score, or CPC. Those numbers live in volume tools built on modeled averages, not in the prediction list, which is also why a niche phrase can autocomplete and still show no search data.
If a phrase autocompletes, people are typing it now. Start there, then go find out how many.
