How to Remove Google Autocomplete Suggestions (and When You Can’t)

You can ask Google to remove autocomplete suggestions that break its autocomplete policies, such as predictions that are sexually explicit, hateful, violent, dangerous, harassing, or that attach disparaging terms to a named person. Use the “Report inappropriate predictions” link under the predictions on desktop, or press and hold a prediction in the Google app. Predictions that are negative but reflect what people genuinely search, like “brand name lawsuit”, usually stay, and the only reliable way to change them is to change what people search for over time.

That distinction matters, because a lot of money gets wasted on services that promise to “delete” autocomplete terms Google has no reason to remove. This guide covers what qualifies, how to report it, and what to do about the rest.

How Google autocomplete works

Google calls these “predictions”, not suggestions, and the difference is deliberate. According to Google’s Search Help, predictions are based on real searches people have done on Google, along with factors such as the language and location of the search, trending interest in a query, and, for signed-in users, their own past searches.

Google doesn’t publish the exact mechanics, and the predictions you see can differ from what someone in another city or on another device sees. What’s clear is that a prediction generally appears because enough people are searching that phrase, not because Google has formed an opinion about you.

That’s why a single bad news story can put “your name + arrested” or “brand + scam” into autocomplete within days. Curiosity spikes, people search, and the prediction follows the searches.

Which predictions Google will remove

Google’s published autocomplete policies describe the kinds of predictions its systems try to block and that it removes when reported. In general terms, they include predictions that are:

• Sexually explicit, vulgar or profane, outside medical or scientific contexts.
• Hateful toward people or groups based on characteristics such as race, religion, gender or other protected attributes.
• Violent or dangerous, including predictions that promote harmful activities or terrorism.
• Harassing toward individuals, including disparaging or bullying terms attached to a named person, and predictions that expose sensitive personal information.
• Serious unproven accusations. Google’s policies also cover predictions that accuse people of serious malevolent acts without supporting evidence.
• Election-related predictions, such as those that could be read as claims for or against a political figure or about the integrity of voting.
• Medically hazardous health claims.
• Spam or piracy-related, which Google has said it may also remove.

Google also provides a separate legal route for predictions you believe are unlawful, for example defamatory. If you’re considering that route, talk to a lawyer about whether the prediction is actually unlawful where you are before you file.

How to report a Google autocomplete prediction

1. Take screenshots first. Capture the prediction with the date, on desktop and mobile if possible. Predictions change, and you’ll want a record.
2. On desktop: go to google.com, type the start of the search until the prediction appears, then click “Report inappropriate predictions” below the list.
3. In the Google app or mobile browser: type the search, then press and hold the prediction and choose the option to report it.
4. Select the prediction and the reason. Pick the category that fits (for example hateful, sexually explicit, or harassing) and add a short, factual explanation.
5. Submit and check back. Google says it reviews reports but may not remove the prediction. It doesn’t usually tell you the outcome, so search again in a week or two from a private window.

Report once, accurately. Filing the same report repeatedly, or recruiting many people to report a prediction that doesn’t break policy, doesn’t change how Google evaluates it.

What won’t be removed, and why

Most negative predictions businesses and professionals worry about are accurate descriptions of what people search. Common examples:

• “[brand] reviews” or “[brand] complaints”, which are ordinary consumer research.
• “[brand] lawsuit” when there really was a lawsuit, even a dismissed one.
• “[name] fired” or “[name] resigns” after a real, reported event.
• “[brand] layoffs”, “[brand] recall” or “[brand] data breach” when those events happened.

Google treats autocomplete as a reflection of search interest. Unless a prediction falls into one of its policy categories, the fact that it’s unflattering isn’t a removal reason, in the same way that a genuine negative review isn’t a removal reason on Google Maps.

What legitimately changes predictions over time

If a prediction reflects search interest, the way to change it is to change the interest. That’s slower than a report and depends on real search interest shifting, but changes that happen this way tend to last.

Fix the underlying issue

If “brand + refund” shows up because refunds really are slow, fix refunds. As fewer people have the problem, fewer people search for it. No search tactic survives an ongoing real complaint.

Publish content people genuinely search for

Give people new, accurate reasons to search your name: a product launch, a useful free tool, a detailed guide, a new location, an event. When those searches grow, related predictions such as “brand + pricing”, “brand + app” or “brand + careers” compete for the limited slots in the list.

Earn coverage and attention

Press coverage, interviews, podcast appearances and partnerships drive real searches. For individuals, speaking engagements, published articles and a well-maintained professional presence do the same thing.

Address the search results as well

A negative prediction is only half the problem. Someone who clicks it lands on results, and those results decide what they conclude. Strengthening what ranks for that search, with accurate, high-quality content about you, is often where the real work is. Our search result suppression service focuses on exactly that.

A worked example

This is an illustrative scenario, not a real client. A regional HVAC company is named in a local news story about a subcontractor who was arrested for something unrelated to the company. Within a week, typing the company name brings up “company name arrest” as a prediction.

The owner reports it. The prediction doesn’t involve sexual, hateful or violent content, and the news story was real, so the report has limited grounds. It stays, at least at first.

So the company works on the searches instead. It asks the newspaper to add a clarification that the subcontractor was not an employee, which the paper does. It publishes a clear seasonal maintenance guide and announces a new service area, both promoted to its customer email list. It keeps asking every customer for a review, which drives people to search “company name reviews”. Over the following months, as interest in the arrest story fades and other searches grow, the prediction tends to drop down the list and eventually out of it. How long that takes depends on how much attention the original story got and how much new search interest the company creates.

Why manipulating autocomplete backfires

Some vendors offer to “flood” autocomplete with positive phrases using bots, click farms or paid searchers. Avoid them.

• It breaks Google’s policies. Google says it may remove predictions it determines to be spam, and artificial search activity is exactly what its systems are designed to spot.
• It usually doesn’t last. When the fake searches stop, or are filtered out, the original prediction tends to come back.
• It can make things worse. Obviously manufactured predictions draw attention, and if a journalist or competitor notices, the story becomes “company tried to game Google”.
• It does nothing for the results. Even if a prediction changes, anyone who searches the negative phrase still sees what ranks for it.

Be wary of anyone who guarantees autocomplete removal. They can’t control Google’s decisions, and the methods behind that promise are usually the ones above.

Autocomplete problems for individuals

For executives, professionals and public figures, a prediction attached to your name can follow you into every job interview and client meeting. The policy on disparaging terms and serious unproven accusations about named individuals gives people more grounds to report than businesses usually have, so check carefully whether a prediction fits. If it doesn’t, the same approach applies: address the cause where you can and build a stronger, accurate presence around your name. Our personal reputation management work often starts here.

Before acting, it helps to see the whole picture. Our guide on how to check your online reputation walks through the searches to run, and if you’re a public figure, getting a Google knowledge panel helps put accurate information at the top of your name search.

Frequently asked questions

Can I pay Google to remove an autocomplete suggestion?

No. Google doesn’t offer paid removals, and anyone claiming to sell one is selling something else. Removals happen only when a prediction breaks Google’s policies or, through the legal route, the law.

How long does Google take to review an autocomplete report?

Google doesn’t publish a timeframe and usually doesn’t confirm the outcome. Check the search again from a private window after a week or two. If the prediction is still there and clearly breaks a policy, you can report it again with a clearer explanation.

Why do I see a different prediction than my colleague?

Predictions vary by location, language and device, and signed-in users may see predictions influenced by their own past searches. Use a private window, and ask people in other places to check too.

Will a negative prediction go away on its own?

Sometimes. If it was driven by a short burst of interest, it often fades as searches drop. If the underlying issue continues, or people keep searching the phrase, it tends to stay until something else replaces that interest.