April 17, 2023

Local Memo: Google Building New AI-Based Search Engine

Damian Rollison

Local Memo: Google Building New AI-Based Search Engine

In this week’s update, learn about Google’s new AI-based search engine; more of Google’s AI initiatives including image generation and a Chrome chatbot; Twitter’s launch of paid verification for businesses; Montana’s vote to ban TikTok; Google’s throttling of Local Packs; and the success story of SpamBrain.

Google Building New AI-Based Search Engine

According to the New York Times, Google is working to build a brand-new search engine based on its AI technology. Separately, the company has launched an initiative called Project Magi that will bring AI features to its existing search platform. The new search experience Google is working on will be extremely personalized, anticipating the needs of individual users by providing suggestions in a conversational format. Project Magi, expected to launch sometime next month — much sooner than the fully revamped search engine, which has no launch timeframe — has 160 Google employees devoted to its development. It will incorporate ads and will have the ability to write computer code. About one million U.S. users will initially be given access, ramping up to 30 million during the year. Another set of feature releases is planned for fall.

Google to Launch Image Generation, Chrome Chatbot

Additional AI features are also in the works, according to the Times article, including a tool that would use AI and Google Earth to help users explore music around the world; an image generation tool called GIFI that will be integrated with Google Images; a language instruction chatbot called Tivoli Tutor; and Searchalong, a tool that lets users ask a chatbot questions while using the Chrome browser to surf the internet. A local search use case was mentioned for the Searchalong tool: users looking to book a stay via Airbnb could use the chatbot to search for things to do in the area.

Brands Can Now Sign Up for Twitter Verified Accounts

Twitter’s “Verification for Organizations” service is now live for businesses globally. Businesses that sign up for the service will get to display a gold checkmark on the platform. Profile photos will show in a square frame instead of the usual circle. Businesses will be able to authorize accounts run by employees, whose affiliation will be shown by a blue checkmark and a brand icon next to their username. Staff will be listed in a tab in the brand profile. Brands will also reportedly receive priority support from the company. Applications will be reviewed by Twitter before approval. The cost is $1,000 per month plus an extra $50 for each verified staff member. Twitter's top 500 advertisers, as well as the 10,000 most followed business accounts on the platform, are being given the new verified status for free. It’s unclear whether the new parent-child structure for Twitter accounts will support a multi-location business model. Twitter has also announced that Twitter Blue subscribers can now post tweets of up to 10,000 characters. Subscribers had previously been permitted to post up to 4,000 characters, as compared to 280 characters for ordinary users. Finally, Twitter will remove all legacy (non-paid) blue checkmarks from user accounts by April 20.

Montana Votes to Ban TikTok

Montana has become the first U.S. state to vote to ban TikTok. The bill that would ban the app passed in the Montana legislature and has been sent to Governor Greg Gianforte for signature. The ban would go into effect on January 1, 2024. A $10,000 fine for each violation would be levied against either TikTok owner ByteDance or the company from whose app store the app was downloaded. Governor Gianforte has been an outspoken critic of TikTok in the past, though it is unknown whether he will sign the bill. Legal challenges are likely if the bill becomes law.

Is Google Throttling Local Packs?

After reports last year that Local Packs were becoming less frequent in Google search results, recent studies have shown that their incidence may have dropped even further, with a new Moz report showing that Local Packs only appeared for about 18% of mobile results and 16% of results on desktop. Studies vary, but Moz reported last October that Local Pack incidence had dropped from 40% to about 24% overall. Greg Sterling speculates that Google may be choosing not to show Local Packs for some local queries, deliberately throttling this result type in order to avoid the appearance of “self-preferencing,” due to ongoing antitrust investigations against the company.

The SpamBrain Success Story

Google’s SpamBrain, the machine learning model designed to ferret out spam content from search results, has had a banner year according to reports from the company, catching 5 times more spam sites in 2022 than it did the previous year, and 200 times more than when the model first launched in 2018. SpamBrain detects several spam types including abusive links and hacked content, and does its work while Google crawls the web, preventing spam content from making its way into Google’s search index.