Perplexity AI, a search artificial intelligence startup, has seen its monthly revenue and usage increase sevenfold because the starting of the 12 months after closing a brand new $250 million funding round.
The AI-powered search engine answered about 250 million questions last month, compared with 500 million queries in all of 2023, Dmitry Shevelenko, Perplexity's chief business officer, told the Financial Times.
The recent figures underscore Perplexity's position as one in every of the fastest-growing generative AI applications since OpenAI's successful launch of ChatGPT in November 2022, despite controversy over the startup's data collection techniques.
San Francisco-based Perplexity, founded just three months before ChatGPT's launch by former Google intern Aravind Srinivas, uses AI software to reply questions using information retrieved in “real time” from the web, including news web sites.
Perplexity began the 12 months with annual revenue of $5 million – an annual revenue projection based on last month's sales – and is now generating greater than $35 million on that basis, an organization insider says.
Now the startup is shifting its business model from subscriptions to promoting, putting it in stronger competition with Google, which dominates the $300 billion search promoting market.
Its growth comes as Google pushes the mixing of AI capabilities into its core search product and launches OpenAI SearchGPT, a prototype AI search tool available to about 10,000 testers.
“Ultimately, the smaller player on this space has two benefits: speed and focus,” said Shevelenko. “Our users and our team consider Perplexity as only one thing: a spot to get answers to their questions. The competition sharpens our focus much more.”
To boost its fight against larger rivals, Perplexity recently closed a brand new $250 million investment, including from SoftBank's Vision Fund 2, people conversant in the deal said, tripling its valuation to $3 billion from $1 billion in April. Bloomberg previously reported on the financing talks.
Existing investors include AI chip maker Nvidia and Amazon founder Jeff Bezos, in addition to several well-known names from the AI industry, comparable to OpenAI co-founder Andrej Karpathy and Meta's chief AI scientist Yann LeCun.
Shevelenko said Perplexity has not been discouraged by competition from better-equipped technology firms, including Microsoft-backed OpenAI, which makes the world's hottest AI chatbot.
“OpenAI does so many various things… They don't concentrate on answering people's questions with high-quality sources,” Shevelenko said. “That's why SearchGPT's reviews show that it doesn't compare well to Perplexity.”
Perplexity has up to now generated most of its revenue from consumer and company subscriptions, however the startup recently announced that it can introduce promoting on its platform by the top of next month.
“Unlike OpenAI, we at all times knew that promoting can be our essential monetization engine,” Shevelenko said.
The company will share a “double-digit” percentage of its revenue from each sponsored article with the news publishers mentioned, Shevelenko said. The company has signed deals with Time, Der Spiegel and Fortune, amongst others.
Before Perplexity announced its latest partnerships with publishers in June, accused of plagiarism by the publications “Forbes” and “Wired”, which criticized the start-up’s reproduction of articles without clear citation of sources and its scraping of internet sites that its crawlers had explicitly blocked.
Shevelenko confirmed the allegations and said the corporate has taken note of the criticism. Perplexity has responded by making changes to its interface to spotlight quotes more clearly and taking steps to make sure its responses don’t summarize web sites.
He said within the two weeks since launch, 50 publishers have asked to affix Perplexity's revenue-sharing program, and the corporate hopes to incorporate as wide a pool of websites as possible.
“For Perplexity to be a useful product on the open web, there must be good business models for publishing recent and updated facts in regards to the world,” Shevelenko said. “If you need to incentivize (journalism) over the long run, revenue sharing is a simpler way than one-time lump sum payments, as OpenAI has done.”
Unlike Google and OpenAI, Perplexity doesn’t develop its AI models itself, which has develop into increasingly costly. Instead, the corporate licenses a mix of AI systems from OpenAI and others.
Perplexity's search engine was originally based on a licensed version of Microsoft's Bing index, like many potential Google competitors, but Shevelenko said Bing is not any longer used as a core system.
“We have our own search index and our own rating system,” Shevelenko said. “We use signals from every kind of search engines like google and yahoo, but we’ve our own crawler and our own rating system.”
The platform is targeting journalism and academia due to vast amounts of reliable information and data they’ve. A one who worked for Perplexity earlier this 12 months said they see this source material as a bonus over traditional search engines like google and yahoo like Google, which draw on a much wider range of internet sites.
“Garbage in and garbage out is an issue that plagues firms, so you should use a greater diversity of sources when training models,” this person said.
However, this person said that the introduction of promoting could scare users away: “When you see promoting, it creates an untrustworthy environment. People at the moment are a bit skeptical in regards to the results on Google.”
Google, which a US judge this week classified as a “monopolist” in a landmark antitrust case, has fended off quite a few attempts to challenge its dominant position within the search space over the past twenty years.
Nevertheless, Joseph Teasdale, chief technology officer at consulting firm Enders Analysis, said the marketplace for AI search engines like google and yahoo is getting “hot.”
“The risk of AI is that general web search as an entire will develop into obsolete through recent ways of connecting users with information, services and products,” he said. But the “big unknown” is whether or not it could possibly be reliable enough for general use.
“AI is stubbornly vulnerable to confabulation,” said Teasdale. “With billions of queries per day, serious failures are inevitable.”
