The rise of generative AI chatbots has ushered in powerful new tools to help us get work done more efficiently. The technology has also introduced a slew of new terms into our lexicon, such as corpus and prompt engineer. But recently, another term has been popping up: “answer engine.”

But just what is an answer engine, and how is it different from a search engine? Here’s what you need to know.

Search engine vs. answer engine

Everyone knows what a search engine is: it’s Google. Or Bing. Or DuckDuckGo. In other words, it’s an online website where you enter a search query (“What are the best Japanese cities to visit?”) and you get a list of results in the form of web links to external websites.

The search engine’s algorithm thinks that the results of these external sources of information are most likely the best places for you to find the answer to your query.

But an answer engine is different. When you enter your query into an answer engine, it gives you the answer to your question right there, in its results. So instead of suggesting external websites that may contain the information to answer your query about the best Japanese cities to visit, the answer engine will just tell you: Tokyo, Kyoto, Hiroshima, Nara, and Osaka.

What are some examples of answer engines?

Many people assume that an “answer engine” has to be a generative AI chatbot—an AI tool that allows you to you ask it what you’d like to know, and it spits out the answer. But answer engines aren’t necessarily required to be powered by artificial intelligence. Rather, they could source their information, not through the magic of large language models but by simply pulling the answer from a large database or via reliable and tested calculations.

A great example of this type of “classic” answer engine is WolframAlpha. Founded in 2009, it was one of the first, and most well-respected, mainstream answer engines. Its data sets and computational knowledge allowed users to get direct answers to their questions in fields ranging from mathematics to chemistry to history. WolframAlpha is generally considered so reliable, in fact, that its answers were directly integrated into early versions of Apple’s Siri (and unlike most Siri answers, the queries answered using WolframAlpha were virtually always accurate).

WolframAlpha is still in use today and is especially beloved by the scientific community.

However, today the term “answer engine” has recently been adopted by some LLM-powered chatbots, too. Most notably, Perplexity AI has openly started billing itself as an answer engine: “An answer engine directly responds to your questions with detailed answers. Perplexity is an answer engine that searches the web and consults partners to provide you with up-to-date information,” a company FAQ page states.

And if Perplexity considers itself an answer engine, most other AI chatbots could, too. Indeed, I asked ChatGPT if it was an answer engine. Here was the response: “ChatGPT can be considered an answer engine in the sense that it is designed to respond to user queries with relevant information, explanations, and answers,” it stated. But then the chatbot went on to boast a little bit, too. “However, it is more than just an answer engine because it can engage in complex dialogues, provide creative writing, assist with problem-solving, and perform tasks that involve understanding and generating human-like text.”

The drawbacks of AI answer engines

The reason that classic answer engines like WolframAlpha have garnered so much respect in some communities is that their answers are pretty consistently reliable. That’s because WolframAlpha doesn’t generate its answers on the fly like LLM AI-powered chatbots do. Instead, it pulls its results directly from vast troves of objective knowledge or from computations that were purpose-built to answer the query.

Because chatbots like Perplexity and ChatGPT are powered by LLMs, they are prone to AI hallucinations because they generate the answer on the spot. AI hallucinations can lead to answers that sound authoritative but ultimately are incorrect or incomplete. 

Some AI chatbot answer engines have tried to alleviate concerns about returning incorrect answers by citing the original sources of information that their answers are based on. However, not all chatbots do this, and for those that do, it just means more work for those seeking the answers—people need to consult the external information sources to verify that the answer the answer engine is providing is correct.

Are answer engines a threat to search engines?

Answer engines might not all be entirely reliable today, but they may represent a threat to search engines like Google in the future, especially as the answer engines get better and provide reliable and accurate information.

Provided that the answer engine is reliable, its main benefit is that it saves time for the user. An answer engine’s answer is nearly instantaneous and put right in front of your eyes. Unlike with search engines, you don’t need to go anywhere else to find an answer your query––there’s no clicking around to other websites and reading multiple articles or sources of information and hoping you can find the information you want.

It’s still too early to say whether answer engines will usurp search engines as the main way people get information online, though without a doubt they will continue to become more popular as the general public becomes more trusting of AI.

---