The best AI chatbots in 2024

What to look for in an AI chatbot

Before we dive in, let’s talk a bit about how ChatGPT works. This will help you understand what’s interesting about each AI chatbot and use it to your advantage.

ChatGPT is an app created by OpenAI that lets users interact with its AI models: GPT-3.5 and GPT-4. The app takes the prompts you write and passes them to the AI model. This model runs the prompt through its systems and returns the results to the app, so you can read them in a conversational chatbot style.

While the app takes care of the features-for example, saving your conversation the AI model takes care of the actual interpretation of your input and the calculations to provide an answer. For more context, take a look at our breakdown of ChatGPT vs. GPT.

Many of the apps on this list are also powered by OpenAI’s GPT models. But even when that’s the case, the app developers can pass additional commands to configure how the model replies, so you may see different results when you try each chatbot, even though they’re running on the same engine. (Some apps on this list use non-GPT models, which are proprietary to the company that built the app.)

With that in mind, here’s what I was keeping an eye out for as I tested each AI chatbot online:

• Accuracy and responsiveness. Given that a lot of AI chatbots are using GPT right now, they’ll generally give you similar-ish content. But there were some duds. I didn’t include any platforms that couldn’t generate coherent language, that delivered wildly inaccurate information consistently, or that couldn’t move a conversation forward naturally.
• Easy access. There shouldn’t be any advanced setup or technical requirements to talk to a chatbot when you can do it quickly (and for free) with ChatGPT.
• Chat experience. There are lots of ways to interact with AI, but the staple of ChatGPT is interacting with the model through a chat window. With a few exceptions, each app on this list is a conversational AI chatbot (I’ll explain any exceptions as I get to them).
• Extra features. Anything on top of entering your prompt and receiving the output is welcome, from multi-language support to connecting directly to the internet.

Based on my research and experiences interacting with them, here are the best AI chatbots for you to try. Have fun- I know you will.

Merlin AI

ChatGPT Powered Chrome Extension

Model: GPT 4, GPT 3.5, GPT Vision, Claude 2, Claude Instant, LLama-2, GPT 3.5 16k, Code-LLaMa

Among many other features of Merlin, its chatbot is most saugth after. It offers a lot of different models that you can use according to your requirements.

Launched in December of 2022, it is extremely beneficial for everyone who uses it regardless of their profession. From lawyers and other professionals to university students, everyone can use it, and that too in as many ways as they want.

ChatGPT

The original AI chatbot

Model: OpenAI GPT-3.5 and 4, DALL·E 2 and 3

It is the original AI chatbot. After its late 2022 release, it took the world by storm in 2023 and put AI back in the news. It’s very simple to use: start a conversation by writing your prompt at the bottom of the screen. As the output comes back, a new entry is created on the left-side menu, so you can keep all your threads separate and come back to them later. If any of these is particularly interesting, you can share a link to it and show it to others (and they can start where you left off).

While GPT-3.5 is already impressive, GPT-4 is even better. It’s trained on a much larger dataset, making it even more flexible, and more accurate with its writing output, and it can even predict what happens next when given a still image.

A lot happened in the year since ChatGPT’s release. You can now do text-to-image prompting (and tweaking), directly in ChatGPT. You can even build your custom bot with unique instructions and data using. These bots can carry out straightforward tasks, simplify processes, or just do zany things for the fun of it. When you add Zapier’s AI Actions to it, you can access thousands of other apps from your GPT too.

Zapier is the leader in no-code automation-integrating with 6,000+ apps from partners like Google, Salesforce, and Microsoft. Build secure, automated systems for your business-critical workflows across your organization’s technology stack. Learn more.

Claude 2

Longest conversation memory

Model: Proprietary

Jumping from the bottom of this list in my last update straight to the second position now, meet. Anthropic’s chatbot aims to be helpful, harmless, and honest. The conversation flows naturally, with responses that are straight to the point, without lengthy introductions and conclusions like ChatGPT sometimes prefers using.

What’s impressive about it is the length of its context window. This is the fancy name for the conversation memory that chatbots are equipped with: it helps it remember what topic you’re talking about, past questions you asked, and how it answered. This feature is what makes answering questions like “Can you summarize everything we’ve talked about into a key takeaway?” possible to answer.

Where ChatGPT can only remember up to 12,000 words worth of conversation, Claude takes this to 75,000 words. Since there’s a file upload feature, this AI model is great for summarizing and asking questions based on long documents. Just make sure to keep the entire word count-questions and answers combined below the limit.

Microsoft Bing AI

Online search, text, and image generation

Model: OpenAI’s GPT and DALL·E models

Early in 2023, Microsoft upped its investment in OpenAI and started developing and rolling out AI features into its products. One of those was Bing, which now has an AI chatting experience that will help you search the web. Once you enter your prompt, it will search the internet for you, process the results, and present you with a reply containing the links it used as a base.

It can now show image results in the chat window, but it doesn’t pick up too well on the intent of image search: it usually prints out a list of image links instead of an image gallery. And if you want to re-read past conversations, you can do so by clicking on each one on the right side of the screen.

Moving away from search, Bing AI is now connected to OpenAI’s DALL·E 3, so I can get my daily fix of AI-generated golden retriever images. State that you want to generate an image while chatting, and enter the prompt. Once it’s out, you can ask to regenerate with new instructions, though in my experience, it doesn’t work well 100% of the time.

It’s fully integrated into Microsoft Edge, the native Windows browser. There are a few cool tools there. The best one lets you generate text in the sidebar and paste it with one click into any input field on a web page. This is useful for generating emails and pasting them into the compose window, for example.

Bing AI is still behaving strangely, sometimes ending conversations abruptly-still, it’s nothing like when it revealed its gaslighting skills. Don’t take it personally if it says it doesn’t want to continue the conversation. Hit New Thread, and keep going.

Here’s a full rundown of ChatGPT vs. Bing Chat.

Google Bard

Integrates with Google products

Model: PaLM 2

Google has been in the AI race for a long time, with a set of AI features already implemented across its product lineup. After an epic hiccup during the initial product demo, left behind the LaMDA model and now uses PaLM 2 to carry out your instructions. It got a lot of flak, but it’s really growing on me.

Bard can connect to the internet to find sources (even offering a handy button that lets you Google it yourself), which is a huge selling point. The search results can even show images directly on the chat window. It also lets you edit your prompt after you’ve sent it and offers up to three drafts of each output, so you can pick the best one. It can keep track of your conversation history, and you can share your conversations with others.

But here’s what I love: it integrates deeply with your Google account and with other Google products such as Hotels, Flights, and YouTube. Want to search your Gmail file jungle with one prompt? Do it. Summarize files inside your Google Drive? Yes, please. Check real-time flight and hotel prices as the AI builds your trip? Schedule that time off: it even gives you a packing list.

When compared with ChatGPT, Bard feels more conversational and less oriented toward text commands. To read more about their differences, here’s a direct comparison: ChatGPT vs. Bard.

Llama 2

Open license AI model

Model: Meta’s Llama 2

Meta joined the AI game in a unique way. Instead of building a commercial chatbot like all the competition, it decided to launch its own AI model with a generous open licensing framework. This means that you can use it and tweak it for free until you hit a revenue limit-but this limit is super high, designed to fence out the big tech competitors from ever using this LLM.

This easy licensing process almost makes it look like an open source model, but you can’t really peek into the details of Llama 2’s development, so it can’t really take that tag.

While the actual process of deploying a Llama 2 model is reserved for developers, you can try it on the website to get a feel for how it responds. The output feels more direct, less tuned than other chatbots, a vanilla model ready to be specialized and tweaked to unique needs. This app implementation offers a chat experience along with a few controls such as the system prompt, the temperature, and the context window-the bare minimum to explore the possibilities and limitations.

Llama’s promising future is tied to its open licensing terms. While using it isn’t as exciting as other options here, it’s definitely a model to keep an eye on. You may end up interacting with multiple implementations of it in other apps in the future.

Pi

For personal use

Model: Inflection-1

Pi: with a short name that stands for personal intelligence, Inflection AI’s new chatbot sets out to be engaging and supportive. It may not be as flexible as ChatGPT-it can’t write articles or search the internet-but it offers a fresh user experience that’s worthy of mention.

The app is minimalistic and filled with loads of cute details and animations. You’ll find that Pi isn’t tuned for long answers. Instead, it prefers shorter bursts of conversation and loves asking questions. It wants you to share your day, mention difficulties you’re having, or talk through problems in your life. It’s friendly, and while vague at times, it always has nice things to say.

In addition to the standard chat mode, you can switch to SupportPi to talk things through, get advice, or just as a “sounding board” for stuff on your mind. You can combine these models with the Discover section, where you can choose a conversation type, with options such as “practice a big conversation,” “get motivated,” or “just vent.”

No big productivity use cases, sure. But maybe there’s no productivity if you don’t keep your personal stuff in order, and that’s exactly Pi’s angle.

It’s free to use at the moment, so you can jump right in. To keep track of your conversation history, you’ll have to provide your name and phone number. This way, Pi will be able to text you from time to time to ask how things are going, a nice reminder to check in and catch up.

Poe

For trying multiple AI models

Model: OpenAI’s GPT and DALL·E models, StableDiffusionXL, Claude 1 and 2, PaLM, Llama (multiple versions), Solar 70B, Mistral 7B

ChatGPT, Claude 2, and StableDiffusion XL walk into a bar. That bar is called Poe, which acts as a kind of AI model aggregator. It’s the best if you want to try out the top models on the market right now for a single-pack price of $20 per month. There will be message limits for the highest quality models, but it’s still better than subscribing to each individual one if you want to explore.

When you’re done chatting, you might want to try creating your own bot. You can set its system prompt, add a knowledge base, and configure a greeting message. This feels like one of OpenAI’s GPTs, with two key differences: you can pick which AI model you want to use as a base, and you can monetize it by setting up your creator account. As I write this, you get $20 whenever a user subscribes to Poe by using one of your chatbots. If you create a good one, maybe your Poe subscription can pay itself multiple times over.

Perplexity

For internet deep dives

Model: OpenAI GPT-3 and GPT-4

Perplexity is another take on AI internet-connected chatbots for handling more information and longer, less organized searches. Here’s why: when getting the output, you’ll see a list of all the sources below it. Then, you can add a new prompt to continue that search, or pick one of the suggested related search terms. All the results will stack on the bottom, so you can scroll up or down to read everything.

You can tick Copilot in the search bar to get some help with product recommendations, best healthy recipes, or travel tips, for example. Once you enter your prompt, Perplexity will ask you a set of qualifying questions to home in on your intent. The resulting output summarizes all the key information, acting as a good starting point for a deep dive.

When you share your chats with others, they can continue the conversation you started without limitations. On your end, you can see the views for shared conversations, likes, and follow-up questions, making the experience more interactive.

A new feature, Discover, rounds up popular searches into one short, snappy article. For example, “the most streamed artist in 2023” brings together the sources that prove that Taylor Swift attained 26.1 billion streams, coupled with a short AI-generated summary unpacking the important information about the topic.

Now better prepared for advanced tasks like solving equations and processing real-time data by leveraging Wolfram Alpha, Perplexity is positioning itself more and more as a powerful and straightforward search tool.